summaryrefslogtreecommitdiffstats
path: root/python/fatcat_tools
diff options
context:
space:
mode:
authorBryan Newbold <bnewbold@robocracy.org>2020-11-05 17:19:37 -0800
committerBryan Newbold <bnewbold@robocracy.org>2020-11-05 17:19:37 -0800
commit8b6b1447cc37fb76865fd80377c55463e59db3b9 (patch)
tree42b2f26171db65e10d786f0bc45cccac79ae8b07 /python/fatcat_tools
parent5afde4690a4653db53fe4962af5da3eb9188d9a2 (diff)
downloadfatcat-8b6b1447cc37fb76865fd80377c55463e59db3b9.tar.gz
fatcat-8b6b1447cc37fb76865fd80377c55463e59db3b9.zip
ingest: basic checks for ingest_type
Diffstat (limited to 'python/fatcat_tools')
-rw-r--r--python/fatcat_tools/importers/ingest.py32
1 files changed, 29 insertions, 3 deletions
diff --git a/python/fatcat_tools/importers/ingest.py b/python/fatcat_tools/importers/ingest.py
index 4b1d3702..c88ec86a 100644
--- a/python/fatcat_tools/importers/ingest.py
+++ b/python/fatcat_tools/importers/ingest.py
@@ -20,7 +20,7 @@ class IngestFileResultImporter(EntityImporter):
assert self.default_link_rel
self.require_grobid = require_grobid
if self.require_grobid:
- print("Requiring GROBID status == 200")
+ print("Requiring GROBID status == 200 (for PDFs)")
else:
print("NOT checking GROBID success")
self.ingest_request_source_whitelist = [
@@ -74,8 +74,22 @@ class IngestFileResultImporter(EntityImporter):
if not row.get('file_meta'):
self.counts['skip-file-meta'] += 1
return False
- if self.require_grobid and row.get('grobid', {}).get('status_code') != 200:
- self.counts['skip-grobid'] += 1
+
+ # type-specific filters
+ if row['request'].get('ingest_type') == 'pdf':
+ if self.require_grobid and row.get('grobid', {}).get('status_code') != 200:
+ self.counts['skip-grobid'] += 1
+ return False
+ if row['file_meta'].get('mimetype') not in ("application/pdf",):
+ self.counts['skip-mimetype'] += 1
+ return False
+ elif row['request'].get('ingest_type') == 'xml':
+ if row['file_meta'].get('mimetype') not in ("application/xml",
+ "application/jats+xml", "application/tei+xml", "text/xml"):
+ self.counts['skip-mimetype'] += 1
+ return False
+ else:
+ self.counts['skip-ingest-type'] += 1
return False
return True
@@ -85,6 +99,18 @@ class IngestFileResultImporter(EntityImporter):
request = row['request']
fatcat = request.get('fatcat')
file_meta = row['file_meta']
+
+ # double check that want() filtered request correctly (eg, old requests)
+ if request.get('ingest_type') not in ('pdf', 'xml'):
+ self.counts['skip-ingest-type'] += 1
+ return None
+ assert (request['ingest_type'], file_meta['mimetype']) in [
+ ("pdf", "application/pdf"),
+ ("xml", "application/xml"),
+ ("xml", "application/jats+xml"),
+ ("xml", "application/tei+xml"),
+ ("xml", "text/xml"),
+ ]
# identify release by fatcat ident, or extid lookup, or biblio-glutton match
release_ident = None