transform and import fixes/tweaks

5 files changed, 92 insertions, 22 deletions

diff --git a/python/fatcat_export.py b/python/fatcat_export.py
index cf8bf1c3..027d6c0a 100755
--- a/python/fatcat_export.py
+++ b/python/fatcat_export.py
@@ -12,9 +12,10 @@ import json
 import argparse
 import fatcat_client
 from fatcat_client.rest import ApiException
-from fatcat_client import ReleaseEntity
+from fatcat_client import ReleaseEntity, ContainerEntity, ChangelogEntry
 from fatcat_tools import uuid2fcid, entity_from_json, entity_to_dict, \
-    release_to_elasticsearch, public_api
+    release_to_elasticsearch, container_to_elasticsearch, \
+    changelog_to_elasticsearch, public_api
 
 
 def run_export_releases(args):
@@ -30,9 +31,27 @@ def run_transform_releases(args):
         line = line.strip()
         if not line:
             continue
-        release = entity_from_json(line, ReleaseEntity)
+        entity = entity_from_json(line, ReleaseEntity)
         args.json_output.write(
-            json.dumps(release_to_elasticsearch(release)) + '\n')
+            json.dumps(release_to_elasticsearch(entity)) + '\n')
+
+def run_transform_containers(args):
+    for line in args.json_input:
+        line = line.strip()
+        if not line:
+            continue
+        entity = entity_from_json(line, ContainerEntity)
+        args.json_output.write(
+            json.dumps(container_to_elasticsearch(entity)) + '\n')
+
+def run_transform_changelogs(args):
+    for line in args.json_input:
+        line = line.strip()
+        if not line:
+            continue
+        entity = entity_from_json(line, ChangelogEntry)
+        args.json_output.write(
+            json.dumps(changelog_to_elasticsearch(entity)) + '\n')
 
 def run_export_changelog(args):
     api = args.api
@@ -74,6 +93,24 @@ def main():
         help="where to send output",
         default=sys.stdout, type=argparse.FileType('w'))
 
+    sub_transform_containers = subparsers.add_parser('transform-containers')
+    sub_transform_containers.set_defaults(func=run_transform_containers)
+    sub_transform_containers.add_argument('json_input',
+        help="JSON-per-line of container entities",
+        default=sys.stdin, type=argparse.FileType('r'))
+    sub_transform_containers.add_argument('json_output',
+        help="where to send output",
+        default=sys.stdout, type=argparse.FileType('w'))
+
+    sub_transform_changelogs = subparsers.add_parser('transform-changelogs')
+    sub_transform_changelogs.set_defaults(func=run_transform_changelogs)
+    sub_transform_changelogs.add_argument('json_input',
+        help="JSON-per-line of changelog entries",
+        default=sys.stdin, type=argparse.FileType('r'))
+    sub_transform_changelogs.add_argument('json_output',
+        help="where to send output",
+        default=sys.stdout, type=argparse.FileType('w'))
+
     sub_changelog = subparsers.add_parser('changelog')
     sub_changelog.set_defaults(func=run_export_changelog)
     sub_changelog.add_argument('--start',
diff --git a/python/fatcat_tools/__init__.py b/python/fatcat_tools/__init__.py
index e2b1e3a2..64c45062 100644
--- a/python/fatcat_tools/__init__.py
+++ b/python/fatcat_tools/__init__.py
@@ -1,4 +1,6 @@
 
 from .api_auth import authenticated_api, public_api
 from .fcid import fcid2uuid, uuid2fcid
-from .transforms import entity_to_dict, entity_from_json, release_to_elasticsearch
+from .transforms import entity_to_dict, entity_from_json, \
+    release_to_elasticsearch, container_to_elasticsearch, \
+    changelog_to_elasticsearch
diff --git a/python/fatcat_tools/importers/common.py b/python/fatcat_tools/importers/common.py
index ebdce56f..a29b3019 100644
--- a/python/fatcat_tools/importers/common.py
+++ b/python/fatcat_tools/importers/common.py
@@ -236,8 +236,8 @@ class EntityImporter:
         self._entity_queue.append(entity)
         if len(self._entity_queue) >= self.edit_batch_size:
             self.insert_batch(self._entity_queue)
-            self.counts['insert'] += len(_entity_queue)
-            self._entity_queue = 0
+            self.counts['insert'] += len(self._entity_queue)
+            self._entity_queue = []
 
     def want(self, raw_record):
         """
diff --git a/python/fatcat_tools/importers/journal_metadata.py b/python/fatcat_tools/importers/journal_metadata.py
index 7f6b1ee8..be62d63a 100644
--- a/python/fatcat_tools/importers/journal_metadata.py
+++ b/python/fatcat_tools/importers/journal_metadata.py
@@ -44,7 +44,7 @@ class JournalMetadataImporter(EntityImporter):
             editgroup_extra=eg_extra)
 
     def want(self, raw_record):
-        if raw_record.get('issnl'):
+        if raw_record.get('issnl') and raw_record.get('name'):
             return True
         return False
 
@@ -55,6 +55,10 @@ class JournalMetadataImporter(EntityImporter):
         returns a ContainerEntity (or None if invalid or couldn't parse)
         """
 
+        if not row.get('name'):
+            # Name is required (by schema)
+            return None
+
         extra = dict()
         for key in ('issne', 'issnp', 'languages', 'country', 'urls', 'abbrev',
             'coden', 'aliases', 'original_name', 'first_year', 'last_year',
@@ -76,8 +80,10 @@ class JournalMetadataImporter(EntityImporter):
         extra_ia = dict()
         # TODO: would like an ia.longtail_ia flag
         if row.get('sim'):
+            # NB: None case of the .get() here is blech, but othrwise
+            # extra['ia'].get('sim') would be false-y, breaking 'any_ia_sim' later on
             extra_ia['sim'] = {
-                'year_spans': row['sim']['year_spans'],
+                'year_spans': row['sim'].get('year_spans'),
             }
         if extra_ia:
             extra['ia'] = extra_ia
diff --git a/python/fatcat_tools/transforms.py b/python/fatcat_tools/transforms.py
index a85c877c..7bb75c3e 100644
--- a/python/fatcat_tools/transforms.py
+++ b/python/fatcat_tools/transforms.py
@@ -231,20 +231,12 @@ def container_to_elasticsearch(entity):
         container_type = entity.container_type,
         issnl = entity.issnl,
         wikidata_qid = entity.wikidata_qid,
-
-        entity_status = entity.entity_status,
-        language = entity.language,
-        license = entity.license_slug,
-        doi = entity.doi,
-        pmid = entity.pmid,
-        isbn13 = entity.isbn13,
-        core_id = entity.core_id,
-        arxiv_id = entity.core_id,
-        jstor_id = entity.jstor_id,
     )
 
     # TODO: region, discipline
     # TODO: single primary language?
+    if not entity.extra:
+        entity.extra = dict()
     for key in ('country', 'languages', 'mimetypes', 'first_year', 'last_year'):
         if entity.extra.get(key):
             t[key] = entity.extra[key]
@@ -285,13 +277,46 @@ def container_to_elasticsearch(entity):
         if extra['ia'].get('sim'):
             any_ia_sim = True
 
-    t['in_doaj'] = is_doaj
-    t['in_road'] = is_road
+    t['in_doaj'] = in_doaj
+    t['in_road'] = in_road
     t['in_doi'] = in_doi
     t['in_sherpa_romeo'] = in_sherpa_romeo
-    t['is_oa'] = in_doaj or in_road or is_longtail_oa or ia_oa
+    t['is_oa'] = in_doaj or in_road or is_longtail_oa or is_oa
     t['is_longtail_oa'] = is_longtail_oa
     t['any_kbart'] = any_ia_sim
     t['any_jstor'] = any_ia_sim
     t['any_ia_sim'] = bool(any_ia_sim)
     return t
+
+
+def changelog_to_elasticsearch(entity):
+
+    editgroup = entity.editgroup
+    t = dict(
+        index=entity.index,
+        editgroup_id=entity.editgroup_id,
+        timestamp=entity.timestamp,
+        editor_id=editgroup.editor_id,
+    )
+
+    extra = editgroup.extra or dict()
+    if extra.get('agent'):
+        t['agent'] = extra['agent']
+
+    t['containers'] = len(editgroup.edits.containers)
+    t['creators'] = len(editgroup.edits.containers)
+    t['files'] = len(editgroup.edits.containers)
+    t['filesets'] = len(editgroup.edits.containers)
+    t['webcaptures'] = len(editgroup.edits.containers)
+    t['releases'] = len(editgroup.edits.containers)
+    t['works'] = len(editgroup.edits.containers)
+
+    # TODO: parse and pull out counts
+    #created = 0
+    #updated = 0
+    #deleted = 0
+    #t['created'] = created
+    #t['updated'] = updated
+    #t['deleted'] = deleted
+    #t['total'] = created + updated + deleted
+    return t