added parallel json output creation

Makefile

@@ -1,15 +1,15 @@
 .PHONY: tsv_files srl_tagged_files json_files env clean
 
-all: json_files
+all: tools/fillpred_model/model.pickle tsv_files srl_tagged_files json_files
 
-json_files: srl_tagged_files
+json_files: # srl_tagged_files
 	cd tools; python3 gen_json.py
 
-srl_tagged_files: tsv_files
+srl_tagged_files: # tsv_files
 	# cd tools/srl-20131216; ./scripts/parse_srl_only_mod.sh; cd -
 	cd tools/srl-20131216; ./tag_all.sh
 
-tsv_files: tools/fillpred_model/model.pickle
+tsv_files: # tools/fillpred_model/model.pickle
 	cd tools; python3 parse_all.py
 
 tools/fillpred_model/model.pickle:

dockerfiles/python-java/Dockerfile

@@ -5,8 +5,9 @@ RUN apt-get install -y \
 vim \
 default-jdk \
 python3 \
-python3-pip \
-sshfs
+python3-pip
+
+RUN apt-get install -y sshfs
 
 RUN pip3 install lxml pandas sklearn
 

tools/gen_json.py

@@ -4,6 +4,7 @@ import configparser
 import json
 import sys
 import logging
+from multiprocessing import Pool
 
 # parse config
 config = configparser.ConfigParser()
@@ -12,11 +13,14 @@ ORIGPATH = Path(config["tools"]["kres_orig"])
 INPATH = Path(config["tools"]["kres_srl"])
 OUTPATH = Path(config["tools"]["kres_json"])
 DEBUG = config["tools"]["debug"] == "True"
-LOGFILE = Path(config["tools"]["logfile"]).absolute().resolve()
+CPU_CORES = int(config["tools"]["cpu_cores"])
+
+LOGFILE = Path(config["tools"]["logfile"]).absolute()
+LOGFILE.touch(exist_ok=True)
+LOGFILE.resolve()
 
 logging.basicConfig(filename=str(LOGFILE), level=logging.INFO)
 
-
 def get_origfile(filename):
     for origfile in ORIGPATH.iterdir():
         if filename.name.split('.')[0] == origfile.name.split('.')[0]:
@@ -55,12 +59,10 @@ def get_dep_rel(token):
             }
     return None
 
-
-par = Parser()
-OUTPATH.mkdir(exist_ok=True)
-
-logging.info("Start generating .josn files.")
-for infile in [x for x in INPATH.iterdir() if x.is_file()]:
+def handle_file(infile_tpl):
+    i = infile_tpl[0]
+    infile = infile_tpl[1]
+    outfile = (OUTPATH / infile.name).with_suffix(".json")
     origfile = get_origfile(infile)
     orig_dict = par.parse_tei(origfile)
 
@@ -94,8 +96,19 @@ for infile in [x for x in INPATH.iterdir() if x.is_file()]:
                 print()
                 print()
 
-	outfile = (OUTPATH / infile.name).with_suffix(".json")
     with outfile.open("w") as fp:
         json.dump(outdata, fp)
-		logging.info("SRL relations written to: ", outfile)
+        logging.info("SRL relations written to: {}".format(outfile))
+
+
+# main
+par = Parser()
+OUTPATH.mkdir(exist_ok=True)
+
+infiles = list(enumerate([x for x in INPATH.iterdir() if x.is_file()]))
+logging.info("Generating JSON SRL files from {} tsv files.".format(len(infiles)))
+
+with Pool(CPU_CORES) as p:
+    p.map(handle_file, infiles)
+
 logging.info("Finished generating .json files.")

tools/tools.cfg

@@ -4,5 +4,5 @@ kres_tsv = ../data/kres_out/1_tsv
 kres_srl = ../data/kres_out/2_srl
 kres_json = ../data/kres_out/final_json
 logfile = ../progress.log
-cpu_cores = 5
+cpu_cores = 1
 debug = False