diff options
Diffstat (limited to 'please')
-rwxr-xr-x | please | 60 |
1 files changed, 60 insertions, 0 deletions
@@ -64,16 +64,47 @@ def run_extract(args): --grobid-uri {grobid_uri} \ -r hadoop \ -c mrjob.conf \ + --output-dir {output} \ + --no-output \ --archive venv-current.tar.gz#venv \ --jobconf mapred.line.input.format.linespermap=8000 \ --jobconf mapreduce.job.queuename=extraction \ --jobconf mapred.task.timeout=3600000 \ {input_cdx} """.format(hbase_host=HBASE_HOST, env=args.env, + output=output, input_cdx=args.input_cdx, grobid_uri=GROBID_URI) subprocess.call(cmd, shell=True) +def run_extract_ungrobided(args): + if args.rebuild: + rebuild_python() + print("Starting extractungrobided job...") + output = "{}/output-{}/{}-extract-ungrobided".format( + HDFS_DIR, + args.env, + datetime.strftime(datetime.now(), "%Y-%m-%d-%H%M.%S")) + cmd = """cd python; + pipenv run ./extraction_ungrobided.py \ + --hbase-host {hbase_host} \ + --hbase-table wbgrp-journal-extract-0-{env} \ + --grobid-uri {grobid_uri} \ + -r hadoop \ + -c mrjob.conf \ + --output-dir {output} \ + --no-output \ + --archive venv-current.tar.gz#venv \ + --jobconf mapred.line.input.format.linespermap=8000 \ + --jobconf mapreduce.job.queuename=extraction \ + --jobconf mapred.task.timeout=3600000 \ + {input_ungrobided} + """.format(hbase_host=HBASE_HOST, env=args.env, + input_ungrobided=args.input_ungrobided, + output=output, + grobid_uri=GROBID_URI) + subprocess.call(cmd, shell=True) + def run_rowcount(args): if args.rebuild: rebuild_scalding() @@ -257,6 +288,27 @@ def run_keysmissingcol(args): env=args.env) subprocess.call(cmd, shell=True) +def run_dumpungrobided(args): + if args.rebuild: + rebuild_scalding() + print("Starting dumpungrobided job...") + output = "{}/output-{}/{}-dumpungrobided".format( + HDFS_DIR, + args.env, + datetime.strftime(datetime.now(), "%Y-%m-%d-%H%M.%S")) + cmd = """hadoop jar \ + scalding/target/scala-2.11/sandcrawler-assembly-0.2.0-SNAPSHOT.jar \ + com.twitter.scalding.Tool sandcrawler.DumpUnGrobidedJob \ + --hdfs \ + --app.conf.path scalding/ia_cluster.conf \ + --hbase-table wbgrp-journal-extract-0-{env} \ + --zookeeper-hosts {zookeeper_hosts} \ + --output {output}""".format( + output=output, + zookeeper_hosts=ZOOKEEPER_HOSTS, + env=args.env) + subprocess.call(cmd, shell=True) + def main(): parser = argparse.ArgumentParser() @@ -281,6 +333,11 @@ def main(): sub_extract.add_argument('input_cdx', help="full HDFS path of CDX file to extract") + sub_extractungrobided = subparsers.add_parser('extract-ungrobided') + sub_extractungrobided.set_defaults(func=run_extract_ungrobided) + sub_extractungrobided.add_argument('input_ungrobided', + help="full HDFS path of 'ungrobided' file to extract") + sub_rowcount = subparsers.add_parser('row-count') sub_rowcount.set_defaults(func=run_rowcount) @@ -320,6 +377,9 @@ def main(): sub_keysmissingcol.add_argument('column', help="column to SCAN for missing keys") + sub_dumpungrobided = subparsers.add_parser('dump-ungrobided') + sub_dumpungrobided.set_defaults(func=run_dumpungrobided) + args = parser.parse_args() if not args.__dict__.get("func"): print("tell me what to do! (try --help)") |