separate parsing of ssj and kres (mem management)

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")
-        links = [x for x in lgrps if x.get("type") == links_type][0]
+        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 +9,18 @@ import sys
 from multiprocessing import Pool
 import time
 
-logging.basicConfig(level=logging.INFO)
+CORPORA = ["kres", "ssj"]
+
+# logging.basicConfig(filename=Path("/var/tmp/corpusparser.log"), filemode='a', level=logging.INFO)
 logger = logging.getLogger(__name__)
 
+lfh = logging.FileHandler("/project/logs/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 +82,13 @@ 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):
@@ -92,11 +101,13 @@ def _get_dbclient(args):
     )
     return dbclient
 
+
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Parsing corpora kres and ssj500k.")
-    parser.add_argument('--kres-folder', required=True)
+    parser.add_argument('--corpus', required=True)
-    parser.add_argument('--kres-srl-folder', required=True)
+    parser.add_argument('--kres-folder', required=False)
-    parser.add_argument('--ssj-file', required=True)
+    parser.add_argument('--kres-srl-folder', required=False)
+    parser.add_argument('--ssj-file', required=False)
     parser.add_argument('--output', required=False, default=None)
     parser.add_argument('--outdir', required=False, default=None)
     parser.add_argument('--dbaddr', required=False, default=None)
@@ -105,48 +116,55 @@ if __name__ == "__main__":
     parser.add_argument('--cores', required=False, default=1)
     args = parser.parse_args()
 
+
+    corpus = args.corpus
+    assert (corpus in CORPORA), "Wrong corpus name."
+
+
     outdir = None
     if args.output == "file":
         outdir = Path(args.outdir)
         outdir.mkdir(parents=True, exist_ok=True)
     elif args.output == "db":
-        # Force unique sid
         dbclient = _get_dbclient(args)
-        for corpus in ["kres", "ssj"]:
+        dbclient.valdb[corpus].ensure_index([("sid", pymongo.ASCENDING)])
-            dbclient.valdb[corpus].ensure_index([("sid", pymongo.ASCENDING)])
+        dbclient.valdb[corpus].ensure_index([("headwords", pymongo.ASCENDING)])
-            dbclient.valdb[corpus].ensure_index([("headwords", pymongo.ASCENDING)])
+        dbclient.valdb[corpus].ensure_index([("functors", 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)
 
 
-    # Kres
+    if corpus == "ssj":
-    logger.info("Parsing Kres: {}".format(args.kres_folder))
+        logger.info("Parsing Ssj: {}".format(args.ssj_file))
-    kres_parser = Parser(
+        ssj_parser = Parser(logger=logger, corpus="ssj")
-        corpus="kres",
+        ssj_data = ssj_parser.parse_xml_file(Path(args.ssj_file))
-        kres_srl_folder=args.kres_srl_folder
+        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)
 
-    # [(idx, filepath)]
-    kres_files = [x for x in Path(args.kres_folder).iterdir()]
-    kres_files = [x for x in enumerate(kres_files)]
-    n_kres_files = len(kres_files)
 
-    p = Pool(int(args.cores))
+    if corpus == "kres":
-    p.map(_handle_kres_file_tpl, kres_files)
+        # Kres
+        logger.info("Parsing Kres: {}".format(args.kres_folder))
+        kres_parser = Parser(
+            logger=logger,
+            corpus="kres",
+            kres_srl_folder=args.kres_srl_folder
+        )
+
+        # [(idx, filepath)]
+        kres_files = [x for x in Path(args.kres_folder).iterdir()]
+        kres_files = [x for x in enumerate(kres_files)]
+        n_kres_files = len(kres_files)
+
+        p = Pool(int(args.cores))
+        p.map(_handle_kres_file_tpl, kres_files)
+
     
     logger.info("Finished parsing.")