added adjective handling (appending _ to headwords)

corpusparser/Parser.py

@@ -107,12 +107,12 @@ class Parser():
             yield from self.xml_file_to_generator(self.ssj_file)
 
     def parse_xml_file(self, xml_file):
-        tstart = time.time()
+        # tstart = time.time()
         file_data = []
         for tpl in self.xml_file_to_generator(xml_file):
             file_data += [tpl[1]]
         tend = time.time()
-        self.logger.info("Parsed {} in {:.4f} s".format(xml_file, tend - tstart))
+        # self.logger.info("Parsed {} in {:.4f} s".format(xml_file, tend - tstart))
         return file_data
 
     def xml_file_to_generator(self, xml_file):
@@ -211,7 +211,7 @@ class Parser():
                         "text": sentence_text,
                         "tokens": sentence_tokens,
                         "jos_links": jos_links,
-                        "srl_links": srl_links_parsed
+                        "srl_links": srl_links_parsed,
                     }
                     self.stats["parsed_count"] += 1
                     yield (xml_file, sentence_entry)

corpusparser/__init__.py

@@ -1 +1,2 @@
 from corpusparser.Parser import Parser
+from corpusparser.main import enriched_lemma

corpusparser/main.py

@@ -4,20 +4,48 @@ import argparse
 import logging
 import json
 from pymongo import MongoClient
+import pymongo
 import sys
 from multiprocessing import Pool
+import time
 
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 n_kres_files = -1  # for logging
 
+
+def enriched_lemma(token):
+    return (token["lemma"] if token["msd"][0] == "G" else token["lemma"] + "_")
+
+
+def _helper_tid_to_token(tid, tokens):
+    for t in tokens:
+        if t["tid"] == tid:
+            return t
+    return None
+
+
+def _db_preprocess(e):
+    if e["srl_links"] is None:
+        e["headwords"] = []
+        e["functors"] = []
+    else:
+        hw_tids = list(set([x["from"] for x in e["srl_links"]]))
+        hw_tokens = [_helper_tid_to_token(tid, e["tokens"]) for tid in hw_tids]
+        headwords = [enriched_lemma(t) for t in hw_tokens]
+        e["headwords"] = headwords
+
+        functors = list(set([x["afun"] for x in e["srl_links"]]))
+        e["functors"] = functors
+    return e
+
+
 # handler for concurrency
 def _handle_kres_file_tpl(kres_file_tpl):
+    tstart = time.time()
     kres_file_idx = kres_file_tpl[0]
     kres_file = kres_file_tpl[1]
-    logging.info("Handling {} ({}/{})".format(
-        kres_file, kres_file_idx, n_kres_files))
     kres_data = kres_parser.parse_xml_file(kres_file)
     if args.output == "file":
         kres_outdir = outdir / "kres_json"
@@ -26,9 +54,43 @@ def _handle_kres_file_tpl(kres_file_tpl):
         with kres_outfile.open("w") as fp:
             json.dump(kres_data, fp)
     elif args.output == "db":
+        # mongoclient needs to be created after forking
+        dbclient = MongoClient(
+            "mongodb://{}".format(args.dbaddr),
+            username=args.dbuser,
+            password=args.dbpass,
+            authSource="valdb",
+            authMechanism='SCRAM-SHA-1'
+        )
+        valdb = dbclient.valdb
         kres_col = valdb["kres"]
+
+        # HUUUUGE BOTTLENECK
+        """
         for sentence in kres_data:
             kres_col.update({"sid": sentence["sid"]}, sentence, upsert=True)
+        """
+
+        # 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(
+                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(
+            kres_file, kres_file_idx, n_kres_files, time.time() - tstart))
+
+def _get_dbclient(args):
+    dbclient = MongoClient(
+        "mongodb://{}".format(args.dbaddr),
+        username=args.dbuser,
+        password=args.dbpass,
+        authSource="valdb",
+        authMechanism='SCRAM-SHA-1'
+    )
+    return dbclient
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser(description="Parsing corpora kres and ssj500k.")
@@ -44,19 +106,16 @@ if __name__ == "__main__":
     args = parser.parse_args()
 
     outdir = None
-    valdb = None
     if args.output == "file":
         outdir = Path(args.outdir)
         outdir.mkdir(parents=True, exist_ok=True)
     elif args.output == "db":
-        dbclient = MongoClient(
-            "mongodb://{}".format(args.dbaddr),
-            username=args.dbuser,
-            password=args.dbpass,
-            authSource="valdb",
-            authMechanism='SCRAM-SHA-1'
-        )
-        valdb = dbclient.valdb
+        # 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([("headwords", pymongo.ASCENDING)])
+            dbclient.valdb[corpus].ensure_index([("functors", pymongo.ASCENDING)])
 
     # SSJ
     logger.info("Parsing Ssj: {}".format(args.ssj_file))
@@ -67,8 +126,11 @@ if __name__ == "__main__":
         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)