diff options
author | Bryan Newbold <bnewbold@archive.org> | 2019-09-25 17:51:07 -0700 |
---|---|---|
committer | Bryan Newbold <bnewbold@archive.org> | 2019-09-25 17:51:07 -0700 |
commit | d7830b4a5aad0a59a588e98798711f0e694d50d6 (patch) | |
tree | 7565cbec74584a146b8ee79bb881fa9f78851f60 /python/tests/test_backfill_hbase_from_cdx.py | |
parent | 6e24eec4b6d1861eba37a0a05220b257e829ebbb (diff) | |
download | sandcrawler-d7830b4a5aad0a59a588e98798711f0e694d50d6.tar.gz sandcrawler-d7830b4a5aad0a59a588e98798711f0e694d50d6.zip |
refactor old python hadoop code into new directory
Diffstat (limited to 'python/tests/test_backfill_hbase_from_cdx.py')
-rw-r--r-- | python/tests/test_backfill_hbase_from_cdx.py | 74 |
1 files changed, 0 insertions, 74 deletions
diff --git a/python/tests/test_backfill_hbase_from_cdx.py b/python/tests/test_backfill_hbase_from_cdx.py deleted file mode 100644 index 070662b..0000000 --- a/python/tests/test_backfill_hbase_from_cdx.py +++ /dev/null @@ -1,74 +0,0 @@ -""" -TODO: could probably refactor to use unittest.mock.patch('happybase') -""" - -import io -import json -import pytest -import mrjob -import happybase_mock -from backfill_hbase_from_cdx import MRCDXBackfillHBase - -@pytest.fixture -def job(): - """ - Note: this mock only seems to work with job.run_mapper(), not job.run(); - the later results in a separate instantiation without the mock? - """ - job = MRCDXBackfillHBase(['--no-conf', '-']) - - conn = happybase_mock.Connection() - conn.create_table('wbgrp-journal-extract-test', - {'file': {}, 'grobid0': {}, 'f': {}}) - job.hb_table = conn.table('wbgrp-journal-extract-test') - - return job - - -def test_some_lines(job): - - raw = io.BytesIO(b""" -com,sagepub,cep)/content/28/9/960.full.pdf 20170705062200 http://cep.sagepub.com/content/28/9/960.full.pdf application/pdf 301 3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ - - 401 313356621 CITESEERX-CRAWL-2017-06-20-20170705061647307-00039-00048-wbgrp-svc284/CITESEERX-CRAWL-2017-06-20-20170705062052659-00043-31209~wbgrp-svc284.us.archive.org~8443.warc.gz -eu,eui,cadmus)/bitstream/handle/1814/36635/rscas_2015_03.pdf;jsessionid=761393014319a39f40d32ae3eb3a853f?sequence=1 20170705062202 http://cadmus.eui.eu/bitstream/handle/1814/36635/RSCAS_2015_03.pdf%3Bjsessionid%3D761393014319A39F40D32AE3EB3A853F?sequence%3D1 application/PDF 200 MPCXVWMUTRUGFP36SLPHKDLY6NGU4S3J - - 854156 328850624 CITESEERX-CRAWL-2017-06-20-20170705061647307-00039-00048-wbgrp-svc284/CITESEERX-CRAWL-2017-06-20-20170705062052659-00043-31209~wbgrp-svc284.us.archive.org~8443.warc.gz -com,pbworks,educ333b)/robots.txt 20170705063311 http://educ333b.pbworks.com/robots.txt text/plain 200 6VAUYENMOU2SK2OWNRPDD6WTQTECGZAD - - 638 398190140 CITESEERX-CRAWL-2017-06-20-20170705062707827-00049-00058-wbgrp-svc284/CITESEERX-CRAWL-2017-06-20-20170705063158203-00053-31209~wbgrp-svc284.us.archive.org~8443.warc.gz -""") - - job.sandbox(stdin=raw) - job.run_mapper() - - assert job.hb_table.row(b'1') == {} - # HTTP 301 - assert job.hb_table.row(b'sha1:3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ') == {} - # valid - assert job.hb_table.row(b'sha1:MPCXVWMUTRUGFP36SLPHKDLY6NGU4S3J') != {} - # text/plain - assert job.hb_table.row(b'sha1:6VAUYENMOU2SK2OWNRPDD6WTQTECGZAD') == {} - - row = job.hb_table.row(b'sha1:MPCXVWMUTRUGFP36SLPHKDLY6NGU4S3J') - assert row[b'file:mime'] == b"application/pdf" - - file_cdx = json.loads(row[b'file:cdx'].decode('utf-8')) - assert int(file_cdx['offset']) == 328850624 - - f_c = json.loads(row[b'f:c'].decode('utf-8')) - assert f_c['u'] == "http://cadmus.eui.eu/bitstream/handle/1814/36635/RSCAS_2015_03.pdf%3Bjsessionid%3D761393014319A39F40D32AE3EB3A853F?sequence%3D1" - assert b'i' not in f_c - -def test_parse_cdx_skip(job): - - job.mapper_init() - - print("CDX prefix") - raw = " com,sagepub,cep)/content/28/9/960.full.pdf 20170705062200 http://cep.sagepub.com/content/28/9/960.full.pdf application/pdf 200 3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ - - 401 313356621 CITESEERX-CRAWL-2017-06-20-20170705062052659-00043-31209~wbgrp-svc284.us.archive.org~8443.warc.gz" - info, status = job.mapper(None, raw).__next__() - assert info is None - assert status['status'] == "invalid" - assert 'prefix' in status['reason'] - - print("mimetype") - raw = "com,sagepub,cep)/content/28/9/960.full.pdf 20170705062200 http://cep.sagepub.com/content/28/9/960.full.pdf text/html 200 3I42H3S6NNFQ2MSVX7XZKYAYSCX5QBYJ - - 401 313356621 CITESEERX-CRAWL-2017-06-20-20170705061647307-00039-00048-wbgrp-svc284/CITESEERX-CRAWL-2017-06-20-20170705062052659-00043-31209~wbgrp-svc284.us.archive.org~8443.warc.gz" - info, status = job.mapper(None, raw).__next__() - assert info is None - assert status['status'] == "skip" - assert 'mimetype' in status['reason'] - |