a

corpusparser/main.py

@@ -21,7 +21,7 @@ lfh.setFormatter(formatter)
 logger.addHandler(lfh)
 logger.setLevel(logging.INFO)
 
-n_chunks = -1  # for logging
+n_chunks = -1
 
 
 def enriched_lemma(token):
@@ -69,7 +69,11 @@ def _handle_kres_file_chunk(kres_file_chunk):
         )
         # dbclient.valdb["kres"]
     for kres_file in kres_chunk:
+        try:
             kres_data = kres_parser.parse_xml_file(kres_file)
+        except:
+            logger.error("Failed to parse file: {}".format(kres_file))
+            continue
         if args.output == "file":
             kres_outdir = outdir / "kres_json"
             kres_outdir.mkdir(parents=True, exist_ok=True)
@@ -78,17 +82,23 @@ def _handle_kres_file_chunk(kres_file_chunk):
                 json.dump(kres_data, fp)
 
         elif args.output == "db":
+            """
             if dbclient.valdb["kres"].find({"sid": kres_data[0]["sid"]}).count() > 0:
                 logger.info("File {} already in DB, closing chunk ({}/{})".format(
                     kres_file, kres_chunk_idx, n_chunks))
                 dbclient.close()
                 return
+            """
 
             kres_data_1 = [_db_preprocess(x) for x in kres_data]
             db_payload += kres_data_1
 
+    try:
         dbclient.valdb["kres"].insert_many(db_payload, ordered=False)  # much much better (just make sure sid has a unique index)
-    logger.info("Inserted kres files chunk ({}/{}) in {:.2f} s".format(
+    except:
+        logger.error("Failed inserting kres files chunk ({}/{}) in {:.2f} s".format(
+            kres_chunk_idx, n_chunks, time.time() - tstart))
+    logger.info("Db insert: chunks ({}/{}) in {:.2f} s".format(
         kres_chunk_idx, n_chunks, time.time() - tstart))
     dbclient.close()