creative importer for bulk JSTOR imports

2 files changed, 23 insertions, 1 deletions

diff --git a/python/fatcat_tools/importers/__init__.py b/python/fatcat_tools/importers/__init__.py
index d0c6656c..663b9812 100644
--- a/python/fatcat_tools/importers/__init__.py
+++ b/python/fatcat_tools/importers/__init__.py
@@ -12,7 +12,7 @@ To run an import you combine two classes; one each of:
 
 """
 
-from .common import EntityImporter, JsonLinePusher, LinePusher, CsvPusher, SqlitePusher, Bs4XmlFilePusher, Bs4XmlLinesPusher, KafkaJsonPusher, make_kafka_consumer, clean, is_cjk, LANG_MAP_MARC
+from .common import EntityImporter, JsonLinePusher, LinePusher, CsvPusher, SqlitePusher, Bs4XmlFilePusher, Bs4XmlLinesPusher, Bs4XmlFileListPusher, KafkaJsonPusher, make_kafka_consumer, clean, is_cjk, LANG_MAP_MARC
 from .crossref import CrossrefImporter, CROSSREF_TYPE_MAP, lookup_license_slug
 from .jalc import JalcImporter
 from .jstor import JstorImporter
diff --git a/python/fatcat_tools/importers/common.py b/python/fatcat_tools/importers/common.py
index 4d7b29fb..2a434693 100644
--- a/python/fatcat_tools/importers/common.py
+++ b/python/fatcat_tools/importers/common.py
@@ -606,6 +606,28 @@ class Bs4XmlFilePusher(RecordPusher):
         return counts
 
 
+class Bs4XmlFileListPusher(RecordPusher):
+
+    def __init__(self, importer, list_file, record_tag, **kwargs):
+        self.importer = importer
+        self.list_file = list_file
+        self.record_tag = record_tag
+
+    def run(self):
+        for xml_path in self.list_file:
+            xml_path = xml_path.strip()
+            if not xml_path or xml_path.startswith("#"):
+                continue
+            with open(xml_path, 'r') as xml_file:
+                soup = BeautifulSoup(xml_file, "xml")
+                for record in soup.find_all(self.record_tag):
+                    self.importer.push_record(record)
+                soup.decompose()
+        counts = self.importer.finish()
+        print(counts)
+        return counts
+
+
 class KafkaJsonPusher(RecordPusher):
 
     def __init__(self, importer, kafka_hosts, kafka_env, topic_suffix, group, **kwargs):