kristjan/cjvt-corpusparser
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge branch 'master' of gitea.cjvt.si:kristjan/cjvt-corpusparser into my-fix

Browse Source
This commit is contained in:
voje 2019-04-21 16:07:15 +02:00
commit 14c607c106
2 changed files with 13 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
corpusparser/Parser.py
View File

@ -5,7 +5,7 @@ from lxml import etree
import logging import logging
import time import time
logging.basicConfig(level=logging.INFO) # logging.basicConfig(level=logging.INFO)
# Read input file(.xml, .json; kres or ssj500k). # Read input file(.xml, .json; kres or ssj500k).
# Create an iterator that outputs resulting sentences (python dict format). # Create an iterator that outputs resulting sentences (python dict format).

16
corpusparser/main.py
View File

@ -9,9 +9,16 @@ import sys
from multiprocessing import Pool from multiprocessing import Pool
import time import time
logging.basicConfig(level=logging.INFO) # logging.basicConfig(filename=Path("/var/tmp/corpusparser.log"), filemode='a', level=logging.INFO)
logger = logging.getLogger(__name__) 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 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 # skip if one of the sentences is already in DB
if kres_col.find({"sid": kres_data[0]["sid"]}).count() > 0: 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)) kres_file, kres_file_idx, n_kres_files))
return return
kres_data_1 = [_db_preprocess(x) for x in kres_data] 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) 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)) kres_file, kres_file_idx, n_kres_files, time.time() - tstart))
def _get_dbclient(args): def _get_dbclient(args):
@ -119,7 +126,7 @@ if __name__ == "__main__":
# SSJ # SSJ
logger.info("Parsing Ssj: {}".format(args.ssj_file)) logger.info("Parsing Ssj: {}".format(args.ssj_file))
ssj_parser = Parser(corpus="ssj") ssj_parser = Parser(logger=logger, corpus="ssj")
ssj_data = ssj_parser.parse_xml_file(Path(args.ssj_file)) ssj_data = ssj_parser.parse_xml_file(Path(args.ssj_file))
if args.output == "file": if args.output == "file":
ssj_outfile = outdir / "ssj500k.json" ssj_outfile = outdir / "ssj500k.json"
@ -137,6 +144,7 @@ if __name__ == "__main__":
# Kres # Kres
logger.info("Parsing Kres: {}".format(args.kres_folder)) logger.info("Parsing Kres: {}".format(args.kres_folder))
kres_parser = Parser( kres_parser = Parser(
logger=logger,
corpus="kres", corpus="kres",
kres_srl_folder=args.kres_srl_folder kres_srl_folder=args.kres_srl_folder
) )