moved parse_ssj into a subprocess

corpusparser/Parser.py

@@ -5,7 +5,7 @@ from lxml import etree
 import logging
 import time
 
-logging.basicConfig(level=logging.INFO)
+# logging.basicConfig(level=logging.INFO)
 
 # Read input file(.xml, .json; kres or ssj500k).  
 # Create an iterator that outputs resulting sentences (python dict format).  
@@ -61,7 +61,10 @@ class Parser():
 
     def parse_any_links_ssj(self, sent_el, links_type):
         lgrps = sent_el.findall(".//linkGrp")
+        try:
             links = [x for x in lgrps if x.get("type") == links_type][0]
+        except:
+            return []
         res_links = []
         for link in links:
             tar = self.parse_ssj_target_arg(link.get("target"))

corpusparser/main.py

@@ -9,9 +9,16 @@ import sys
 from multiprocessing import Pool
 import time
 
-logging.basicConfig(level=logging.INFO)
+# logging.basicConfig(filename=Path("/var/tmp/corpusparser.log"), filemode='a', level=logging.INFO)
 logger = logging.getLogger(__name__)
 
+# lfh = logging.FileHandler("/var/tmp/fill-database.log")
+lfh = logging.StreamHandler(sys.stdout)
+formatter = logging.Formatter("%(asctime)s %(levelname)s %(message)s")
+lfh.setFormatter(formatter)
+logger.addHandler(lfh)
+logger.setLevel(logging.INFO)
+
 n_kres_files = -1  # for logging
 
 
@@ -73,13 +80,13 @@ def _handle_kres_file_tpl(kres_file_tpl):
 
         # skip if one of the sentences is already in DB
         if kres_col.find({"sid": kres_data[0]["sid"]}).count() > 0:
-            logging.info("File {} already in DB ({}/{})".format(
+            logger.info("File {} already in DB ({}/{})".format(
                 kres_file, kres_file_idx, n_kres_files))
             return
 
         kres_data_1 = [_db_preprocess(x) for x in kres_data]
         kres_col.insert_many(kres_data_1)  # much much better (just make sure sid has a unique index)
-        logging.info("Inserted data from {} ({}/{}) in {:.2f} s".format(
+        logger.info("Inserted data from {} ({}/{}) in {:.2f} s".format(
             kres_file, kres_file_idx, n_kres_files, time.time() - tstart))
 
 def _get_dbclient(args):
@@ -92,6 +99,27 @@ def _get_dbclient(args):
     )
     return dbclient
 
+
+# wrap it in a function for better garbage collection
+def parse_ssj(args):
+    logger.info("Parsing Ssj: {}".format(args.ssj_file))
+    ssj_parser = Parser(logger=logger, corpus="ssj")
+    ssj_data = ssj_parser.parse_xml_file(Path(args.ssj_file))
+    if args.output == "file":
+        ssj_outfile = outdir / "ssj500k.json"
+        with ssj_outfile.open("w") as fp:
+            json.dump(ssj_data, fp)
+    elif args.output == "db":
+        dbclient = _get_dbclient(args)
+        valdb = dbclient.valdb
+        ssj_col = valdb["ssj"]
+        for sentence in ssj_data:
+            sentence = _db_preprocess(sentence)
+            ssj_col.update({"sid": sentence["sid"]}, sentence, upsert=True)
+    del ssj_parser
+    del ssj_data
+
+
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Parsing corpora kres and ssj500k.")
     parser.add_argument('--kres-folder', required=True)
@@ -117,26 +145,17 @@ if __name__ == "__main__":
             dbclient.valdb[corpus].ensure_index([("headwords", pymongo.ASCENDING)])
             dbclient.valdb[corpus].ensure_index([("functors", pymongo.ASCENDING)])
 
-    # SSJ
-    logger.info("Parsing Ssj: {}".format(args.ssj_file))
-    ssj_parser = Parser(corpus="ssj")
-    ssj_data = ssj_parser.parse_xml_file(Path(args.ssj_file))
-    if args.output == "file":
-        ssj_outfile = outdir / "ssj500k.json"
-        with ssj_outfile.open("w") as fp:
-            json.dump(ssj_data, fp)
-    elif args.output == "db":
-        dbclient = _get_dbclient(args)
-        valdb = dbclient.valdb
-        ssj_col = valdb["ssj"]
-        for sentence in ssj_data:
-            sentence = _db_preprocess(sentence)
-            ssj_col.update({"sid": sentence["sid"]}, sentence, upsert=True)
 
+    # SSJ
+    p = Pool(1)
+    p.map(parse_ssj, [args])
+
+    time.sleep(30)
 
     # Kres
     logger.info("Parsing Kres: {}".format(args.kres_folder))
     kres_parser = Parser(
+        logger=logger,
         corpus="kres",
         kres_srl_folder=args.kres_srl_folder
     )