initial implementation of file_meta importer

3 files changed, 86 insertions, 0 deletions

diff --git a/python/fatcat_import.py b/python/fatcat_import.py
index 252ab3a5..498683e0 100755
--- a/python/fatcat_import.py
+++ b/python/fatcat_import.py
@@ -236,6 +236,11 @@ def run_datacite(args):
     else:
         JsonLinePusher(dci, args.json_file).run()
 
+def run_file_meta(args):
+    fmi = FileMetaImporter(args.api,
+        edit_batch_size=100)
+    JsonLinePusher(fmi, args.json_file).run()
+
 def main():
     parser = argparse.ArgumentParser(
         formatter_class=argparse.ArgumentDefaultsHelpFormatter)
@@ -556,6 +561,16 @@ def main():
         auth_var="FATCAT_AUTH_WORKER_DATACITE",
     )
 
+    sub_file_meta = subparsers.add_parser('file-meta',
+        help="simple update-only importer for file metadata")
+    sub_file_meta.set_defaults(
+        func=run_file_meta,
+        auth_var="FATCAT_API_AUTH_TOKEN",
+    )
+    sub_file_meta.add_argument('json_file',
+        help="File with jsonlines from file_meta schema to import from",
+        default=sys.stdin, type=argparse.FileType('r'))
+
     args = parser.parse_args()
     if not args.__dict__.get("func"):
         print("tell me what to do!")
diff --git a/python/fatcat_tools/importers/__init__.py b/python/fatcat_tools/importers/__init__.py
index c26446fd..b82eb11a 100644
--- a/python/fatcat_tools/importers/__init__.py
+++ b/python/fatcat_tools/importers/__init__.py
@@ -29,3 +29,4 @@ from .wayback_static import auto_wayback_static
 from .cdl_dash_dat import auto_cdl_dash_dat
 from .ingest import IngestFileResultImporter, SavePaperNowFileImporter
 from .shadow import ShadowLibraryImporter
+from .file_meta import FileMetaImporter
diff --git a/python/fatcat_tools/importers/file_meta.py b/python/fatcat_tools/importers/file_meta.py
new file mode 100644
index 00000000..c3728570
--- /dev/null
+++ b/python/fatcat_tools/importers/file_meta.py
@@ -0,0 +1,70 @@
+
+import fatcat_openapi_client
+from .common import EntityImporter
+
+
+class FileMetaImporter(EntityImporter):
+    """
+    The purpose of this importer is to update file-level metadata for file
+    entities that are missing some fields.
+
+    It should *only* update entities, never create (insert) them.
+
+    In particular, during early boostramping over 18 million file entities were
+    imported which were missing file size, mimetype, md5, and/or sha256.
+    """
+
+    def __init__(self, api, require_grobid=True, **kwargs):
+
+        eg_desc = kwargs.pop('editgroup_description', None) or "File metadata updates"
+        eg_extra = kwargs.pop('editgroup_extra', dict())
+        eg_extra['agent'] = eg_extra.get('agent', 'fatcat_tools.FileMetaImporter')
+        kwargs['do_updates'] = kwargs.get("do_updates", True)
+        super().__init__(api,
+            editgroup_description=eg_desc,
+            editgroup_extra=eg_extra,
+            **kwargs)
+
+    def want(self, row):
+        for k in ('sha1hex', 'sha256hex', 'md5hex', 'size_bytes', 'mimetype'):
+            if not row.get(k):
+                return False
+        return True
+
+    def parse_record(self, row):
+
+        file_meta = row
+        fe = fatcat_openapi_client.FileEntity(
+            md5=file_meta['md5hex'],
+            sha1=file_meta['sha1hex'],
+            sha256=file_meta['sha256hex'],
+            size=file_meta['size_bytes'],
+            mimetype=file_meta['mimetype'],
+        )
+        return fe
+
+    def try_update(self, fe):
+        # lookup sha1, or create new entity
+        existing = None
+        try:
+            existing = self.api.lookup_file(sha1=fe.sha1)
+        except fatcat_openapi_client.rest.ApiException as err:
+            if err.status != 404:
+                raise err
+
+        if not existing:
+            self.counts['skip-no-match']
+            return False
+
+        if (existing.md5 and existing.sha256 and existing.size_bytes and existing.mimetype):
+            self.counts['skip-existing-complete']
+            return False
+
+        existing.md5 = existing.md5 or fe.md5
+        existing.sha256 = existing.sha256 or fe.sha256
+        existing.file_bytes = existing.file_bytes or fe.file_bytes
+        existing.mimetype = existing.mimetype or fe.mimetype
+
+        self.api.update_container(self.get_editgroup_id(), existing.ident, existing)
+        self.counts['update'] += 1
+        return False