ord_schema.artifacts#
ord_schema.artifacts.arrivals#
Each reaction’s first commit to ord-data, one artifact per dataset.
When a reaction entered ORD’s public corpus is not in its record, so it is not in the
projection either, which restates the record and nothing else. It is read from
ord-data’s git history into an index (ord_schema.artifacts.history), and this
artifact carries the index’s answer for one dataset: one row per reaction its projection
holds, keyed by reaction_id, with the commit, its time as git recorded it, and the
dataset whose file first held the reaction.
It derives from the projection, as the structures artifact does, and passes the projection’s source hash through, so it pairs with its projection the way every other artifact does. Its currency has a second input the stamps every artifact carries cannot see: the index. It is stamped with the index’s walk version and head, and stays current while that head is on the chain a later index walked, since a walk only appends. A reaction the index lacks refuses the derivation rather than reading as one ord-data never held.
A query reads this artifact only when it names first_commit; see
ord_schema.search.query. Every other query reads the projection alone and costs
what it did before the artifact existed.
- ord_schema.artifacts.arrivals.is_current(path: str | PathLike[str], source_md5: str, index: Index) bool#
Returns whether
pathrestatessource_md5underindexby this library.- Parameters:
path – The artifact. Need not exist.
source_md5 – Hash of the source dataset it is expected to restate.
index – The index of first commits it is expected to agree with.
- Returns:
Whether its stamps are current and name a head on the index’s chain.
- ord_schema.artifacts.arrivals.write_arrivals(source: str | PathLike[str], output: str | PathLike[str], *, index: Index, source_md5: str | None = None, source_dataset_id: str | None = None, compression: str = 'zstd', row_group_size: int = 100000) int#
Derives an arrivals artifact from a projection and writes it.
Only the projection’s
reaction_idcolumn is read. The output is published atomically, so a failure leaves any existing artifact untouched.- Parameters:
source – Path to the projection whose reactions to look up.
output – Path to write the artifact to.
index – The index of first commits, walked to at least the commit the projection’s source was read from.
source_md5 – Hash of the source dataset to stamp, if the caller already read one. Taken from the projection’s stamps when omitted.
source_dataset_id – Source dataset ID to stamp, if the caller already read one. Taken from the projection’s stamps when omitted.
compression – Parquet codec, any name
pq.ParquetWriteraccepts.row_group_size – Rows per output row group. Large, because the artifact has few leaves and a footer per row group is what a reader’s cache holds.
- Returns:
one per reaction in the projection.
- Return type:
Number of rows written
- Raises:
ValueError – If
row_group_sizeis less than one, refused before the source is read; ifsourceis not a current projection; or if the index lacks one of its reactions, named with the source and the index’s head.
ord_schema.artifacts.base#
Shared mechanics for derived artifacts: footer stamps and the derive-a-tree driver.
A derived artifact restates a source dataset in a queryable shape. It adds no information, so it is wrong if it disagrees with its source and stale if its source has moved on. These stamps make both conditions detectable without reading column data.
More than one artifact is expected – a projection carrying every field, structures pulled out of it for search, and pivoted indexes over the paths that turn out to be hot – so the parts that do not vary between them live here: the stamps, the output-path mapping, and the driver that walks a glob, refuses to write over what it reads, and skips what is current. An artifact module supplies only its schema, its projection, and its name.
An artifact may derive from another rather than from a source dataset, which is how the structures artifact reaches the projection’s columns without recomputing them. Only the reading differs: the stamps describe the originating dataset either way, so a chain is invisible to a consumer checking whether a file is current.
Every artifact carries these keys, in the ord. namespace that ord_schema.parquet
already uses for source datasets:
ord.artifact– which artifact this is, so a reader can tell one from another without inspecting the schema.ord.artifact_schema_version– the version of the schema the artifact was derived to: its columns and what each holds. One for every artifact, because they are published and fetched as one build.ord.source_md5– the content of the source dataset it restates. An artifact derived from another artifact passes this through rather than hashing its parent, so every artifact names the dataset it reflects however many derivations away it sits, and one comparison answers “is this current for that dataset?”ord.rdkit_version– the RDKit that derived it. Every artifact leans on RDKit – canonical SMILES in the projection, fingerprints in the structures artifact – and RDKit releases have changed both canonicalization and pattern fingerprints, either of which silently breaks an artifact whose consumers run the new RDKit against files built by the old. Optional to read so that a file stamped before the key existed still loads; absent reads as stale, which rebuilds it.ord.source_dataset_id– the source’s ID, written only when the source records one, and so the only key not required to read stamps back.
is_current requires the artifact name, the source content, the artifact schema
version, and the RDKit version to all match, since any of the four changing can
change a value or mean the file answers a different question.
No key names the ord-schema release. A release need not change any artifact, and a
development build reports the last release’s number, so a release stamp says nothing
about whether a file is current – ord.artifact_schema_version is what says the
columns moved – while making a rebuild of an unchanged artifact write different
bytes, which an upload then carries as a change.
- class ord_schema.artifacts.base.ArtifactWriter(*args, **kwargs)#
Bases:
ProtocolWrites one derived artifact from the Parquet file it derives from.
The parent is passed as a path rather than an open reader, because parents are not all the same kind of file: a projection reads a source dataset, and the structures artifact reads a projection. Each writer opens what it knows how to read.
The first two arguments are positional-only, so an implementation is free to name them for its own artifact.
- class ord_schema.artifacts.base.Stamps(artifact: str, source_dataset_id: str | None, source_md5: str, artifact_schema_version: str, rdkit_version: str | None = None)#
Bases:
objectFooter stamps recording what an artifact was derived from, and by what.
rdkit_versionis None only when read from a file stamped before the key existed; such a file reads as stale.- artifact: str#
- artifact_schema_version: str#
- rdkit_version: str | None = None#
- source_dataset_id: str | None#
- source_md5: str#
- ord_schema.artifacts.base.current_stamps(artifact: str, source_dataset_id: str | None, source_md5: str) Stamps#
Returns the stamps to write for an artifact derived by the current library.
- Parameters:
artifact – Artifact name, e.g.
"structures"or"projection".source_dataset_id – Source dataset ID, or None if the source records none.
source_md5 – Hash of the source, from
parquet.DatasetView.md5.
- Returns:
Stamps carrying the artifact schema version and the RDKit version in use.
- ord_schema.artifacts.base.derive_tree(input_pattern: str, output_dir: str, *, artifact: str, write: ArtifactWriter, schema: Schema | None = None, force: bool = False, parent_artifact: str | None = None, root: str | PathLike[str] | None = None, current: Callable[[Path], bool] | None = None) tuple[int, int, int]#
Derives one artifact per file matching
input_pattern.Outputs mirror the inputs’ directory layout beneath
output_dir. Matches that are not the kind of parent this derivation reads are ignored, so a recursive pattern reaching the output tree stays re-runnable. Artifacts whose footer already records the current source content and versions are skipped, so re-running is cheap.- Parameters:
input_pattern – Glob matching the files to derive from.
output_dir – Directory to write artifacts beneath.
artifact – Artifact name, used for the staleness check and the footer stamp.
write – Writer taking
(parent, output, source_md5=..., source_dataset_id=...)and returning rows.schema – The columns
writeproduces, when the caller has them. Given, an artifact lacking one is derived again rather than skipped.force – Rewrite artifacts that are already current.
parent_artifact – Artifact the parents hold, for a derivation that reads another artifact rather than a source dataset. None means the parents are source datasets.
root – The directory outputs mirror from, for a caller that has one rather than a pattern. Defaults to
input_pattern’s wildcard-free leading directories, which is what a caller who wrote the pattern means by it.current – A further test an existing artifact has to pass to be skipped, for a writer reading an input the stamps do not describe. Given the artifact’s path; False derives it again.
- Returns:
artifacts derived, artifacts already current, and matches that were not the expected kind of parent.
- Return type:
(written, skipped, ignored)counts- Raises:
ValueError – If the input pattern matches nothing, if any match cannot be read as Parquet, if any destination would land on a parent, or if any destination already holds a file this library did not write.
- ord_schema.artifacts.base.glob_root(pattern: str) PurePath#
Returns the leading directories of a glob pattern that contain no wildcards.
Output paths are built relative to this, so
data/*/x.parquetunderoutput_dir=structureslands atstructures/<subdir>/x.parquet. A pattern naming a single file has no wildcard to stop at, so its own last component is the file rather than a directory and the root is the directory holding it.A bracket holding only
*,?, or[matches that character alone, which is howglob.escapewrites a directory name that contains one, so such a bracket is read as the character:glob.escape("data[1]") + "/*/x.parquet"has the rootdata[1].- Parameters:
pattern – Glob pattern, with
*,?, or[marking the first wildcard.- Returns:
The wildcard-free leading directories, unescaped, or
.when the pattern names no directory or begins with a wildcard.
- ord_schema.artifacts.base.is_artifact(path: str | PathLike[str]) bool#
Returns whether
pathcarries derived-artifact stamps.- Parameters:
path – Path to a Parquet file.
- Returns:
True if
pathrecords artifact stamps, False if it is readable Parquet that does not.- Raises:
ValueError – If
pathcannot be read as Parquet at all. This predicate decides whether to treat a file as a source, so answering False for a truncated one would feed it to a reader that fails later without naming it.
- ord_schema.artifacts.base.is_current(path: str | PathLike[str], artifact: str, source_md5: str, schema: Schema | None = None) bool#
Returns whether
pathholdsartifactderived fromsource_md5by us.All of the source content, the artifact schema version, and the RDKit version must match; any of them changing can change a value. A missing, unreadable, or wrong-artifact file is not current.
- Parameters:
path – Path to check. Need not exist.
artifact – Artifact name the file is expected to hold.
source_md5 – Hash the file is expected to restate.
schema – The columns the artifact carries, when the caller has them. Given, a file lacking one is not current, which is what makes a column added without a version bump derive again instead of being skipped.
- Returns:
Whether the file is one this library would write today.
- ord_schema.artifacts.base.left_behind(directory: str | PathLike[str], artifact: str, schemas: Mapping[str, Schema], artifact_paths: Callable[[str | PathLike[str], str], list[Path]], derived: Collection[str]) dict[str, str]#
Returns the keys already under
directorythat a run left short of a column.An artifact split across a tree – one subdirectory per pivot level, per indexed path – is derived a subset at a time, which is the cheaper run and often the only one wanted. The tree that run writes into holds whatever earlier runs put there, and those files stamp current forever, so a schema that grew a column leaves every key nobody named short of it. A corpus reads the whole tree rather than the part the run touched.
Columns only, matching what a reader refuses whatever its policy: an artifact short a column fails every corpus that opens it, while one whose stamps are merely old is refused where
require_currentsays so and read otherwise. A build that stopped on the second would decide that question on the reader’s behalf, and leave no way to finish a subset run against a tree an older library wrote.- Parameters:
directory – The directory the keys sit under.
artifact – Artifact name the files are expected to hold; anything else is left alone, since a tree is not a build’s to police.
schemas – Every key this artifact covers, mapped to the columns it carries.
artifact_paths – Lists one key’s artifacts exactly as a reader finds them, taking
directoryand the key.derived – The keys the run covered, which are current by construction.
- Returns:
One entry per uncovered key whose artifacts no longer match its schema, naming the first artifact that does not and what it lacks. Keys with no artifacts are not included: there is nothing there to be out of date.
- ord_schema.artifacts.base.load_stamps(path: str | PathLike[str]) Stamps#
Reads an artifact’s footer stamps.
- Parameters:
path – Path to a derived artifact Parquet file.
- Returns:
The stamps recorded when the artifact was written.
- Raises:
ValueError – If
pathis missing a required stamp, e.g. because it is a source dataset rather than a derived artifact.
- ord_schema.artifacts.base.missing_columns(path: str | PathLike[str], schema: Schema) list[str]#
Returns the columns
schemadeclares thatpathdoes not carry.Stamps say nothing about a file’s columns, so this is what sees a column added to an artifact’s definition: a file written before it stamps exactly as a file written after it, and the difference shows up only in the schema.
Compared down through structs, lists and maps, not across the top level alone. The projection is one struct per message and nearly every field it has is nested, so a top-level comparison would call a file current that is missing everything added since it was written – and a reader binding the new column would fail on a file nothing marked stale.
A leaf whose type changed counts as missing too, because a path comparison alone cannot see it: widening a column or moving a timestamp from seconds to microseconds leaves every name where it was, so a file holding the old type would read as current and keep serving values the current definition does not describe.
- Parameters:
path – Path to check. Need not exist.
schema – The columns this library’s version of the artifact carries.
- Returns:
The declared paths the file lacks, dotted from the top level and in declaration order. A path present with the wrong type carries the type it should hold, so the message says what changed rather than naming a column the file visibly has. A file that cannot be read at all lacks all of them.
- ord_schema.artifacts.base.output_path(source: str, pattern: str, output_dir: str, *, root: str | PathLike[str] | None = None) Path#
Maps a source path to its artifact path under
output_dir.- Parameters:
source – Path the pattern matched.
pattern – The glob it matched, whose wildcard-free prefix the output mirrors from when
rootis not given.output_dir – Directory to write beneath.
root – The directory to mirror from, for a caller that has one; otherwise
glob_root(pattern).
- Returns:
Where
source’s artifact goes.
- ord_schema.artifacts.base.read_stamps(path: str | PathLike[str]) Stamps | None#
Returns
path’s artifact stamps, or None if it carries none.One footer read where a predicate followed by
load_stampswould be two, which matters toderive_tree: it needs the answer and the stamps behind it for every match of a corpus-wide glob.- Parameters:
path – Path to a Parquet file.
- Returns:
The stamps, or None if
pathis readable Parquet that records none.- Raises:
ValueError – If
pathcannot be read as Parquet at all. This decides whether a glob match is a candidate parent, so answering None for a truncated one would either feed it to a reader that fails later without naming it, or drop it from the run without a word.
- ord_schema.artifacts.base.stamps_are_current(value: Stamps, artifact: str) bool#
Returns whether
valuerecordsartifactas this library defines it.The source content is deliberately not compared: a caller holding the stamps of a parent asks whether the file was written by the current definition, and the hash it carries is the answer to a different question.
- Parameters:
value – Stamps read from an artifact.
artifact – Artifact name the file is expected to hold.
- Returns:
Whether the artifact name, the artifact schema version, and the RDKit version all match. A file with no RDKit stamp is not current.
- ord_schema.artifacts.base.stamps_from_metadata(raw: Mapping[bytes, bytes]) Stamps#
Reads stamps from Parquet footer key-value metadata, for a file read elsewhere.
- Parameters:
raw – The footer’s key-value metadata.
- Returns:
The stamps recorded when the artifact was written.
- Raises:
ValueError – If
rawis missing a required stamp, e.g. because it is a source dataset’s footer rather than a derived artifact’s.
ord_schema.artifacts.history#
Where and when each reaction first reached ord-data, read from its git history.
A record says when its depositor created it and when the submission pipeline modified
it, but not when it entered ORD’s public corpus: the pipeline stamps its own
record_modified event when it processes a submission, which can precede the merge
by days. ord-data’s history holds the answer, so this module walks it into an index,
from which ord_schema.artifacts.arrivals derives one artifact per dataset.
The fact stays out of the records, which hold what the depositor and the pipeline
wrote, and out of the projection, which restates them.
A reaction’s first commit is the first commit on main, followed by first parent,
whose version of a dataset file under data/ holds the reaction’s ID:
First parent. Early submissions merged in two steps, into a submission branch and then into
main, and the reaction entered the corpus at the second.Every version of every file. A dataset file’s own first commit is not the answer: every dataset has been rewritten into another format at least once, and USPTO has been resharded more than once, so every version of every file the corpus has held is read, in whichever format it was written.
Per reaction. A dataset updated later holds reactions that arrived at different commits.
The committer date. When the commit was made on
main; for a merged pull request, the merge.timestampholds its wall clock as git recorded it andutc_offset_minutesthe offset git stored with it, the pair every projected time is read as, so nothing git records is dropped.The dataset the file recorded, which need not be the one holding the reaction now.
A reaction whose ID a migration reassigned first appears under its current ID at that migration, since nothing in either file links the two IDs. ord-data holds two such migrations: the Nanomole flow dataset merged in ord-data#41, given new IDs in #73; and the USPTO grants merged in #66 through #69, given new IDs by the reshard in #151, #153, and #155.
The index holds every reaction ID main has held, not only the current ones, and
records the chain of commits it walked. History only grows, so a later walk extends an
earlier index from its head rather than starting over, and a reaction that leaves the
corpus and returns keeps the commit it first arrived at.
Only the two IDs are read from each version, through a message declaring nothing else
rather than today’s Dataset: a version written under an older schema need not parse
under today’s, and a full parse decodes every field of every reaction to reach one.
Neither ID has changed its field number or name since before ord-data held a dataset.
The walk needs the whole history as it was committed. A shallow clone presents its oldest commit as a root, and a grafts file gives a commit whatever parents it names, so either would place every reaction such a commit holds at that commit; both refuse the walk. Replace refs, which do the same, are ignored.
Dataset files live in git LFS, and the walk reads only objects already fetched. One that is not refuses the walk before any dataset file is read, naming the command that fetches it, rather than fetching from whichever store the checkout points at.
- class ord_schema.artifacts.history.Index(table: Table, chain: Sequence[str])#
Bases:
objectEvery reaction ID ord-data’s
mainhas held, and its first commit.- Variables:
table – One row per reaction ID, in
INDEX_SCHEMA, in the order the walk met them.chain – Every commit walked, oldest first; the last is the head.
- covers(path: str | PathLike[str]) bool#
Returns whether
pathwas derived with this index or one it extends.Every first commit an earlier index records is one this index records too, since a walk only appends, so an artifact stamped with any head on this chain holds what this index would give it. One stamped with a head off the chain was derived from a history that has since been rewritten.
- Parameters:
path – An arrivals artifact. Need not exist.
- Returns:
Whether its stamp names this walk version and a commit on this chain. A file stamped by no index, or unreadable, is not covered.
- property head: str#
The last commit walked.
- lookup(reaction_ids: Collection[str]) dict[str, dict[str, Any]]#
Returns the first commit of each reaction ID the index holds.
- Parameters:
reaction_ids – The IDs to look up.
- Returns:
Rows in
TYPE, keyed by reaction ID. An ID the index does not hold is absent rather than mapped to None.
- stamps() dict[str, str]#
Returns the footer keys an artifact derived with this index carries.
- ord_schema.artifacts.history.commit_time(seconds: int, offset: str) tuple[datetime, int]#
Returns the wall clock and UTC offset of a time git stores as seconds and offset.
- Parameters:
seconds – Seconds since the epoch, as
--date=rawprints them.offset – The offset git recorded, as
-0400or+0130.
- Returns:
The wall clock at that offset, naive, and the offset in minutes east of UTC. Together they are everything git records, so the instant stays recoverable.
- Raises:
ValueError – If
offsetis not a signed four-digit offset.
- ord_schema.artifacts.history.read_ids(path: str, content: bytes | Path) tuple[str, list[str]]#
Returns a dataset file’s dataset ID and the reaction IDs it holds.
- Parameters:
path – The file’s name, whose suffix says how it is written.
content – Its bytes, or the local file holding them.
- Returns:
The dataset ID the file records, empty where it records none, and every reaction ID in file order, empty ones included.
- Raises:
ValueError – If the suffix is none a dataset file has been written with.
- ord_schema.artifacts.history.read_index(path: str | PathLike[str]) Index#
Reads an index written by
write_index.- Parameters:
path – The index file.
- Returns:
The index.
- Raises:
ValueError – If
pathis not an index, or was walked under anotherVERSION, whose first commits this library does not stand behind.
- ord_schema.artifacts.history.read_stamp(path: str | PathLike[str]) tuple[str, str] | None#
Returns the walk version and index head an artifact was derived with.
- Parameters:
path – An arrivals artifact.
- Returns:
derived without an index, missing, or not Parquet.
- Return type:
(version, head), or None wherepathcarries no index stamp
- ord_schema.artifacts.history.walk(repository: str | PathLike[str], *, revision: str = 'main', previous: Index | None = None) Index#
Walks ord-data’s first-parent history into an index of first commits.
- Parameters:
repository – An ord-data checkout holding every LFS object the walk reads.
revision – The branch or commit to walk to, followed by first parent from the root.
previous – An index to extend, walked to a commit on the same chain. Only the commits after its head are read.
- Returns:
An index of every reaction ID the chain’s dataset files have held, each at the first commit that held it, walked to
revision.- Raises:
ValueError – If the checkout’s history is shallow or grafted, if
previouswas walked along a chain that is not the start of this one, if an LFS object is missing, if a file records no dataset ID, or if a reaction first appears in two datasets in one commit, which would leave its first dataset unknowable.
ord_schema.artifacts.occurrences#
One row per structure occurrence, for the paths a structure query can be routed to.
A quantifier whose body is a structure predicate on the element’s own smiles is
answered by a semi-join rather than by reading the projection: which reactions hold a
match is a separate cost from the chemistry, and the larger one. The relation that
answers it is one row per occurrence, carrying the reaction and the structure and
nothing of the element. Whether the molecule went in or came out is the indexed path,
which names the directory a file sits under. Every element field, the role among them,
is in the pivot this is derived from, and a query binding one is answered there: a row
is what a semi-join reads, and every column on it is paid for by every occurrence in the
corpus.
Dataset-local IDs, always. An occurrence names its structure by the structure_id
its own dataset assigned, never by a corpus-wide one. A corpus-wide ID is that plus an
offset that is a running total over the corpus’s datasets, so it belongs to no artifact:
the same rows read beside a different set of datasets need a different offset, and an ID
baked in here would stay in range while naming another dataset’s molecule – passing on
the whole corpus and failing only on a subset of it. The corpus joins the offset on at
open, keyed by the source dataset every artifact names; see
ord_schema.search.execute.
That rule is what lets this be an artifact at all rather than a relation assembled at
open. Assembled, it is a table held in memory for the life of the corpus, and unnesting
it from the projections has a memory floor and spills heavily to get there, before the
first query can be answered. The same rows derived here are a small fraction of that as
Parquet, and a corpus given a directory covering every indexed path reads them as a
view instead of building anything, which is faster to open and holds far less than the
table. See ord_schema.search.execute.Corpus and its occurrences_dir, and
the search README for the measurements.
Derived from the pivot over the level an indexed path ranges within, which already
holds one row per element: this artifact is that projected down to two columns, so the
two cannot disagree about which elements exist. A path descending from its level through
singular struct fields – an authentic standard is one compound per measurement – is
read from that level’s pivot through the remainder; see ord_schema.artifacts.pivot.
- ord_schema.artifacts.occurrences.artifact_paths(occurrences_dir: str | PathLike[str], path: str) list[Path]#
Returns the artifacts a reader finds for one indexed path, in a stable order.
The one definition of what a path’s artifacts are, for the same reason
pivot.artifact_pathsis: a build judging its own output by a wider rule than a reader applies calls a tree healthy that no corpus can read.Recursive, because the tree mirrors the projections it descends from: a sharded corpus puts its files under
<path>/<shard>/rather than directly under the path.- Parameters:
occurrences_dir – The directory the indexed paths sit under.
path – The indexed path, as the query grammar names it.
- Returns:
The paths, sorted. Empty if the path has no directory at all.
- ord_schema.artifacts.occurrences.indexed_paths(schema: ~pyarrow.lib.Schema = smiles: string identifiers: list<item: struct<type: string, details: string, value: string, is_mapped: bool>> child 0, item: struct<type: string, details: string, value: string, is_mapped: bool> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,REACTION_SMILES,REACTION_CXSMILES,RDF' + 24 child 1, details: string child 2, value: string child 3, is_mapped: bool inputs: map<string, struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: (... 2062 chars omitted) child 0, entries: struct<key: string not null, value: struct<components: list<item: struct<smiles: string, structure_i (... 2086 chars omitted) not null child 0, key: string not null child 1, value: struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: s (... 2049 chars omitted) child 0, components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 1105 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC' + 117 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORK' + 70 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED,DRI' + 14 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,N' + 98 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the times' + 4 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 1, crude_components: list<item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: stru (... 292 chars omitted) child 0, item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: struct<mass_gra (... 280 chars omitted) child 0, reaction_id: string child 1, includes_workup: bool child 2, has_derived_amount: bool child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 2, addition_order: int32 child 3, addition_time_seconds: double child 4, addition_time_precision_seconds: double child 5, addition_speed: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,ALL_AT_ONCE,FAST,SLOW,DROPWISE,CONTINUOU' + 13 child 1, details: string child 6, addition_duration_seconds: double child 7, addition_duration_precision_seconds: double child 8, flow_rate_milliliters_per_minute: double child 9, flow_rate_precision_milliliters_per_minute: double child 10, addition_device: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,SYRINGE,CANNULA,ADDITION_FUN' + 83 child 1, details: string child 11, addition_temperature_kelvin: double child 12, addition_temperature_precision_kelvin: double child 13, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SOLI' + 38 child 1, details: string setup: struct<vessel: struct<type: string, details: string, material: struct<type: string, details: string> (... 524 chars omitted) child 0, vessel: struct<type: string, details: string, material: struct<type: string, details: string>, preparations: (... 234 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,ROUND_BOTTOM_FLASK,VIAL,WELL_PLATE,MI' + 130 child 1, details: string child 2, material: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,GLASS,POLYPROPYLENE,PLASTIC,METAL,Q' + 5 child 1, details: string child 3, preparations: list<item: struct<type: string, details: string>> child 0, item: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,OVEN_DRIED,FLAME_DRIED,EVACU' + 22 child 1, details: string child 4, attachments: list<item: struct<type: string, details: string>> child 0, item: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,NONE,CUSTOM,SEPTUM,CAP,MAT,REFLUX_CONDEN' + 146 child 1, details: string child 5, volume_liters: double child 6, volume_precision_liters: double child 7, vessel_id: string child 8, position: string child 9, row: string child 10, col: string child 1, is_automated: bool child 2, automation_platform: string child 3, automation_code: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 4, environment: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,FUME_HOOD,BENCH_TOP,GLOVE_BOX,GLOVE_B' + 2 child 1, details: string conditions: struct<temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, (... 1761 chars omitted) child 0, temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, setpoint_precision_k (... 199 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,OIL_BATH,WATER_BATH,SAND_BA' + 77 child 1, details: string child 1, setpoint_kelvin: double child 2, setpoint_precision_kelvin: double child 3, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 69 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, temperat (... 57 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,THERMOCOUPLE_INTERNAL,THERMOCOUPL' + 19 child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, temperature_kelvin: double child 5, temperature_precision_kelvin: double child 1, pressure: struct<control: struct<type: string, details: string>, setpoint_kilopascals: double, setpoint_precis (... 264 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,SLIGHT_POSITIVE,SEALED,PRES' + 7 child 1, details: string child 1, setpoint_kilopascals: double child 2, setpoint_precision_kilopascals: double child 3, atmosphere: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AIR,NITROGEN,ARGON,OXYGEN,HYDROGEN,' + 71 child 1, details: string child 4, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 73 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, pressure (... 61 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,PRESSURE_TRANSDUCER' child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, pressure_kilopascals: double child 5, pressure_precision_kilopascals: double child 2, stirring: struct<type: string, details: string, rate: struct<type: string, details: string, rpm: int32>> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,STIR_BAR,OVERHEAD_MIXER,AGITATIO' + 25 child 1, details: string child 2, rate: struct<type: string, details: string, rpm: int32> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,HIGH,MEDIUM,LOW' child 1, details: string child 2, rpm: int32 child 3, illumination: struct<type: string, details: string, peak_wavelength_nanometers: double, peak_wavelength_precision_ (... 124 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,DARK,LED,HALOGEN_LAMP,DEUTERI' + 38 child 1, details: string child 2, peak_wavelength_nanometers: double child 3, peak_wavelength_precision_nanometers: double child 4, color: string child 5, distance_to_vessel_centimeters: double child 6, distance_to_vessel_precision_centimeters: double child 4, electrochemistry: struct<type: string, details: string, current_amperes: double, current_precision_amperes: double, vo (... 446 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,CONSTANT_CURRENT,CONSTANT_VOLTAGE' child 1, details: string child 2, current_amperes: double child 3, current_precision_amperes: double child 4, voltage_volts: double child 5, voltage_precision_volts: double child 6, anode_material: string child 7, cathode_material: string child 8, electrode_separation_centimeters: double child 9, electrode_separation_precision_centimeters: double child 10, measurements: list<item: struct<time_seconds: double, time_precision_seconds: double, current_amperes: double, cur (... 88 chars omitted) child 0, item: struct<time_seconds: double, time_precision_seconds: double, current_amperes: double, current_precis (... 76 chars omitted) child 0, time_seconds: double child 1, time_precision_seconds: double child 2, current_amperes: double child 3, current_precision_amperes: double child 4, voltage_volts: double child 5, voltage_precision_volts: double child 11, cell: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,DIVIDED_CELL,UNDIVIDED_CELL' child 1, details: string child 5, flow: struct<type: string, details: string, pump_type: string, tubing: struct<type: string, details: strin (... 73 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,PLUG_FLOW_REACTOR,CONTINUOUS_STIRRED_' + 31 child 1, details: string child 2, pump_type: string child 3, tubing: struct<type: string, details: string, diameter_centimeters: double, diameter_precision_centimeters: (... 7 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,STEEL,COPPER,PFA,FEP,TEFLONAF,PTFE,' + 25 child 1, details: string child 2, diameter_centimeters: double child 3, diameter_precision_centimeters: double child 6, reflux: bool child 7, ph: float child 8, conditions_are_dynamic: bool child 9, details: string notes: struct<is_heterogeneous: bool, forms_precipitate: bool, is_exothermic: bool, offgasses: bool, is_sen (... 133 chars omitted) child 0, is_heterogeneous: bool child 1, forms_precipitate: bool child 2, is_exothermic: bool child 3, offgasses: bool child 4, is_sensitive_to_moisture: bool child 5, is_sensitive_to_oxygen: bool child 6, is_sensitive_to_light: bool child 7, safety_notes: string child 8, procedure_details: string observations: list<item: struct<time_seconds: double, time_precision_seconds: double, comment: string, image: stru (... 139 chars omitted) child 0, item: struct<time_seconds: double, time_precision_seconds: double, comment: string, image: struct<float_va (... 127 chars omitted) child 0, time_seconds: double child 1, time_precision_seconds: double child 2, comment: string child 3, image: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string workups: list<item: struct<type: string, details: string, duration_seconds: double, duration_precision_second (... 2903 chars omitted) child 0, item: struct<type: string, details: string, duration_seconds: double, duration_precision_seconds: double, (... 2891 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,ADDITION,ALIQUOT,TEMPERATURE,CONCENTR' + 166 child 1, details: string child 2, duration_seconds: double child 3, duration_precision_seconds: double child 4, input: struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: s (... 2049 chars omitted) child 0, components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 1105 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC' + 117 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORK' + 70 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED,DRI' + 14 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,N' + 98 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the times' + 4 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 1, crude_components: list<item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: stru (... 292 chars omitted) child 0, item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: struct<mass_gra (... 280 chars omitted) child 0, reaction_id: string child 1, includes_workup: bool child 2, has_derived_amount: bool child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 2, addition_order: int32 child 3, addition_time_seconds: double child 4, addition_time_precision_seconds: double child 5, addition_speed: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,ALL_AT_ONCE,FAST,SLOW,DROPWISE,CONTINUOU' + 13 child 1, details: string child 6, addition_duration_seconds: double child 7, addition_duration_precision_seconds: double child 8, flow_rate_milliliters_per_minute: double child 9, flow_rate_precision_milliliters_per_minute: double child 10, addition_device: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,SYRINGE,CANNULA,ADDITION_FUN' + 83 child 1, details: string child 11, addition_temperature_kelvin: double child 12, addition_temperature_precision_kelvin: double child 13, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SOLI' + 38 child 1, details: string child 5, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATED' child 1, details: string child 7, volume_includes_solutes: bool child 6, temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, setpoint_precision_k (... 199 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,OIL_BATH,WATER_BATH,SAND_' + 79 child 1, details: string child 1, setpoint_kelvin: double child 2, setpoint_precision_kelvin: double child 3, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 69 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, temperat (... 57 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,THERMOCOUPLE_INTERNAL,THERMOCOU' + 21 child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, temperature_kelvin: double child 5, temperature_precision_kelvin: double child 7, keep_phase: string child 8, stirring: struct<type: string, details: string, rate: struct<type: string, details: string, rpm: int32>> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,STIR_BAR,OVERHEAD_MIXER,AGITAT' + 27 child 1, details: string child 2, rate: struct<type: string, details: string, rpm: int32> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,HIGH,MEDIUM,LOW' child 1, details: string child 2, rpm: int32 child 9, target_ph: float child 10, is_automated: bool outcomes: list<item: struct<reaction_time_seconds: double, reaction_time_precision_seconds: double, conversion (... 2983 chars omitted) child 0, item: struct<reaction_time_seconds: double, reaction_time_precision_seconds: double, conversion: struct<va (... 2971 chars omitted) child 0, reaction_time_seconds: double child 1, reaction_time_precision_seconds: double child 2, conversion: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 3, products: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 2422 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 2410 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC_N' + 115 child 1, details: string child 2, value: string child 3, is_desired_product: bool child 4, measurements: list<item: struct<analysis_key: string, type: string, details: string, uses_internal_standard: bool, (... 1984 chars omitted) child 0, item: struct<analysis_key: string, type: string, details: string, uses_internal_standard: bool, is_normali (... 1972 chars omitted) child 0, analysis_key: string child 1, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,IDENTITY,YIELD,SELECTIVITY,PU' + 33 child 2, details: string child 3, uses_internal_standard: bool child 4, is_normalized: bool child 5, uses_authentic_standard: bool child 6, authentic_standard: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,I' + 121 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TIT' + 5 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,' + 74 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED' + 18 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_1' + 102 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the t' + 8 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPH' + 46 child 1, details: string child 7, percentage: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 8, float_value: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 9, string_value: string child 10, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRA' + 3 child 1, details: string child 7, volume_includes_solutes: bool child 11, retention_time_seconds: double child 12, retention_time_precision_seconds: double child 13, mass_spec_details: struct<type: string, details: string, tic_minimum_mz: float, tic_maximum_mz: float, eic_masses: list (... 14 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,TIC,TIC_POSITIVE,TIC_NEGATI' + 6 child 1, details: string child 2, tic_minimum_mz: float child 3, tic_maximum_mz: float child 4, eic_masses: list<item: float> child 0, item: float child 14, selectivity: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,EE,ER,DR,EZ,ZE' child 1, details: string child 15, wavelength_nanometers: double child 16, wavelength_precision_nanometers: double child 5, isolated_color: string child 6, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SO' + 40 child 1, details: string child 7, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 8, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORKUP' + 68 child 4, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,NMR_OTH' + 92 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' provenance: struct<experimenter: struct<username: string, name: string, orcid: string, organization: string, ema (... 824 chars omitted) child 0, experimenter: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 1, city: string child 2, experiment_start: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 3, doi: string child 4, patent: string child 5, publication_url: string child 6, record_created: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, person: str (... 105 chars omitted) child 0, time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 1, person: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 2, details: string child 7, record_modified: list<item: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, (... 117 chars omitted) child 0, item: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, person: str (... 105 chars omitted) child 0, time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 1, person: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 2, details: string child 8, reaction_metadata: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, is_mined: bool reaction_id: string) dict[str, tuple[RepeatedLevel, tuple[str, ...]]]#
Returns the paths this artifact covers, and the pivot each is read from.
Taken from the same schema walk the structures artifact is built from, so the occurrences reach the elements the projection does by construction rather than by a list here agreeing with one there. Every path the schema can carry a structure at has to come out covered: one dropped would leave the structures sitting only there reachable by no query, which reads as a corpus holding none of them.
- Parameters:
schema – The projection schema.
- Returns:
Each structure-bearing path, mapped to the repeated level whose pivot holds it and the segments from that level’s element down to the structure.
- Raises:
ValueError – If a structure-bearing path is reached by no repeated level, so no pivot can answer for it, or if the schema carries no structure at all. Raised at import, naming the path, because the answer is to change this module rather than to retry.
- ord_schema.artifacts.occurrences.occurrence_path(path: str | PathLike[str]) str | None#
Returns the indexed path an artifact holds, or None if it holds none.
- Parameters:
path – Path to a Parquet file.
- Returns:
The dotted path stamped in the footer, or None if the file is not an occurrences artifact.
- ord_schema.artifacts.occurrences.write_occurrences(source: str | PathLike[str], output: str | PathLike[str], /, *, path: str, source_md5: str | None = None, source_dataset_id: str | None = None, compression: str = 'zstd', row_group_size: int = 100000) int#
Derives the occurrences at one indexed path from a pivot artifact.
- Parameters:
source – The pivot over the level
pathranges within.output – Where to write.
path – The indexed path, which must be a key of
PATHS.source_md5 – The source dataset’s hash; taken from
sourcewhen unset.source_dataset_id – The source dataset’s ID; taken from
sourcewhen unset.compression – Parquet codec.
row_group_size – Rows per group.
- Returns:
How many occurrences were written. Zero is ordinary – most corpora record no authentic standards – and produces a valid empty file rather than none.
- Raises:
ValueError – If
pathis not an indexed path, or ifsourceis not a current pivot over the level that path ranges within.
ord_schema.artifacts.pivot#
The repeated levels of the projection, and what a pivot over one holds.
A quantifier asks whether some element of a repeated level satisfies a body. Answering that from the projection means decoding the whole nested column and filtering its elements; answering it from a pivot – one row per element – means a semi-join against a flat table. Over ORD the second is far faster than the first, because the cost was never the predicate but the reconstruction of lists of structs.
A pivot row carries three things:
reaction_id, which is what the semi-join joins on.The ordinal of every repeated level from the root down to and including this one. Binding an element in the nested form gives co-membership structurally – a measurement reached through a product belongs to it. A flat row has to say so, and a correlation joining on anything short of the full prefix over-returns.
element, a struct holding the element’s own fields with every repeated field removed, recursively. Struct nesting is kept, so a path reaches the same leaf by the same spelling on either route, and a body reaching a dropped field simply fails to resolve – which is how a quantifier the pivot cannot answer declines to the projection without a flag anyone has to keep honest.
Keeping the element’s fields inside one struct rather than hoisting them is what makes
the key columns safe: inputs.components.preparations carries a reaction_id of
its own, and hoisted it would collide with the row’s.
A pivot is unfiltered – every element gets a row, including one whose fields are all
NULL. That completeness is what lets it answer forall, which an index holding only
the elements that match something never can – and what makes an empty one dangerous
rather than merely useless, since forall over it is then vacuously true of every
reaction in the corpus. See Counts.
- class ord_schema.artifacts.pivot.Counts(identity: FileIdentity, at: Mapping[str, int])#
Bases:
Mapping[str,int]How many elements a projection records at each of several levels.
A read-only mapping of level to count, carrying the identity of the file they were read from – which is what lets the build that unnests check that it is unnesting what was counted. A count of zero is the answer nothing downstream can contradict: it skips the unnest, so the check against the rows produced has nothing to disagree with, and the empty pivot that follows costs what this module’s docstring describes.
Handed out whole rather than as a number the caller pairs up itself, because a caller holding a bare count and a bare identity can pass an honest count of one level against an honest identity of the file, for a different level entirely. Held behind a proxy for the same reason one step on: a build shares a projection’s counts across every level it derives, so a write through one level’s handle would be answering for the rest of the run.
- Variables:
identity (ord_schema.artifacts.pivot.FileIdentity) – The file the counts were read from, as it stood across the read.
at (collections.abc.Mapping[str, int]) – The count per level, in the order asked.
- at: Mapping[str, int]#
- identity: FileIdentity#
- class ord_schema.artifacts.pivot.FileIdentity(device: int, inode: int, size: int, modified: int)#
Bases:
NamedTupleWhat tells one version of a file from another, without reading it.
- Variables:
device (int) – The filesystem the file sits on, which is what makes the inode mean anything: inode numbers are unique per device, not across them.
inode (int) – The file’s inode, which an atomic replacement changes; atomic_io writes a sibling and renames over the destination, so every publication is one.
size (int) – Size in bytes.
modified (int) – Modification time in nanoseconds.
- device: int#
Alias for field number 0
- inode: int#
Alias for field number 1
- modified: int#
Alias for field number 3
- size: int#
Alias for field number 2
- class ord_schema.artifacts.pivot.RepeatedLevel(path: str, steps: tuple[Step, ...], element_type: DataType)#
Bases:
objectA repeated level of the projection, and the shape of a pivot over it.
- Variables:
path (str) – The level as the query grammar names it, with no wrapper segments.
steps (tuple[ord_schema.artifacts.pivot.Step, ...]) – One per repeated level from the root down to and including this one, outermost first. Unnesting them in order is what a build walks, and their ordinals are what a row carries to say which element it was.
element_type (pyarrow.lib.DataType) – The element’s struct type with repeated fields removed recursively, which is what a body must resolve against to be answerable from the pivot.
- element_type: DataType#
- property ordinals: tuple[str, ...]#
Returns the ordinal column names, outermost first.
- path: str#
- class ord_schema.artifacts.pivot.Step(segments: tuple[str, ...], is_map: bool, ordinal: str)#
Bases:
objectOne repeated level on the way to another, and how to reach it.
- Variables:
segments (tuple[str, ...]) – The dotted hops from the previous element – or from the row, for the first step – down to the repeated field itself.
is_map (bool) – Whether the repeated field is a map, whose values are the elements.
ordinal (str) – The column naming this element’s position among its siblings.
- expression(source: str | None) str#
Returns the list to unnest at this step.
- Parameters:
source – The variable bound to the previous element, or None at the row.
- Returns:
A DuckDB expression evaluating to a list. A map contributes its values, since those are the elements a query quantifies over.
- is_map: bool#
- ordinal: str#
- segments: tuple[str, ...]#
- ord_schema.artifacts.pivot.artifact_paths(pivots_dir: str | PathLike[str], level_path: str) list[Path]#
Returns the artifacts a reader finds for one level, in a stable order.
The one definition of what a level’s artifacts are. A build that judges its own output by a wider rule than a reader applies calls a tree healthy that no query can read: the level then has no artifacts as far as the corpus is concerned, and every quantifier over it falls back to unnesting the projection, or is refused.
Recursive, because a pivot tree mirrors the projections it was derived from: a sharded corpus puts them under
<level>/<shard>/rather than directly under the level. Named rather than stamped, so listing a level costs no reads – and so a build writing anything else has written something nothing here will find.Through
glob, which descends a symlinked shard directory and passes over a file whose name begins with a dot. Both matter: a deployment is free to put a shard on other storage and link it into place, and the tree may sit on a filesystem whose tools leave dotted siblings of their own. It is also how the projections and structures beside these are found, so one corpus cannot read part of itself.- Parameters:
pivots_dir – The directory the levels sit under.
level_path – The level, as the query grammar names it.
- Returns:
The paths, sorted. Empty if the level has no directory at all.
- ord_schema.artifacts.pivot.check_levels(level_paths: Iterable[str]) list[str]#
Returns
level_pathsas a list, refusing any the schema does not have.- Parameters:
level_paths – The levels to check, as the query grammar names them.
- Returns:
The same paths, in order.
- Raises:
ValueError – If any path is not a repeated level of the projection schema, or if one is named twice, which would leave a caller fewer answers than it asked for.
- ord_schema.artifacts.pivot.count_levels(source: str | PathLike[str], level_paths: Iterable[str]) Counts#
Returns how many elements a projection records at each of
level_paths.One query carrying one aggregate per level, because the count reads the same columns for every level sharing a prefix and a query per level reads them again each time.
- Parameters:
source – Path to the projection to count.
level_paths – The levels to count, as the query grammar names them.
- Returns:
The counts, carrying the identity of the file they were read from. The identity is taken on both sides of the read and must agree, so a projection replaced while it was being counted yields no counts rather than counts of two files.
- Raises:
ValueError – If
sourceis not a current projection, if any path is not a repeated level of the projection schema, if one is named twice, or ifsourcewas replaced while it was being counted.OSError – If
sourcecannot be stat’ed.
- ord_schema.artifacts.pivot.element_expression(dtype: DataType, source: str) str#
Returns a struct literal rebuilding
sourcewithout its repeated fields.The pivot stores what a predicate can read rather than the whole element, which is where the size comes down: the lists a build already unnested into their own levels would otherwise be carried again on every row above them.
- Parameters:
dtype – The pruned element type, as
RepeatedLevel.element_typegives it.source – The DuckDB expression bound to the element.
- Returns:
A struct literal holding each kept field, recursing into kept structs.
- ord_schema.artifacts.pivot.file_identity(source: str | PathLike[str]) FileIdentity#
Returns the identity of the file at
source.- Parameters:
source – Path to the file.
- Returns:
Its device, inode, size, and modification time.
- Raises:
OSError – If
sourcecannot be stat’ed.
- ord_schema.artifacts.pivot.pivot_path(path: str | PathLike[str]) str | None#
Returns the level a pivot artifact holds, or None if it holds none.
- Parameters:
path – Path to a Parquet file.
- Returns:
The dotted level path stamped in the footer, or None if the file is not a pivot artifact.
- ord_schema.artifacts.pivot.reach(path: str, levels: dict[str, RepeatedLevel] | None = None) tuple[RepeatedLevel, tuple[str, ...], DataType] | None#
Returns the pivot a quantifier over
pathranges within, if one does.A quantifier’s path need not name a repeated level itself. Descending from one through singular struct fields reaches one value per element rather than a list of its own –
outcomes.products.measurements.authentic_standardis one compound per measurement – so the level it ranges over is the nearest repeated ancestor, and the pivot over that level already carries the struct.- Parameters:
path – The dotted path the quantifier ranges over.
levels – Levels to search; the projection’s by default.
- Returns:
The level whose pivot holds it, the remaining segments from that level’s element down to the value, and the type they reach – or None if no repeated ancestor covers the path, or the descent leaves the element’s pruned type, which is where every repeated field it dropped would have been.
- ord_schema.artifacts.pivot.repeated_levels(schema: ~pyarrow.lib.Schema = smiles: string identifiers: list<item: struct<type: string, details: string, value: string, is_mapped: bool>> child 0, item: struct<type: string, details: string, value: string, is_mapped: bool> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,REACTION_SMILES,REACTION_CXSMILES,RDF' + 24 child 1, details: string child 2, value: string child 3, is_mapped: bool inputs: map<string, struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: (... 2062 chars omitted) child 0, entries: struct<key: string not null, value: struct<components: list<item: struct<smiles: string, structure_i (... 2086 chars omitted) not null child 0, key: string not null child 1, value: struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: s (... 2049 chars omitted) child 0, components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 1105 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC' + 117 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORK' + 70 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED,DRI' + 14 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,N' + 98 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the times' + 4 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 1, crude_components: list<item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: stru (... 292 chars omitted) child 0, item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: struct<mass_gra (... 280 chars omitted) child 0, reaction_id: string child 1, includes_workup: bool child 2, has_derived_amount: bool child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 2, addition_order: int32 child 3, addition_time_seconds: double child 4, addition_time_precision_seconds: double child 5, addition_speed: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,ALL_AT_ONCE,FAST,SLOW,DROPWISE,CONTINUOU' + 13 child 1, details: string child 6, addition_duration_seconds: double child 7, addition_duration_precision_seconds: double child 8, flow_rate_milliliters_per_minute: double child 9, flow_rate_precision_milliliters_per_minute: double child 10, addition_device: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,SYRINGE,CANNULA,ADDITION_FUN' + 83 child 1, details: string child 11, addition_temperature_kelvin: double child 12, addition_temperature_precision_kelvin: double child 13, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SOLI' + 38 child 1, details: string setup: struct<vessel: struct<type: string, details: string, material: struct<type: string, details: string> (... 524 chars omitted) child 0, vessel: struct<type: string, details: string, material: struct<type: string, details: string>, preparations: (... 234 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,ROUND_BOTTOM_FLASK,VIAL,WELL_PLATE,MI' + 130 child 1, details: string child 2, material: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,GLASS,POLYPROPYLENE,PLASTIC,METAL,Q' + 5 child 1, details: string child 3, preparations: list<item: struct<type: string, details: string>> child 0, item: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,OVEN_DRIED,FLAME_DRIED,EVACU' + 22 child 1, details: string child 4, attachments: list<item: struct<type: string, details: string>> child 0, item: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,NONE,CUSTOM,SEPTUM,CAP,MAT,REFLUX_CONDEN' + 146 child 1, details: string child 5, volume_liters: double child 6, volume_precision_liters: double child 7, vessel_id: string child 8, position: string child 9, row: string child 10, col: string child 1, is_automated: bool child 2, automation_platform: string child 3, automation_code: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 4, environment: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,FUME_HOOD,BENCH_TOP,GLOVE_BOX,GLOVE_B' + 2 child 1, details: string conditions: struct<temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, (... 1761 chars omitted) child 0, temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, setpoint_precision_k (... 199 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,OIL_BATH,WATER_BATH,SAND_BA' + 77 child 1, details: string child 1, setpoint_kelvin: double child 2, setpoint_precision_kelvin: double child 3, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 69 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, temperat (... 57 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,THERMOCOUPLE_INTERNAL,THERMOCOUPL' + 19 child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, temperature_kelvin: double child 5, temperature_precision_kelvin: double child 1, pressure: struct<control: struct<type: string, details: string>, setpoint_kilopascals: double, setpoint_precis (... 264 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,SLIGHT_POSITIVE,SEALED,PRES' + 7 child 1, details: string child 1, setpoint_kilopascals: double child 2, setpoint_precision_kilopascals: double child 3, atmosphere: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AIR,NITROGEN,ARGON,OXYGEN,HYDROGEN,' + 71 child 1, details: string child 4, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 73 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, pressure (... 61 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,PRESSURE_TRANSDUCER' child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, pressure_kilopascals: double child 5, pressure_precision_kilopascals: double child 2, stirring: struct<type: string, details: string, rate: struct<type: string, details: string, rpm: int32>> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,STIR_BAR,OVERHEAD_MIXER,AGITATIO' + 25 child 1, details: string child 2, rate: struct<type: string, details: string, rpm: int32> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,HIGH,MEDIUM,LOW' child 1, details: string child 2, rpm: int32 child 3, illumination: struct<type: string, details: string, peak_wavelength_nanometers: double, peak_wavelength_precision_ (... 124 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,DARK,LED,HALOGEN_LAMP,DEUTERI' + 38 child 1, details: string child 2, peak_wavelength_nanometers: double child 3, peak_wavelength_precision_nanometers: double child 4, color: string child 5, distance_to_vessel_centimeters: double child 6, distance_to_vessel_precision_centimeters: double child 4, electrochemistry: struct<type: string, details: string, current_amperes: double, current_precision_amperes: double, vo (... 446 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,CONSTANT_CURRENT,CONSTANT_VOLTAGE' child 1, details: string child 2, current_amperes: double child 3, current_precision_amperes: double child 4, voltage_volts: double child 5, voltage_precision_volts: double child 6, anode_material: string child 7, cathode_material: string child 8, electrode_separation_centimeters: double child 9, electrode_separation_precision_centimeters: double child 10, measurements: list<item: struct<time_seconds: double, time_precision_seconds: double, current_amperes: double, cur (... 88 chars omitted) child 0, item: struct<time_seconds: double, time_precision_seconds: double, current_amperes: double, current_precis (... 76 chars omitted) child 0, time_seconds: double child 1, time_precision_seconds: double child 2, current_amperes: double child 3, current_precision_amperes: double child 4, voltage_volts: double child 5, voltage_precision_volts: double child 11, cell: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,DIVIDED_CELL,UNDIVIDED_CELL' child 1, details: string child 5, flow: struct<type: string, details: string, pump_type: string, tubing: struct<type: string, details: strin (... 73 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,PLUG_FLOW_REACTOR,CONTINUOUS_STIRRED_' + 31 child 1, details: string child 2, pump_type: string child 3, tubing: struct<type: string, details: string, diameter_centimeters: double, diameter_precision_centimeters: (... 7 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,STEEL,COPPER,PFA,FEP,TEFLONAF,PTFE,' + 25 child 1, details: string child 2, diameter_centimeters: double child 3, diameter_precision_centimeters: double child 6, reflux: bool child 7, ph: float child 8, conditions_are_dynamic: bool child 9, details: string notes: struct<is_heterogeneous: bool, forms_precipitate: bool, is_exothermic: bool, offgasses: bool, is_sen (... 133 chars omitted) child 0, is_heterogeneous: bool child 1, forms_precipitate: bool child 2, is_exothermic: bool child 3, offgasses: bool child 4, is_sensitive_to_moisture: bool child 5, is_sensitive_to_oxygen: bool child 6, is_sensitive_to_light: bool child 7, safety_notes: string child 8, procedure_details: string observations: list<item: struct<time_seconds: double, time_precision_seconds: double, comment: string, image: stru (... 139 chars omitted) child 0, item: struct<time_seconds: double, time_precision_seconds: double, comment: string, image: struct<float_va (... 127 chars omitted) child 0, time_seconds: double child 1, time_precision_seconds: double child 2, comment: string child 3, image: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string workups: list<item: struct<type: string, details: string, duration_seconds: double, duration_precision_second (... 2903 chars omitted) child 0, item: struct<type: string, details: string, duration_seconds: double, duration_precision_seconds: double, (... 2891 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,ADDITION,ALIQUOT,TEMPERATURE,CONCENTR' + 166 child 1, details: string child 2, duration_seconds: double child 3, duration_precision_seconds: double child 4, input: struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: s (... 2049 chars omitted) child 0, components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 1105 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC' + 117 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORK' + 70 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED,DRI' + 14 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,N' + 98 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the times' + 4 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 1, crude_components: list<item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: stru (... 292 chars omitted) child 0, item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: struct<mass_gra (... 280 chars omitted) child 0, reaction_id: string child 1, includes_workup: bool child 2, has_derived_amount: bool child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 2, addition_order: int32 child 3, addition_time_seconds: double child 4, addition_time_precision_seconds: double child 5, addition_speed: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,ALL_AT_ONCE,FAST,SLOW,DROPWISE,CONTINUOU' + 13 child 1, details: string child 6, addition_duration_seconds: double child 7, addition_duration_precision_seconds: double child 8, flow_rate_milliliters_per_minute: double child 9, flow_rate_precision_milliliters_per_minute: double child 10, addition_device: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,SYRINGE,CANNULA,ADDITION_FUN' + 83 child 1, details: string child 11, addition_temperature_kelvin: double child 12, addition_temperature_precision_kelvin: double child 13, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SOLI' + 38 child 1, details: string child 5, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATED' child 1, details: string child 7, volume_includes_solutes: bool child 6, temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, setpoint_precision_k (... 199 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,OIL_BATH,WATER_BATH,SAND_' + 79 child 1, details: string child 1, setpoint_kelvin: double child 2, setpoint_precision_kelvin: double child 3, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 69 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, temperat (... 57 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,THERMOCOUPLE_INTERNAL,THERMOCOU' + 21 child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, temperature_kelvin: double child 5, temperature_precision_kelvin: double child 7, keep_phase: string child 8, stirring: struct<type: string, details: string, rate: struct<type: string, details: string, rpm: int32>> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,STIR_BAR,OVERHEAD_MIXER,AGITAT' + 27 child 1, details: string child 2, rate: struct<type: string, details: string, rpm: int32> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,HIGH,MEDIUM,LOW' child 1, details: string child 2, rpm: int32 child 9, target_ph: float child 10, is_automated: bool outcomes: list<item: struct<reaction_time_seconds: double, reaction_time_precision_seconds: double, conversion (... 2983 chars omitted) child 0, item: struct<reaction_time_seconds: double, reaction_time_precision_seconds: double, conversion: struct<va (... 2971 chars omitted) child 0, reaction_time_seconds: double child 1, reaction_time_precision_seconds: double child 2, conversion: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 3, products: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 2422 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 2410 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC_N' + 115 child 1, details: string child 2, value: string child 3, is_desired_product: bool child 4, measurements: list<item: struct<analysis_key: string, type: string, details: string, uses_internal_standard: bool, (... 1984 chars omitted) child 0, item: struct<analysis_key: string, type: string, details: string, uses_internal_standard: bool, is_normali (... 1972 chars omitted) child 0, analysis_key: string child 1, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,IDENTITY,YIELD,SELECTIVITY,PU' + 33 child 2, details: string child 3, uses_internal_standard: bool child 4, is_normalized: bool child 5, uses_authentic_standard: bool child 6, authentic_standard: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,I' + 121 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TIT' + 5 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,' + 74 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED' + 18 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_1' + 102 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the t' + 8 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPH' + 46 child 1, details: string child 7, percentage: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 8, float_value: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 9, string_value: string child 10, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRA' + 3 child 1, details: string child 7, volume_includes_solutes: bool child 11, retention_time_seconds: double child 12, retention_time_precision_seconds: double child 13, mass_spec_details: struct<type: string, details: string, tic_minimum_mz: float, tic_maximum_mz: float, eic_masses: list (... 14 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,TIC,TIC_POSITIVE,TIC_NEGATI' + 6 child 1, details: string child 2, tic_minimum_mz: float child 3, tic_maximum_mz: float child 4, eic_masses: list<item: float> child 0, item: float child 14, selectivity: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,EE,ER,DR,EZ,ZE' child 1, details: string child 15, wavelength_nanometers: double child 16, wavelength_precision_nanometers: double child 5, isolated_color: string child 6, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SO' + 40 child 1, details: string child 7, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 8, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORKUP' + 68 child 4, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,NMR_OTH' + 92 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' provenance: struct<experimenter: struct<username: string, name: string, orcid: string, organization: string, ema (... 824 chars omitted) child 0, experimenter: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 1, city: string child 2, experiment_start: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 3, doi: string child 4, patent: string child 5, publication_url: string child 6, record_created: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, person: str (... 105 chars omitted) child 0, time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 1, person: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 2, details: string child 7, record_modified: list<item: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, (... 117 chars omitted) child 0, item: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, person: str (... 105 chars omitted) child 0, time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 1, person: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 2, details: string child 8, reaction_metadata: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, is_mined: bool reaction_id: string) dict[str, RepeatedLevel]#
Returns every repeated level of
schema, by the path a query names it.Walked rather than listed by hand, in the same spirit as
structures.structure_levels: a repeated field added upstream becomes a level nobody had to remember to add, and one walk cannot disagree with itself the way two lists can.A list or a map is a repeated level; both are levels a query quantifies over, and the grammar spells neither with a wrapper segment, so
inputs.componentsnames the components under a map of inputs.- Parameters:
schema – Schema to walk; the projection schema by default.
- Returns:
Each level’s path mapped to what a pivot over it holds. A list of scalars is absent – there is no element struct to pivot – and so is a level whose elements carry nothing but repeated fields.
- Raises:
ValueError – If two levels on one path yield the same ordinal name, which would give a row two columns of that name and let the inner one silently win. Raised at import, naming the ordinal, because the answer is to name it differently here rather than to retry.
- ord_schema.artifacts.pivot.schema(level: RepeatedLevel) Schema#
Returns the schema of a pivot artifact over
level.- Parameters:
level – The level the artifact holds.
- Returns:
reaction_id, one ordinal per repeated level above and including this one, and the pruned element.
- ord_schema.artifacts.pivot.select(level: RepeatedLevel, table: str) str#
Returns the query building the pivot over
level.Unnests one repeated level at a time rather than taking the flattened traversal the compiler uses, because the ordinals are what each unnest contributes and a list already flattened has lost them.
- Parameters:
level – The level to pivot.
table – The relation holding the reactions.
- Returns:
A SELECT over
table, one row per element of the level.
- ord_schema.artifacts.pivot.stage(connection: DuckDBPyConnection, level: RepeatedLevel, table: str) str#
Unnests
levela step at a time through temp tables, returning the last query.A step given a table of its own can run on its own, where a whole level reached in one query – every unnest and the element prune in a single pipeline, which is what
selectreturns – runs largely serially. The work is the same either way; what differs is how much of the machine it can occupy. Over an ORD projection the two cost about the same CPU and staging spends it faster in wall clock on a many-core machine, so expect the margin to fall with the core count.Every step but the last gets a table. Fusing more than one re-serializes them, so a single-step level is the case that must be staged rather than the case that can be skipped.
The tables hold the element whole, unlike the artifact: pruning is what removes the repeated fields, and a child reaches its own through the field a pruned parent would have dropped.
- Parameters:
connection – Connection to create the temp tables in, holding
table. They live as long as it does and are named for the level’s ancestors, so staging a second level that shares one raises rather than answering short. Give each call a connection it can have to itself.level – The level to pivot.
table – The relation holding the reactions.
- Returns:
A SELECT of
reaction_id, the level’s ordinals, and the pruned element, one row per element of the level, reading the last temp table rather thantable.- Raises:
duckdb.Error – If a table this level would stage is already in
connection.
- ord_schema.artifacts.pivot.table_name(path: str) str#
Returns the relation name holding the pivot over
path.- Parameters:
path – The level as the query grammar names it.
- Returns:
A bare identifier, so it reaches SQL without quoting.
- ord_schema.artifacts.pivot.write_pivot(source: str | PathLike[str], output: str | PathLike[str], /, *, level_path: str, source_md5: str | None = None, source_dataset_id: str | None = None, compression: str = 'zstd', row_group_size: int = 100000, counts: Counts | None = None) int#
Derives a pivot artifact over one repeated level and writes it.
One artifact holds one level of one projection, because the levels have different schemas and a reader globs the one it wants. The rows stream out of DuckDB a batch at a time, so the caller holds a batch of a level that explodes – a corpus of long measurement lists – rather than the whole level. DuckDB holds more, under its own memory limit rather than the caller’s: the sort that orders the rows holds the whole level, and
stageholds each ancestor whole, repeated fields and all.The output is published atomically, so a failure partway leaves any existing artifact untouched.
- Parameters:
source – Path to the projection to pivot.
output – Path to write the artifact to.
level_path – The repeated level to pivot, as the query grammar names it.
source_md5 – Hash of the source dataset to stamp, if the caller already read one. Taken from the projection’s own stamps when omitted, so the artifact names the dataset it reflects rather than the file it read.
source_dataset_id – Source dataset ID to stamp, if the caller already read one. Taken from the projection’s stamps when omitted.
compression – Parquet codec, any name
pq.ParquetWriteraccepts.row_group_size – Rows per output row group, which is also how many elements the caller holds at a time.
counts – What
count_levelsread offsource, where the caller has already counted. Counted here when omitted. A build deriving every level counts them together, since one query answers for all of them at the cost of reading the projection once. The level’s own count is taken from these rather than passed alongside them, and the identity they carry is checked againstsource: a count of zero skips the unnest, so unlike a nonzero one it is never checked against the rows produced, and this is what stands in place of that check.
- Returns:
the number of elements at that level.
- Return type:
Number of rows written
- Raises:
ValueError – If
sourceis not a current projection, if the schema has no such repeated level, ifsourcehas changed since it was counted, or if the element count and the unnest disagree.KeyError – If
countsdoes not coverlevel_path.OSError – If
countsis given andsourcecannot be stat’ed.
ord_schema.artifacts.projection#
A queryable projection of the whole Reaction proto.
Source datasets store reactions as opaque wire-format bytes, so any question about them
costs a full deserialization. A projection restates the same reactions as nested
Parquet – message to STRUCT, repeated field to LIST, map to MAP – so a
query engine can read one leaf without touching the rest.
No field is left out. Every field of every message reachable from Reaction
appears, because a field left out is a question nobody can ask, and the point of the
artifact is to support questions nobody enumerated in advance. The schema is generated
from the proto descriptors rather than written by hand, so a new field appears here
without anyone deciding it is worth carrying.
That works only because the message graph is acyclic: no message reaches itself, so the
recursion terminates. build_schema enforces this rather than assuming it – a
recursive message added upstream would otherwise recurse forever at import time.
Three normalizations are applied, and only three. Each costs no query and removes a real trap:
United messages become canonical floats. A
{value, precision, units}triple projected verbatim meansWHERE temperature > 350silently misses every row recorded in Celsius. Each becomes two columns named for its unit –setpoint_kelvinandsetpoint_precision_kelvin– so a question in other units stays expressible while the mixed-unit comparison that would quietly return the wrong rows does not. The source records precision in the same units as the value, so it converts with the value and is null under the same conditions.Dates become typed values.
DateTimeis a bare string the depositor wrote, and the corpus holds many shapes of it, so comparing the text answers arbitrarily: a prefix match on a year misses that year’s records written in another shape, and matches a dataset whose stamps merely begin with those digits. EachDateTimegains atimestampand autc_offset_minutesbeside the recorded string, which stays: the projection adds a reading rather than replacing what the source says, and a shape none of the formats cover is a visible null.Every time keeps the wall clock it was written with, in
timestamp, and the UTC offset it was written under, inutc_offset_minutes: minutes east of UTC, named for its unit as the converted quantities are, and null where the record names none. Together they hold everything the string says. A zone-aware column could not: it stores one instant per value under one zone for the whole column, so it would drop each value’s own offset and read a time with none as UTC.NN/NN/NNNNis the part that needs more than a format list, since the string never says which field is the month. It is settled per dataset and before any row is written – the convention belongs to the depositor, not to a value – from_DAY_FIRST, from the dataset’s own unambiguous values, and from month first otherwise. A source that records no dataset ID is the exception: no list can name it, so only its own values speak for it, and where none does its ambiguous timestamps are null, because a guess written here is indistinguishable afterwards from a date the source recorded.The string the depositor wrote stays beside the timestamp, and is internal. Keeping it is what lets such a null be read back once the order is known; hiding it is what keeps a date question off a column the corpus spells ten ways.
Structural identifiers collapse to one ``smiles``.
SMILES,CXSMILES,INCHI, andMOLBLOCKall answer “what is this molecule,” so the projection answers it once, throughmessage_helpers.smiles_from_compound– the same CXSMILES-first choice the tabular view makes, so the two never disagree about a molecule.REACTION_SMILESandREACTION_CXSMILEScollapse the same way at the reaction level, into asmilesthatmessage_helpers.derived_reaction_smilesnormalizes from the recorded value – agents removed and the result canonicalized, so a reaction deposited asA>B>Creads asA>>C– or generates from the components when the source records nothing readable. Atom mapping is kept; fragment grouping is not.The drop is all-or-nothing per message and conditional on the collapse producing something: a compound whose identifiers none of RDKit’s loaders can read keeps all of them, rather than reading as a compound whose structure the source never recorded.
Every other identifier is kept, as a list. NAME alone covers compounds that no
structural identifier reaches, and a compound may carry several of them, so pivoting
identifiers into named scalar fields would silently drop data.
One derived column rides beside the compound-level smiles: a structure_id
numbering the dataset’s distinct structures in first-seen order, the join key into the
dataset’s structures artifact (ord_schema.artifacts.structures), where the
fingerprints live. It exists because a structure predicate is evaluated outside the
query – screened and verified against the structures artifact – and its match set has
to re-enter the query at element granularity: DuckDB does not allow a subquery inside a
lambda expression, so the components a quantifier binds cannot semi-join against a match
table, but they can test one integer against a bitmap parameter. The IDs are meaningful
only alongside the projection that assigned them, so the column is marked internal and
stays out of what a model is told it may query.
Anything else – the original units, the identifier types that were collapsed – remains
in the source reaction column, which stays authoritative for byte-exact
round-tripping. The projection promises that every field is readable, not that it is
reproducible from the projection alone.
Warning
Query repeated levels with list lambdas, not UNNEST in the FROM clause.
UNNEST materializes the exploded intermediate, which is far slower where it
finishes at all; the artifacts README has the measurements. It also emits one row
per matching identifier rather than one per reaction, so the two spellings below need
the DISTINCT to count alike.
-- fast
SELECT reaction_id FROM p
WHERE len(list_filter(
flatten(list_transform(map_values(inputs), i -> i.components)),
c -> len(list_filter(c.identifiers,
x -> x.type = 'NAME' AND x.value = 'THF')) > 0)) > 0
-- same answer, does not finish
SELECT DISTINCT reaction_id
FROM p, UNNEST(map_values(inputs)) t(i), UNNEST(i.components) u(c),
UNNEST(c.identifiers) v(x)
WHERE x.type = 'NAME' AND x.value = 'THF'
- ord_schema.artifacts.projection.build_schema() Schema#
Returns the Arrow schema projected from the Reaction descriptor.
- Returns:
One top-level field per Reaction field, plus the collapsed
smiles.- Raises:
ValueError – If the message graph contains a cycle, which a total projection cannot represent.
- ord_schema.artifacts.projection.column_name(field: FieldDescriptor) str#
Returns the column name for
field, carrying the unit where one applies.
- ord_schema.artifacts.projection.dataset_slash_orientation(dataset_id: str | None, view: DatasetView) bool | None#
Returns whether a dataset writes slash dates day first, as its projection reads.
Every row of a dataset is projected under this one answer, which no single row carries, so code that re-projects one row, such as a fidelity check, reads it here.
- Parameters:
dataset_id – The source’s
ord_dataset-ID, as the projection is stamped with it, or None where the source records none.view – The source dataset, whose recorded dates are scanned until one settles the order.
- Returns:
True for day first, False for month first, and None where nothing settles it; see
slash_orientation.- Raises:
ValueError – If a witness in the data contradicts
_DAY_FIRST.
- ord_schema.artifacts.projection.day_first_dates(values: Iterable[str | None]) bool | None#
Returns whether a dataset writes slash dates day first, or None if it cannot say.
The convention is a property of the depositor rather than of a value, and a value only reveals it where one of the two leading numbers exceeds twelve. No ORD dataset carries evidence both ways, so one answer per dataset is the right shape – and some carry none at all, which is what None is for: those hold one bulk-import stamp repeated across every row, so a date guessed for them would be guessed for every reaction in the dataset at once.
- Parameters:
values – Every date string the dataset records, in any order. Nones and values in other formats are ignored, since only a slash date is ambiguous.
- Returns:
True where the dataset writes day/month, False where it writes month/day, and None where no value settles it.
- ord_schema.artifacts.projection.enum_members(field: Field) tuple[str, ...] | None#
Returns the values
fieldmay hold, or None if it is not an enum column.- Parameters:
field – A field of
SCHEMA, or of a struct within it.- Returns:
The member names in declaration order, or None.
- ord_schema.artifacts.projection.is_current(path: str | PathLike[str], source_md5: str) bool#
Returns whether
pathis a projection ofsource_md5by this library.
- ord_schema.artifacts.projection.is_internal(field: Field) bool#
Returns whether
fieldis artifact-internal machinery rather than data.- Parameters:
field – A field of
SCHEMA, or of a struct within it.- Returns:
True for columns like
structure_idwhose values join artifacts together and are not stable across builds. They are hidden from the schema description a model sees, and a query naming one is refused.
- ord_schema.artifacts.projection.message_row(message: Message, structure_ids: MutableMapping[str, int] | None = None, *, day_first: bool | None = None) dict[str, Any]#
Projects
messageto a dict matching its struct type inSCHEMA.Unset fields are None rather than the proto default, so a consumer can tell “the source is silent” from “the source says zero”.
- Parameters:
message – Any message reachable from Reaction.
structure_ids – Mapping from SMILES to
structure_id, extended in first-seen order as compounds are projected. None – the default, for a message projected outside a dataset – leaves everystructure_idnull, since an ID is meaningful only against the artifact that shares the mapping.day_first – Whether this dataset writes slash dates day first, from
dataset_slash_orientation. None leaves an ambiguous one unparsed; seeparse_timestamp.
- Returns:
A dict keyed by projected column name.
- ord_schema.artifacts.projection.parse_timestamp(value: str, *, day_first: bool | None = None) datetime | None#
Returns the time a recorded date names, as written, or None where no shape fits.
- Parameters:
value – The string a depositor wrote.
day_first – Whether this dataset writes slash dates day first, from
dataset_slash_orientation. None leaves an ambiguous slash date unparsed rather than guessing, which is the difference between a null a reader can see and a date that is silently the wrong month.
- Returns:
The parsed time as written, without the UTC offset
parse_utc_offsetreads; None where no format matched or the value is ambiguous and the dataset does not say which way to read it.
- ord_schema.artifacts.projection.parse_utc_offset(value: str) int | None#
Returns the UTC offset a recorded time was written under, in minutes east of UTC.
Together with
parse_timestamp, which keeps the time as written, this holds everything such a value says, so the instant stays recoverable where the record names one.- Parameters:
value – The string a depositor wrote.
- Returns:
The offset in minutes, negative west of UTC as in ISO 8601 (
-05:00is -300), or None where the value states none: a date, a time written without one, or a shape validation does not accept.
- ord_schema.artifacts.projection.precision_column_name(field: FieldDescriptor) str#
Returns the column name for
field’s precision.The unit stays the trailing element, matching the value column, so a reader takes the units off the end of either name without knowing which is which.
- Parameters:
field – A united field, as identified by
_canonical_unit.- Returns:
The precision column’s name.
- Raises:
ValueError – If
fieldis not united, and so has no precision to name.
- ord_schema.artifacts.projection.reaction_row(reaction: Reaction, reaction_id: str | None, structure_ids: MutableMapping[str, int] | None = None, *, day_first: bool | None = None) dict[str, Any]#
Projects one row of a source dataset to a dict matching
SCHEMA.The message is authoritative for every projected value, including the ID; the source’s
reaction_idcolumn – the keyDatasetView.get_reactionindexes – is checked against it rather than chosen between. A check and not a fallback becauseDatasetView.md5hashes the Reaction blobs and not that column: an ID read from the column would sit outside the staleness hash, so correcting the column alone would leave the artifact looking current forever.- Parameters:
reaction – Reaction to project.
reaction_id – The row’s
reaction_idcolumn value, exactly as read – None where the column is null. Required rather than defaulted, so a null cannot arrive looking like a caller who read no column and skip the check. Project a Reaction from anywhere else withmessage_row.structure_ids – Mapping from SMILES to
structure_id, shared across a dataset’s rows and extended in first-seen order. None leaves everystructure_idnull.day_first – Whether this dataset writes slash dates day first, from
dataset_slash_orientation. None leaves an ambiguous one unparsed; seeparse_timestamp.
- Returns:
A dict keyed by projected column name.
- Raises:
ValueError – If
reaction_idis null or empty, or disagrees with the ID the message records. Each means the source cannot be joined on its own key.
- ord_schema.artifacts.projection.slash_orientation(dataset_id: str | None, values: Iterable[str | None]) bool | None#
Returns whether a dataset writes slash dates day first, preferring what is known.
_DAY_FIRSTanswers first, holding evidence a value cannot carry – a sibling submission or the submitter’s answer, alongside the datasets that witness themselves. The scan then answers for a dataset nobody has looked at yet, which is what keeps a day-first dataset added later from silently reading month first. Month first is what is left, and what every settled dataset outside_DAY_FIRSTreads.- Parameters:
dataset_id – The source’s
ord_dataset-ID, or None where it records none.values – Every date string the dataset records. Read whether or not a list names the dataset, since a listed verdict is checked against any witness among them – but read lazily, so a dataset that produces one stops there. A dataset settled by something other than a witness has none to find, and that scan runs to the end, which is still small beside the projection itself.
- Returns:
True for day first, False for month first, and None for a source recording no ID whose own values settle nothing – which leaves ambiguous dates unparsed rather than guessed.
- Raises:
ValueError – If a witness in the data contradicts
_DAY_FIRST. One of the two is then wrong about the dataset, and projecting either reading would record a date this library has evidence against.
- ord_schema.artifacts.projection.write_projection(source: str | PathLike[str], output: str | PathLike[str], *, source_md5: str | None = None, source_dataset_id: str | None = None, compression: str = 'zstd', row_group_size: int = 10000) int#
Projects a source Parquet dataset and writes it.
Reactions are read one source row group at a time and written one output row group at a time, so peak memory is bounded by one of each rather than by the dataset. The output is published atomically, so a failure partway leaves any existing projection untouched.
- Parameters:
source – Path to the source Parquet dataset.
output – Path to write the projection to.
source_md5 – Source hash to stamp, if the caller already computed one. Hashed here when omitted, which costs a full pass over the source.
source_dataset_id – Source dataset ID to stamp, if the caller already read one. Read from the source when omitted.
compression – Parquet compression codec.
row_group_size – Rows per output row group, independent of the source’s. The footer holds statistics for every leaf of every row group and is parsed before any data is read, so fewer, larger row groups make a projection cheaper to open.
- Returns:
Number of rows written.
- Raises:
ValueError – If
row_group_sizeis less than one, which no row group can hold, refused before the source is read; ifsourceis not a source dataset; or if any row’sreaction_idcolumn is missing or disagrees with its Reaction. The message names the source, since a corpus-wide run has thousands of datasets to choose between.
ord_schema.artifacts.structures#
A structure-search artifact: one row per distinct structure in a projection.
Structure search runs in two steps that this artifact serves in turn. A screen tests the query’s fingerprint bits against every row – a bitwise scan a query engine runs in milliseconds without a chemistry library – and is complete but not exact: it never misses a true match, and passes false ones. Verification then runs real subgraph isomorphism over the survivors, from the serialized molecule carried here, which skips re-parsing SMILES and runs faster for it. Neither step needs an index; the reason a structure search elsewhere needs one is the cost of scanning rows one at a time, and a columnar scan of every distinct structure in the corpus is cheap.
The artifact is deduplicated, which is what makes both steps affordable: the corpus
holds roughly fifteen component occurrences per distinct structure, so screening and
verification run against the distinct set and their results fan back out through the
projection. The join is structure_id – assigned by write_projection in
first-seen order, carried on every compound there that records a structure and numbered
0..n-1 here, so a match set travels back into a query as a bitmap indexed by ID.
The two files are two statements of one derivation and are meaningful only together:
IDs are not stable across builds, nothing outside the artifacts should record them, and
a projection rewritten in place needs its structures artifact rederived with it – the
currency stamps name the source dataset rather than the projection file, so the skip
check cannot see that the pairing changed.
The screen’s completeness is an invariant, not a hope: if query Q is a subgraph of
target T, every bit of Q’s pattern fingerprint is set in T’s, so
bit_count(fp & q) = popcount(q) never rejects a true match – checked across
query/target pairs, including explicit-hydrogen queries, with no exception. What a
query layer still has to handle is that these molecules are built from SMILES and so
hold their hydrogens implicitly: a SMARTS naming one as its own atom matches nothing,
which ord_schema.search.query rewrites rather than runs. Similarity needs no
verification at all: Tanimoto is defined on the Morgan fingerprint, so the screen is
the answer, and morgan_popcount bounds it (popcount(B) must lie within [t *
popcount(A), popcount(A) / t] for Tanimoto >= t).
Beside the SMILES the projection derived, a row carries a mol_hash: an identity
for the molecule that ignores how it was drawn. Equality on the stored spelling is
exact and answers the wrong question surprisingly often – acetic acid and acetate, an
amine and its ammonium, a 2-pyridone and its 2-hydroxypyridine tautomer are each one
reagent written two ways, and each compares unequal.
It is RDKit’s registration hash, taken of the uncharged molecule. Uncharging is what makes the scheme insensitive to protonation as well as to tautomer: one of its layers is the molecular formula, and a formula states the charge. Registration is the right API rather than a bare tautomer hash because it normalizes what a bare hash does not – atom-map labels, unnecessary explicit hydrogens, enhanced stereo, S-group data – so a compound written with atom maps is the same compound. It is a hash rather than a canonical tautomer because canonicalizing one costs what the corpus does not have: over ORD’s own molecules, RDKit’s tautomer enumerator is several times slower than this hash, before the per-file parallelism the build already has. The stored SMILES beside it is what a reader looks at; the hash is only ever compared.
Fragments are left alone, so sodium acetate stays distinct from acetic acid. The
question that counts a reagent and the salt it was sold as as one is a different one,
and it gets its own column: a parent_hash, the same hash taken of the molecule with
its recognized counterions removed. Two columns rather than one choice between them,
because either answer is wrong for the other question – sodium and potassium carbonate
are one reagent to a query about carbonate and two to a query about the sodium – and a
corpus-wide column is the wrong place to guess which was meant. See salt_parent for
which counterions are recognized and where stripping stops.
The hash is RDKit’s and moves with it, which is why base stamps the RDKit version
and is_current refuses an artifact built by another one. Without that stamp an
RDKit upgrade would silently change what a stored column means while every query kept
running.
A structure whose SMILES RDKit cannot parse keeps its row – the ID space must stay dense – with every derived column null. It can never match a structure query, and the count of such rows is logged per dataset.
- ord_schema.artifacts.structures.is_current(path: str | PathLike[str], source_md5: str) bool#
Returns whether
pathis a structures artifact ofsource_md5by us.Delegates to
base.is_current, which requires the artifact name, the source content, the artifact schema version, and the RDKit version to all match. A missing or unreadable file is not current.- Parameters:
path – Path to check. Need not exist.
source_md5 – Hash of the source dataset the artifact should restate, which is the dataset’s own even though the artifact derives from a projection.
- Returns:
Whether the file is a structures artifact this library would write today.
- ord_schema.artifacts.structures.mol_hash(structure: Mol | str) str | None#
Returns an identity for a molecule that ignores tautomer and protonation state.
Two drawings of one reagent hash the same: acetic acid and acetate, an amine and its ammonium, a 2-pyridone and its 2-hydroxypyridine tautomer, a compound written with atom-map labels and the same compound without them. Fragments are left alone, so sodium acetate hashes differently from acetic acid, and so is stereochemistry, so enantiomers stay apart.
The molecule is uncharged before it is hashed, which is what makes the scheme insensitive to protonation as well: one of its layers is the molecular formula, and a formula states the charge.
- Parameters:
structure – A parsed molecule, or SMILES to read one from.
- Returns:
The hash, or None where RDKit cannot read or hash the structure, which leaves the column null rather than storing a value the query side would not reproduce.
- ord_schema.artifacts.structures.morgan_fingerprint(molecule: Mol) ExplicitBitVect#
Returns the Morgan fingerprint with the parameters this artifact stores.
The query side of a similarity search must fingerprint identically to the artifact or the comparison is meaningless, so this is the one place the parameters live.
- Parameters:
molecule – The molecule to fingerprint.
- Returns:
A
MORGAN_FP_SIZE-bit fingerprint at radiusMORGAN_RADIUS.
- ord_schema.artifacts.structures.parent_hash(structure: Mol | str) str | None#
Returns the hash of a molecule’s salt parent.
What
mol_hashignores, this ignores too, and counterions besides. Two reagents whose parents differ still differ: potassium and sodium carbonate share a parent and hash alike, while sodium carbonate and sodium acetate do not.- Parameters:
structure – A parsed molecule, or SMILES to read one from.
- Returns:
The hash, or None where RDKit cannot read, strip, or hash the structure, which leaves the column null rather than storing a value the query side would not reproduce.
- ord_schema.artifacts.structures.salt_parent(molecule: Mol) Mol#
Returns the molecule without the counterions it was sold as.
Sodium acetate becomes acetate and triethylamine hydrochloride becomes triethylamine, so a reagent recorded with a spectator counterion and the same reagent recorded without one come out the same.
“Recognized” is RDKit’s own list and no wider: the halides, lithium, sodium, potassium, calcium and magnesium, and the usual acid counterions – nitrate, sulfate, phosphate, hexafluorophosphate, mesylate, tosylate, acetate, trifluoroacetate, oxalate, tartrate, fumarate and maleate. A counterion off that list is not a counterion here: cesium carbonate keeps its cesium and so does not share a parent with the potassium carbonate that loses its potassium. Widening the list is not obviously right – the metals a corpus of reactions carries beside an anion are mostly reagents rather than spectators, and many of this corpus’s multi-fragment structures hold iron, zinc, copper, silver or aluminium – so the list stays RDKit’s, and the RDKit version the artifact stamps says which one.
Stripping stops where the molecule is the salt: nothing is removed unless what survives still holds carbon, which leaves sodium hydride whole rather than turning it into hydrogen, and palladium acetate whole rather than into palladium. That rule is why this is not RDKit’s fragment parent, which keeps the largest fragment and would answer palladium acetate with acetic acid.
- Parameters:
molecule – A parsed molecule.
- Returns:
The stripped molecule, or the original where stripping would leave no carbon.
- ord_schema.artifacts.structures.structure_levels(schema: ~pyarrow.lib.Schema = smiles: string identifiers: list<item: struct<type: string, details: string, value: string, is_mapped: bool>> child 0, item: struct<type: string, details: string, value: string, is_mapped: bool> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,REACTION_SMILES,REACTION_CXSMILES,RDF' + 24 child 1, details: string child 2, value: string child 3, is_mapped: bool inputs: map<string, struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: (... 2062 chars omitted) child 0, entries: struct<key: string not null, value: struct<components: list<item: struct<smiles: string, structure_i (... 2086 chars omitted) not null child 0, key: string not null child 1, value: struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: s (... 2049 chars omitted) child 0, components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 1105 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC' + 117 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORK' + 70 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED,DRI' + 14 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,N' + 98 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the times' + 4 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 1, crude_components: list<item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: stru (... 292 chars omitted) child 0, item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: struct<mass_gra (... 280 chars omitted) child 0, reaction_id: string child 1, includes_workup: bool child 2, has_derived_amount: bool child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 2, addition_order: int32 child 3, addition_time_seconds: double child 4, addition_time_precision_seconds: double child 5, addition_speed: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,ALL_AT_ONCE,FAST,SLOW,DROPWISE,CONTINUOU' + 13 child 1, details: string child 6, addition_duration_seconds: double child 7, addition_duration_precision_seconds: double child 8, flow_rate_milliliters_per_minute: double child 9, flow_rate_precision_milliliters_per_minute: double child 10, addition_device: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,SYRINGE,CANNULA,ADDITION_FUN' + 83 child 1, details: string child 11, addition_temperature_kelvin: double child 12, addition_temperature_precision_kelvin: double child 13, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SOLI' + 38 child 1, details: string setup: struct<vessel: struct<type: string, details: string, material: struct<type: string, details: string> (... 524 chars omitted) child 0, vessel: struct<type: string, details: string, material: struct<type: string, details: string>, preparations: (... 234 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,ROUND_BOTTOM_FLASK,VIAL,WELL_PLATE,MI' + 130 child 1, details: string child 2, material: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,GLASS,POLYPROPYLENE,PLASTIC,METAL,Q' + 5 child 1, details: string child 3, preparations: list<item: struct<type: string, details: string>> child 0, item: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,OVEN_DRIED,FLAME_DRIED,EVACU' + 22 child 1, details: string child 4, attachments: list<item: struct<type: string, details: string>> child 0, item: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,NONE,CUSTOM,SEPTUM,CAP,MAT,REFLUX_CONDEN' + 146 child 1, details: string child 5, volume_liters: double child 6, volume_precision_liters: double child 7, vessel_id: string child 8, position: string child 9, row: string child 10, col: string child 1, is_automated: bool child 2, automation_platform: string child 3, automation_code: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 4, environment: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,FUME_HOOD,BENCH_TOP,GLOVE_BOX,GLOVE_B' + 2 child 1, details: string conditions: struct<temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, (... 1761 chars omitted) child 0, temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, setpoint_precision_k (... 199 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,OIL_BATH,WATER_BATH,SAND_BA' + 77 child 1, details: string child 1, setpoint_kelvin: double child 2, setpoint_precision_kelvin: double child 3, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 69 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, temperat (... 57 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,THERMOCOUPLE_INTERNAL,THERMOCOUPL' + 19 child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, temperature_kelvin: double child 5, temperature_precision_kelvin: double child 1, pressure: struct<control: struct<type: string, details: string>, setpoint_kilopascals: double, setpoint_precis (... 264 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,SLIGHT_POSITIVE,SEALED,PRES' + 7 child 1, details: string child 1, setpoint_kilopascals: double child 2, setpoint_precision_kilopascals: double child 3, atmosphere: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AIR,NITROGEN,ARGON,OXYGEN,HYDROGEN,' + 71 child 1, details: string child 4, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 73 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, pressure (... 61 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,PRESSURE_TRANSDUCER' child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, pressure_kilopascals: double child 5, pressure_precision_kilopascals: double child 2, stirring: struct<type: string, details: string, rate: struct<type: string, details: string, rpm: int32>> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,STIR_BAR,OVERHEAD_MIXER,AGITATIO' + 25 child 1, details: string child 2, rate: struct<type: string, details: string, rpm: int32> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,HIGH,MEDIUM,LOW' child 1, details: string child 2, rpm: int32 child 3, illumination: struct<type: string, details: string, peak_wavelength_nanometers: double, peak_wavelength_precision_ (... 124 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,DARK,LED,HALOGEN_LAMP,DEUTERI' + 38 child 1, details: string child 2, peak_wavelength_nanometers: double child 3, peak_wavelength_precision_nanometers: double child 4, color: string child 5, distance_to_vessel_centimeters: double child 6, distance_to_vessel_precision_centimeters: double child 4, electrochemistry: struct<type: string, details: string, current_amperes: double, current_precision_amperes: double, vo (... 446 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,CONSTANT_CURRENT,CONSTANT_VOLTAGE' child 1, details: string child 2, current_amperes: double child 3, current_precision_amperes: double child 4, voltage_volts: double child 5, voltage_precision_volts: double child 6, anode_material: string child 7, cathode_material: string child 8, electrode_separation_centimeters: double child 9, electrode_separation_precision_centimeters: double child 10, measurements: list<item: struct<time_seconds: double, time_precision_seconds: double, current_amperes: double, cur (... 88 chars omitted) child 0, item: struct<time_seconds: double, time_precision_seconds: double, current_amperes: double, current_precis (... 76 chars omitted) child 0, time_seconds: double child 1, time_precision_seconds: double child 2, current_amperes: double child 3, current_precision_amperes: double child 4, voltage_volts: double child 5, voltage_precision_volts: double child 11, cell: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,DIVIDED_CELL,UNDIVIDED_CELL' child 1, details: string child 5, flow: struct<type: string, details: string, pump_type: string, tubing: struct<type: string, details: strin (... 73 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,PLUG_FLOW_REACTOR,CONTINUOUS_STIRRED_' + 31 child 1, details: string child 2, pump_type: string child 3, tubing: struct<type: string, details: string, diameter_centimeters: double, diameter_precision_centimeters: (... 7 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,STEEL,COPPER,PFA,FEP,TEFLONAF,PTFE,' + 25 child 1, details: string child 2, diameter_centimeters: double child 3, diameter_precision_centimeters: double child 6, reflux: bool child 7, ph: float child 8, conditions_are_dynamic: bool child 9, details: string notes: struct<is_heterogeneous: bool, forms_precipitate: bool, is_exothermic: bool, offgasses: bool, is_sen (... 133 chars omitted) child 0, is_heterogeneous: bool child 1, forms_precipitate: bool child 2, is_exothermic: bool child 3, offgasses: bool child 4, is_sensitive_to_moisture: bool child 5, is_sensitive_to_oxygen: bool child 6, is_sensitive_to_light: bool child 7, safety_notes: string child 8, procedure_details: string observations: list<item: struct<time_seconds: double, time_precision_seconds: double, comment: string, image: stru (... 139 chars omitted) child 0, item: struct<time_seconds: double, time_precision_seconds: double, comment: string, image: struct<float_va (... 127 chars omitted) child 0, time_seconds: double child 1, time_precision_seconds: double child 2, comment: string child 3, image: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string workups: list<item: struct<type: string, details: string, duration_seconds: double, duration_precision_second (... 2903 chars omitted) child 0, item: struct<type: string, details: string, duration_seconds: double, duration_precision_seconds: double, (... 2891 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,ADDITION,ALIQUOT,TEMPERATURE,CONCENTR' + 166 child 1, details: string child 2, duration_seconds: double child 3, duration_precision_seconds: double child 4, input: struct<components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: s (... 2049 chars omitted) child 0, components: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 1105 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC' + 117 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORK' + 70 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED,DRI' + 14 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,N' + 98 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the times' + 4 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 1, crude_components: list<item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: stru (... 292 chars omitted) child 0, item: struct<reaction_id: string, includes_workup: bool, has_derived_amount: bool, amount: struct<mass_gra (... 280 chars omitted) child 0, reaction_id: string child 1, includes_workup: bool child 2, has_derived_amount: bool child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATE' + 1 child 1, details: string child 7, volume_includes_solutes: bool child 4, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_' + 42 child 1, details: string child 2, addition_order: int32 child 3, addition_time_seconds: double child 4, addition_time_precision_seconds: double child 5, addition_speed: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,ALL_AT_ONCE,FAST,SLOW,DROPWISE,CONTINUOU' + 13 child 1, details: string child 6, addition_duration_seconds: double child 7, addition_duration_precision_seconds: double child 8, flow_rate_milliliters_per_minute: double child 9, flow_rate_precision_milliliters_per_minute: double child 10, addition_device: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,SYRINGE,CANNULA,ADDITION_FUN' + 83 child 1, details: string child 11, addition_temperature_kelvin: double child 12, addition_temperature_precision_kelvin: double child 13, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SOLI' + 38 child 1, details: string child 5, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRATED' child 1, details: string child 7, volume_includes_solutes: bool child 6, temperature: struct<control: struct<type: string, details: string>, setpoint_kelvin: double, setpoint_precision_k (... 199 chars omitted) child 0, control: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,AMBIENT,OIL_BATH,WATER_BATH,SAND_' + 79 child 1, details: string child 1, setpoint_kelvin: double child 2, setpoint_precision_kelvin: double child 3, measurements: list<item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: doubl (... 69 chars omitted) child 0, item: struct<type: string, details: string, time_seconds: double, time_precision_seconds: double, temperat (... 57 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,THERMOCOUPLE_INTERNAL,THERMOCOU' + 21 child 1, details: string child 2, time_seconds: double child 3, time_precision_seconds: double child 4, temperature_kelvin: double child 5, temperature_precision_kelvin: double child 7, keep_phase: string child 8, stirring: struct<type: string, details: string, rate: struct<type: string, details: string, rpm: int32>> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,STIR_BAR,OVERHEAD_MIXER,AGITAT' + 27 child 1, details: string child 2, rate: struct<type: string, details: string, rpm: int32> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,HIGH,MEDIUM,LOW' child 1, details: string child 2, rpm: int32 child 9, target_ph: float child 10, is_automated: bool outcomes: list<item: struct<reaction_time_seconds: double, reaction_time_precision_seconds: double, conversion (... 2983 chars omitted) child 0, item: struct<reaction_time_seconds: double, reaction_time_precision_seconds: double, conversion: struct<va (... 2971 chars omitted) child 0, reaction_time_seconds: double child 1, reaction_time_precision_seconds: double child 2, conversion: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 3, products: list<item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, (... 2422 chars omitted) child 0, item: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 2410 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,IUPAC_N' + 115 child 1, details: string child 2, value: string child 3, is_desired_product: bool child 4, measurements: list<item: struct<analysis_key: string, type: string, details: string, uses_internal_standard: bool, (... 1984 chars omitted) child 0, item: struct<analysis_key: string, type: string, details: string, uses_internal_standard: bool, is_normali (... 1972 chars omitted) child 0, analysis_key: string child 1, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,IDENTITY,YIELD,SELECTIVITY,PU' + 33 child 2, details: string child 3, uses_internal_standard: bool child 4, is_normalized: bool child 5, uses_authentic_standard: bool child 6, authentic_standard: struct<smiles: string, structure_id: uint32, identifiers: list<item: struct<type: string, details: s (... 1093 chars omitted) child 0, smiles: string child 1, structure_id: uint32 -- field metadata -- ord.internal: 'joins the dataset's structures artifact' child 2, identifiers: list<item: struct<type: string, details: string, value: string>> child 0, item: struct<type: string, details: string, value: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SMILES,INCHI,MOLBLOCK,I' + 121 child 1, details: string child 2, value: string child 3, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TIT' + 5 child 1, details: string child 7, volume_includes_solutes: bool child 4, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,' + 74 child 5, is_limiting: bool child 6, preparations: list<item: struct<type: string, details: string, reaction_id: string>> child 0, item: struct<type: string, details: string, reaction_id: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,NONE,REPURIFIED,SPARGED' + 18 child 1, details: string child 2, reaction_id: string child 7, source: struct<vendor: string, catalog_id: string, lot: string> child 0, vendor: string child 1, catalog_id: string child 2, lot: string child 8, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_1' + 102 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the t' + 8 child 10, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPH' + 46 child 1, details: string child 7, percentage: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 8, float_value: struct<value: float, precision: float> child 0, value: float child 1, precision: float child 9, string_value: string child 10, amount: struct<mass_grams: double, mass_precision_grams: double, moles_moles: double, moles_precision_moles: (... 146 chars omitted) child 0, mass_grams: double child 1, mass_precision_grams: double child 2, moles_moles: double child 3, moles_precision_moles: double child 4, volume_liters: double child 5, volume_precision_liters: double child 6, unmeasured: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,SATURATED,CATALYTIC,TITRA' + 3 child 1, details: string child 7, volume_includes_solutes: bool child 11, retention_time_seconds: double child 12, retention_time_precision_seconds: double child 13, mass_spec_details: struct<type: string, details: string, tic_minimum_mz: float, tic_maximum_mz: float, eic_masses: list (... 14 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,TIC,TIC_POSITIVE,TIC_NEGATI' + 6 child 1, details: string child 2, tic_minimum_mz: float child 3, tic_maximum_mz: float child 4, eic_masses: list<item: float> child 0, item: float child 14, selectivity: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,EE,ER,DR,EZ,ZE' child 1, details: string child 15, wavelength_nanometers: double child 16, wavelength_precision_nanometers: double child 5, isolated_color: string child 6, texture: struct<type: string, details: string> child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,POWDER,CRYSTAL,OIL,AMORPHOUS_SO' + 40 child 1, details: string child 7, features: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 8, reaction_role: string -- field metadata -- ord.enum: 'UNSPECIFIED,REACTANT,REAGENT,SOLVENT,CATALYST,WORKUP' + 68 child 4, analyses: map<string, struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data (... 295 chars omitted) child 0, entries: struct<key: string not null, value: struct<type: string, details: string, chmo_id: int32, is_of_isol (... 319 chars omitted) not null child 0, key: string not null child 1, value: struct<type: string, details: string, chmo_id: int32, is_of_isolated_species: bool, data: map<string (... 282 chars omitted) child 0, type: string -- field metadata -- ord.enum: 'UNSPECIFIED,CUSTOM,LC,GC,IR,NMR_1H,NMR_13C,NMR_OTH' + 92 child 1, details: string child 2, chmo_id: int32 child 3, is_of_isolated_species: bool child 4, data: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 5, instrument_manufacturer: string child 6, instrument_last_calibrated: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' provenance: struct<experimenter: struct<username: string, name: string, orcid: string, organization: string, ema (... 824 chars omitted) child 0, experimenter: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 1, city: string child 2, experiment_start: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 3, doi: string child 4, patent: string child 5, publication_url: string child 6, record_created: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, person: str (... 105 chars omitted) child 0, time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 1, person: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 2, details: string child 7, record_modified: list<item: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, (... 117 chars omitted) child 0, item: struct<time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string>, person: str (... 105 chars omitted) child 0, time: struct<timestamp: timestamp[us], utc_offset_minutes: int16, value: string> child 0, timestamp: timestamp[us] child 1, utc_offset_minutes: int16 child 2, value: string -- field metadata -- ord.internal: 'the recorded spelling; query the timestamp' child 1, person: struct<username: string, name: string, orcid: string, organization: string, email: string> child 0, username: string child 1, name: string child 2, orcid: string child 3, organization: string child 4, email: string child 2, details: string child 8, reaction_metadata: map<string, struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: stri (... 54 chars omitted) child 0, entries: struct<key: string not null, value: struct<float_value: float, integer_value: int32, bytes_value: bi (... 78 chars omitted) not null child 0, key: string not null child 1, value: struct<float_value: float, integer_value: int32, bytes_value: binary, string_value: string, url: str (... 41 chars omitted) child 0, float_value: float child 1, integer_value: int32 child 2, bytes_value: binary child 3, string_value: string child 4, url: string child 5, description: string child 6, format: string child 9, is_mined: bool reaction_id: string) list[tuple[str, str, DataType]]#
Returns the levels of
schemawhose elements carry a structure.Walked rather than listed by hand, so a compound field added upstream is found wherever it lands – components, products, workup inputs, authentic standards – without anyone updating a list. Every caller that has to agree about where a structure can sit reads this one walk, since two walks are two answers waiting to disagree.
- Parameters:
schema – Schema to walk; the projection schema this library defines by default, or a written file’s
schema_arrowto ask what that file holds.- Returns:
columnis the Parquet physical prefix, carrying thelist.elementandkey_value.valuesegments a file has;pathis the same level as the query grammar names it, which has no wrapper segments; anddtypeis the struct, so a caller can ask what else the element carries.- Return type:
One
(column, path, dtype)per structure-bearing struct, in schema order
- ord_schema.artifacts.structures.write_structures(source: str | PathLike[str], output: str | PathLike[str], *, source_md5: str | None = None, source_dataset_id: str | None = None, compression: str = 'zstd', row_group_size: int = 10000) int#
Derives a structures artifact from a projection and writes it.
Only the smiles and structure_id leaves are decoded, one projection row group at a time. The output is published atomically, so a failure partway leaves any existing artifact untouched.
- Parameters:
source – Path to the projection whose structures to featurize.
output – Path to write the artifact to.
source_md5 – Hash of the source dataset to stamp, if the caller already read one. Taken from the projection’s own stamps when omitted, so the artifact names the dataset it reflects rather than the file it read.
source_dataset_id – Source dataset ID to stamp, if the caller already read one. Taken from the projection’s stamps when omitted.
compression – Parquet codec, any name
pq.ParquetWriteraccepts.row_group_size – Rows per output row group, which is also how many structures are featurized and held in memory at a time.
- Returns:
the number of distinct structures in the projection.
- Return type:
Number of rows written
- Raises:
ValueError – If
sourceis not a projection, or does not state a single dense (smiles, structure_id) mapping.