You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
125 lines
4.0 KiB
125 lines
4.0 KiB
from pathlib import Path
|
|
from corpusparser import Parser
|
|
import argparse
|
|
import logging
|
|
import json
|
|
from pymongo import MongoClient
|
|
import sys
|
|
|
|
logging.basicConfig(level=logging.INFO)
|
|
logger = logging.getLogger(__name__)
|
|
|
|
## Main handles command line arguments and writing to files / DB.
|
|
|
|
def ssj_to_json_file(sentence_generator, outfolder):
|
|
# this funciton is based on the fact that files are parsed sequentially
|
|
outfolder = Path(outfolder)
|
|
outfolder.mkdir(parents=True, exist_ok=True)
|
|
outfile = outfolder / "ssj500k.json"
|
|
|
|
data_buffer = []
|
|
for s in sentence_generator:
|
|
sdata = s[1]
|
|
data_buffer += [sdata]
|
|
|
|
# outfile = Path(outfile)
|
|
with outfile.open("w") as fp:
|
|
logger.info("Writing to {}".format(outfile))
|
|
json.dump(data_buffer, fp)
|
|
|
|
def kres_to_json_files(sentence_generator, outfolder):
|
|
outfolder = Path(outfolder) / "kres_json"
|
|
outfolder.mkdir(parents=True, exist_ok=True)
|
|
|
|
def write_buffer_to_file(outfile, outfile_buffer):
|
|
logger.info("Writing file: {}".format(outfile))
|
|
with outfile.open("w") as fp:
|
|
json.dump(outfile_buffer, fp)
|
|
|
|
outfile_buffer = None
|
|
current_outfile = None
|
|
for s in sentence_generator:
|
|
infile = s[0]
|
|
outfile = outfolder / Path(infile.name.split(".")[0]).with_suffix(".json")
|
|
|
|
# parser sequentially parses files; when we're done with a file, write it out
|
|
if current_outfile is None:
|
|
current_outfile = outfile
|
|
outfile_buffer = []
|
|
elif outfile != current_outfile:
|
|
write_buffer_to_file(current_outfile, outfile_buffer)
|
|
current_outfile = outfile
|
|
outfile_buffer = []
|
|
|
|
# update buffer
|
|
sdata = s[1]
|
|
outfile_buffer += [sdata]
|
|
write_buffer_to_file(current_outfile, outfile_buffer)
|
|
|
|
def data_to_valdb(sentence_generator, dbaddr, username, password, collection_name):
|
|
logger.info("Connecting to: {}".format(dbaddr))
|
|
client = MongoClient(
|
|
"mongodb://{}".format(dbaddr),
|
|
username=username,
|
|
password=password,
|
|
authSource="valdb",
|
|
authMechanism='SCRAM-SHA-256'
|
|
)
|
|
valdb = client.valdb
|
|
logger.info("Writing data to {}.".format(collection_name))
|
|
col = valdb[collection_name]
|
|
for s in sentence_generator:
|
|
sdata = s[1]
|
|
col.insert_one(sdata)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
parser = argparse.ArgumentParser(description="Parsing corpora kres and ssj500k.")
|
|
parser.add_argument('--kres-folder', required=True)
|
|
parser.add_argument('--kres-srl-folder', required=True)
|
|
parser.add_argument('--ssj-file', required=True)
|
|
parser.add_argument('--output', required=False, default=None)
|
|
parser.add_argument('--outdir', required=False, default=None)
|
|
parser.add_argument('--dbaddr', required=False, default=None)
|
|
parser.add_argument('--dbuser', required=False, default=None)
|
|
parser.add_argument('--dbpass', required=False, default=None)
|
|
args = parser.parse_args()
|
|
|
|
|
|
# parse ssj
|
|
logger.info("Parsing ssj500k: {}".format(args.ssj_file))
|
|
ssj_parser = Parser(
|
|
corpus="ssj",
|
|
infiles=[args.ssj_file],
|
|
)
|
|
# ssj to json
|
|
if args.output == "file":
|
|
ssj_to_json_file(ssj_parser.sentence_generator(), args.outdir)
|
|
elif args.output == "db":
|
|
data_to_valdb(
|
|
ssj_parser.sentence_generator(),
|
|
args.dbaddr,
|
|
args.dbuser,
|
|
args.dbpass,
|
|
collection_name="ssj"
|
|
)
|
|
|
|
sys.exit()
|
|
# parse kres
|
|
logger.info("Parsing Kres: {}".format(args.ssj_file))
|
|
kres_parser = Parser(
|
|
corpus="kres",
|
|
infiles=[args.kres_folder, args.kres_srl_folder],
|
|
)
|
|
# kres to json
|
|
if args.output == "file":
|
|
kres_to_json_files(kres_parser.sentence_generator(), args.outdir)
|
|
elif args.output == "db":
|
|
data_to_valdb(
|
|
ssj_parser.sentence_generator(),
|
|
args.dbaddr,
|
|
args.dbuser,
|
|
args.dbpass,
|
|
collection_name="kres"
|
|
)
|