dosage/tests/modules/check_comics.py
Tobias Gruetzmacher 0428fd52b3 Speed up comic checks by avoiding redundant tests
We don't need to test the "navigation" of each and every comic on the
same hoster, if those test give us no new information (this is true for
most "modern" hosters which don't allow individual designs/HTML per
comic)
2021-01-19 00:55:18 +01:00

157 lines
5.3 KiB
Python

# SPDX-License-Identifier: MIT
# Copyright (C) 2004-2008 Tristan Seligmann and Jonathan Jacobs
# Copyright (C) 2012-2014 Bastian Kleineidam
# Copyright (C) 2015-2021 Tobias Gruetzmacher
import json
import multiprocessing
import os
import re
import warnings
from urllib.parse import urlsplit
# Dictionary with per-host locks.
_locks = {}
# Allowed number of connections per host
MaxConnections = 2
# Maximum number of strips to get to test a comic
MaxStrips = 5
# Match (already-escaped) archive.org URL
ARCHIVE_ORG_MATCH = re.compile(r'(?<=web\\.archive\\.org/web)/\d+/')
# Matches some (maybe-escaped - because Python 2) printf-style format specifiers
PRINTF_MATCH = re.compile(r'\\?%[0-9]*[sd]')
# Classes where the modules are very similar, so that testing the history of
# each modules doesn't make much sense
standarized_modules = {
'ComicSherpa',
'ComicsKingdom',
'GoComics',
'MangaDex',
'WebToons',
}
# Already seen classes
seen_modules = set()
def get_lock(host):
"""Get bounded semphore for given host."""
if host not in _locks:
_locks[host] = multiprocessing.BoundedSemaphore(MaxConnections)
return _locks[host]
def test_comicmodule(tmpdir, scraperobj, worker_id):
'''Test a scraper. It must be able to traverse backward for at least 5
strips from the start, and find strip images on at least 4 pages.'''
# Limit number of connections to one host.
host = urlsplit(scraperobj.url).hostname
with get_lock(host):
maxstrips = MaxStrips
parts = scraperobj.name.split('/', maxsplit=1)
if len(parts) > 1 and parts[0] in standarized_modules:
if parts[0] in seen_modules:
maxstrips = 1
else:
seen_modules.add(parts[0])
_test_comic(str(tmpdir), scraperobj, maxstrips)
def _test_comic(outdir, scraperobj, maxstrips):
num_strips = 0
strip = None
files = []
PROXYMAP.apply(scraperobj.name)
for strip in scraperobj.getStrips(maxstrips):
files.append(_check_strip(outdir, strip,
scraperobj.multipleImagesPerStrip))
if num_strips > 0:
_check_stripurl(strip, scraperobj)
num_strips += 1
if scraperobj.prevSearch and not scraperobj.hitFirstStripUrl:
# subtract the number of skipped URLs with no image from the expected
# image number
num_strips_expected = maxstrips - len(scraperobj.skippedUrls)
msg = 'Traversed %d strips instead of %d.' % (num_strips,
num_strips_expected)
if strip:
msg += " Check the prevSearch pattern at %s" % strip.strip_url
assert num_strips == num_strips_expected, msg
if strip:
_check_scraperesult(files, num_strips_expected, strip, scraperobj)
def _check_strip(outdir, strip, multipleImagesPerStrip):
'''Check that a specific page yields images and the comic module correctly
declares if there are multiple images per page.'''
images = []
files = []
for image in strip.getImages():
images.append(image.url)
# write a fake image (to download less)
fakeimg = image._fnbase(outdir) + '.fake'
with open(fakeimg, 'w') as f:
f.write("fake image for testing")
fn, _ = image.save(outdir)
files.append(fn)
assert images, 'failed to find images at %s' % strip.strip_url
if not multipleImagesPerStrip:
assert len(images) == 1, 'found more than 1 image at {}: {}'.format(
strip.strip_url, images)
return files
def _check_scraperesult(saved_images, num_images_expected, strip, scraperobj):
'''Check that exactly or for multiple pages at least num_strips images are
saved. This checks saved files, ie. it detects duplicate filenames.'''
num_images = len(saved_images)
attrs = (num_images, saved_images, num_images_expected)
if scraperobj.multipleImagesPerStrip:
err = 'saved %d %s instead of at least %d images' % attrs
assert num_images >= num_images_expected, err
else:
err = 'saved %d %s instead of %d images' % attrs
assert num_images == num_images_expected, err
def _check_stripurl(strip, scraperobj):
if not scraperobj.stripUrl:
# no indexing support
return
# test that the stripUrl regex matches the retrieved strip URL
urlmatch = re.escape(scraperobj.stripUrl)
urlmatch = PRINTF_MATCH.sub('.+', urlmatch)
urlmatch = ARCHIVE_ORG_MATCH.sub(r'/\\d+/', urlmatch)
ro = re.compile(urlmatch)
mo = ro.match(strip.strip_url)
if not mo:
warnings.warn('strip URL {!r} does not match stripUrl pattern {}'.format(
strip.strip_url, urlmatch))
class ProxyConfig:
"""Loads proxy config from an environment variable and applies it for each test."""
def __init__(self):
self.config = {}
if 'PROXYMAP' in os.environ:
for regex, server in json.loads(os.environ['PROXYMAP']).items():
self.config[re.compile(regex)] = server
def apply(self, name):
useserver = ''
for regex, server in self.config.items():
if regex.match(name):
useserver = server
break
os.environ['http_proxy'] = useserver
os.environ['https_proxy'] = useserver
# External proxy config to fetch some modules via proxies
PROXYMAP = ProxyConfig()