Files
threshing-atlas/collection/publish_metadata.py
T
ekcSchianoandClaude Opus 5.5 63b5d40e54 Step 5: data dictionary and Croissant metadata
- data/dictionary.yaml: every published file, and every column of runs_clean.csv
  with its type, meaning and allowed values
- collection/publish_metadata.py writes data/croissant.json (Croissant 1.0) from
  it, with a sha256 per file; `npm run metadata`, run by the daily update
- tests/core/test_dictionary.py: every column and value described, types hold,
  croissant.json up to date and checksums match
- attempt is now written as a whole number (was 1.0, 2.0); no statistic changes
- the site ships both files and the Methods text links them

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UCzTd31GphpMXDhQqWrCMQ
2026-11-04 01:37:10 +00:00

164 lines
5.2 KiB
Python

"""Write data/croissant.json (Croissant 1.0 JSON-LD) from data/dictionary.yaml.
Croissant is the MLCommons metadata format for datasets: it says what each file is, its checksum, and what each
column means, so tools (and people) can load and cite the data without reading this repo's code.
python3 collection/publish_metadata.py # write data/croissant.json
python3 collection/publish_metadata.py --check # exit 1 if data/croissant.json is out of date
"""
import hashlib
import json
import sys
from pathlib import Path
import yaml
ROOT = Path(__file__).resolve().parents[1]
DICTIONARY = ROOT / "data" / "dictionary.yaml"
SOURCES = ROOT / "data" / "sources.json"
OUT = ROOT / "data" / "croissant.json"
CONTEXT = {
"@language": "en",
"@vocab": "https://schema.org/",
"citeAs": "cr:citeAs",
"column": "cr:column",
"conformsTo": "dct:conformsTo",
"cr": "http://mlcommons.org/croissant/",
"data": {"@id": "cr:data", "@type": "@json"},
"dataType": {"@id": "cr:dataType", "@type": "@vocab"},
"dct": "http://purl.org/dc/terms/",
"examples": {"@id": "cr:examples", "@type": "@json"},
"extract": "cr:extract",
"field": "cr:field",
"fileObject": "cr:fileObject",
"fileProperty": "cr:fileProperty",
"format": "cr:format",
"includes": "cr:includes",
"isLiveDataset": "cr:isLiveDataset",
"jsonPath": "cr:jsonPath",
"key": "cr:key",
"md5": "cr:md5",
"parentField": "cr:parentField",
"path": "cr:path",
"recordSet": "cr:recordSet",
"references": "cr:references",
"regex": "cr:regex",
"repeated": "cr:repeated",
"replace": "cr:replace",
"sc": "https://schema.org/",
"separator": "cr:separator",
"source": "cr:source",
"subField": "cr:subField",
"transform": "cr:transform",
}
# Croissant has no boolean-as-text type; the CSV writes True/False, which sc:Boolean parses.
DATA_TYPES = {"text": "sc:Text", "integer": "sc:Integer", "boolean": "sc:Boolean", "date": "sc:Date"}
def sha256(path):
h = hashlib.sha256()
with open(path, "rb") as f:
for block in iter(lambda: f.read(1 << 20), b""):
h.update(block)
return h.hexdigest()
def load_dictionary():
with open(DICTIONARY, encoding="utf-8") as f:
return yaml.safe_load(f)
def field_description(spec):
text = " ".join(spec["description"].split())
if spec.get("values"):
text += " Values: " + "; ".join(str(v) for v in spec["values"]) + "."
return text
def build(dictionary=None):
d = dictionary or load_dictionary()
ds = d["dataset"]
with open(SOURCES, encoding="utf-8") as f:
sources = json.load(f)
distribution = []
for key, spec in d["files"].items():
path = ROOT / spec["path"]
entry = {
"@type": "cr:FileObject",
"@id": key,
"name": key,
"description": " ".join(spec["description"].split()),
"contentUrl": spec.get("url") or f"{ds['repository']}/blob/main/{spec['path']}",
"encodingFormat": spec["format"],
"sha256": sha256(path),
"contentSize": f"{path.stat().st_size} B",
}
distribution.append(entry)
fields = [
{
"@type": "cr:Field",
"@id": f"runs/{name}",
"name": name,
"description": field_description(spec),
"dataType": DATA_TYPES[spec["type"]],
"source": {"fileObject": {"@id": "runs_clean"}, "extract": {"column": name}},
}
for name, spec in d["columns"].items()
]
return {
"@context": CONTEXT,
"@type": "sc:Dataset",
"conformsTo": "http://mlcommons.org/croissant/1.0",
"name": ds["name"],
"description": " ".join(ds["description"].split()),
"url": ds["url"],
"sameAs": ds["repository"],
"license": ds["license"],
"creator": {"@type": "sc:Person", "name": ds["creator"], "url": f"https://github.com/{ds['creator']}"},
"keywords": ds["keywords"],
"citeAs": ds["citation"],
"isLiveDataset": True,
"isBasedOn": [
{"@type": "sc:CreativeWork", "name": s["name"], "url": s["view"]}
for key, s in sources.items()
if key != "pdf" # the route PDF is only fingerprinted, never redistributed
],
"distribution": distribution,
"recordSet": [
{
"@type": "cr:RecordSet",
"@id": "runs",
"name": "runs",
"description": "One coded threshing run per row (data/derived/runs_clean.csv).",
"key": {"@id": "runs/run_id"},
"field": fields,
}
],
}
def render(doc):
return json.dumps(doc, indent=2, ensure_ascii=False) + "\n"
def main(argv):
text = render(build())
if "--check" in argv:
if not OUT.exists() or OUT.read_text(encoding="utf-8") != text:
print("data/croissant.json is out of date: run python3 collection/publish_metadata.py", file=sys.stderr)
return 1
print("data/croissant.json is up to date")
return 0
OUT.write_text(text, encoding="utf-8")
print(f"wrote {OUT.relative_to(ROOT)}")
return 0
if __name__ == "__main__":
sys.exit(main(sys.argv[1:]))