Mercurial > public > sg101

annotate core/management/commands/ssl_images.py @ 986:26de15fb5a80

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Guard against bad image urls in ssl_images.
author Brian Neal <bgneal@gmail.com>
date Sun, 25 Oct 2015 14:47:29 -0500
parents 7db9037915c4
children 76525f5ac2b1
rev   line source
bgneal@859 1 """
bgneal@859 2 ssl_images is a custom manage.py command to convert forum post and comment
bgneal@859 3 images to https. It does this by rewriting the markup:
bgneal@859 4 - Images with src = http://surfguitar101.com/something are rewritten to be
bgneal@859 5 /something.
bgneal@859 6 - Non SG101 images that use http: are downloaded, resized, and uploaded to
bgneal@859 7 an S3 bucket. The src attribute is replaced with the new S3 URL.
bgneal@859 8 """
bgneal@882 9 import base64
bgneal@895 10 import datetime
bgneal@899 11 import json
bgneal@859 12 import logging
bgneal@859 13 from optparse import make_option
bgneal@888 14 import os
bgneal@863 15 import re
bgneal@863 16 import signal
bgneal@868 17 import urlparse
bgneal@881 18 import uuid
bgneal@859 19
bgneal@859 20 from django.core.management.base import NoArgsCommand, CommandError
bgneal@859 21 from django.conf import settings
bgneal@894 22 from lxml import etree
bgneal@863 23 import markdown.inlinepatterns
bgneal@881 24 from PIL import Image
bgneal@979 25 import requests
bgneal@859 26
bgneal@860 27 from comments.models import Comment
bgneal@860 28 from forums.models import Post
bgneal@979 29 from core.download import download_file
bgneal@979 30 from core.functions import remove_file
bgneal@881 31 from core.s3 import S3Bucket
bgneal@860 32
bgneal@860 33
bgneal@859 34 LOGFILE = os.path.join(settings.PROJECT_PATH, 'logs', 'ssl_images.log')
bgneal@859 35 logger = logging.getLogger(__name__)
bgneal@859 36
bgneal@871 37 IMAGE_LINK_RE = re.compile(markdown.inlinepatterns.IMAGE_LINK_RE,
bgneal@871 38 re.DOTALL | re.UNICODE)
bgneal@871 39 IMAGE_REF_RE = re.compile(markdown.inlinepatterns.IMAGE_REFERENCE_RE,
bgneal@871 40 re.DOTALL | re.UNICODE)
bgneal@863 41
bgneal@868 42 SG101_HOSTS = set(['www.surfguitar101.com', 'surfguitar101.com'])
bgneal@963 43 WHITELIST_HOSTS = set(settings.USER_IMAGES_SOURCES)
bgneal@866 44 MODEL_CHOICES = ['comments', 'posts']
bgneal@866 45
bgneal@881 46 PHOTO_MAX_SIZE = (660, 720)
bgneal@979 47 PHOTO_BASE_URL = settings.HOT_LINK_PHOTOS_BASE_URL
bgneal@979 48 PHOTO_BUCKET_NAME = settings.HOT_LINK_PHOTOS_BUCKET
bgneal@881 49
bgneal@899 50 CACHE_FILENAME = 'ssl_images_cache.json'
bgneal@899 51
bgneal@863 52 quit_flag = False
bgneal@881 53 bucket = None
bgneal@881 54 url_cache = {}
bgneal@899 55 bad_hosts = set()
bgneal@980 56 request_timeout = None
bgneal@863 57
bgneal@863 58
bgneal@863 59 def signal_handler(signum, frame):
bgneal@863 60 """SIGINT signal handler"""
bgneal@863 61 global quit_flag
bgneal@863 62 quit_flag = True
bgneal@863 63
bgneal@859 64
bgneal@859 65 def _setup_logging():
bgneal@859 66 logger.setLevel(logging.DEBUG)
bgneal@859 67 logger.propagate = False
bgneal@859 68 handler = logging.FileHandler(filename=LOGFILE, encoding='utf-8')
bgneal@859 69 formatter = logging.Formatter('%(asctime)s %(levelname)s %(message)s')
bgneal@859 70 handler.setFormatter(formatter)
bgneal@859 71 logger.addHandler(handler)
bgneal@859 72
bgneal@981 73 requests_log = logging.getLogger("requests.packages.urllib3")
bgneal@981 74 requests_log.setLevel(logging.INFO)
bgneal@981 75 requests_log.propagate = True
bgneal@981 76 requests_log.addHandler(handler)
bgneal@981 77
bgneal@982 78 dl_log = logging.getLogger("core.download")
bgneal@982 79 dl_log.setLevel(logging.INFO)
bgneal@982 80 dl_log.propagate = True
bgneal@982 81 dl_log.addHandler(handler)
bgneal@982 82
bgneal@859 83
bgneal@979 84 def resize_image(img_path):
bgneal@979 85 """Resizes the image found at img_path if necessary.
bgneal@979 86
bgneal@979 87 Returns True if the image was resized or resizing wasn't necessary.
bgneal@979 88 Returns False if the image could not be read or processed.
bgneal@881 89 """
bgneal@979 90 try:
bgneal@979 91 image = Image.open(img_path)
bgneal@979 92 except IOError as ex:
bgneal@979 93 logger.error("Error opening %s: %s", img_path, ex)
bgneal@979 94 return False
bgneal@881 95
bgneal@881 96 if image.size > PHOTO_MAX_SIZE:
bgneal@881 97 logger.info('Resizing from %s to %s', image.size, PHOTO_MAX_SIZE)
bgneal@981 98 try:
bgneal@981 99 image.thumbnail(PHOTO_MAX_SIZE, Image.ANTIALIAS)
bgneal@981 100 image.save(img_path)
bgneal@981 101 except IOError as ex:
bgneal@981 102 logger.error("Error resizing image from %s: %s", img_path, ex)
bgneal@981 103 return False
bgneal@881 104
bgneal@979 105 return True
bgneal@979 106
bgneal@881 107
bgneal@882 108 def gen_key():
bgneal@882 109 """Return a random key."""
bgneal@882 110 return base64.b64encode(uuid.uuid4().bytes, '-_').rstrip('=')
bgneal@882 111
bgneal@882 112
bgneal@881 113 def upload_image(img_path):
bgneal@881 114 """Upload image file located at img_path to our S3 bucket.
bgneal@881 115
bgneal@881 116 Returns the URL of the image in the bucket or None if an error occurs.
bgneal@881 117 """
bgneal@881 118 logger.info("upload_image starting")
bgneal@881 119 # Make a unique name for the image in the bucket
bgneal@881 120 ext = os.path.splitext(img_path)[1]
bgneal@882 121 file_key = gen_key() + ext
bgneal@881 122 try:
bgneal@881 123 return bucket.upload_from_filename(file_key, img_path, public=True)
bgneal@881 124 except IOError as ex:
bgneal@881 125 logger.error("Error uploading file: %s", ex)
bgneal@881 126 return None
bgneal@881 127
bgneal@881 128
bgneal@888 129 def convert_to_ssl(parsed_url):
bgneal@888 130 """Top-level function for moving an image to SSL."""
bgneal@888 131
bgneal@888 132 src = parsed_url.geturl()
bgneal@888 133
bgneal@899 134 if parsed_url.hostname in bad_hosts:
bgneal@899 135 logger.info("Host known to be bad, skipping: %s", src)
bgneal@899 136 return None
bgneal@899 137
bgneal@899 138 # Check the cache
bgneal@897 139 try:
bgneal@897 140 new_url = url_cache[src]
bgneal@897 141 except KeyError:
bgneal@897 142 # cache miss, try to get the file
bgneal@899 143 new_url = save_image_to_cloud(parsed_url)
bgneal@897 144 url_cache[src] = new_url
bgneal@897 145 else:
bgneal@897 146 if new_url:
bgneal@897 147 logger.info("Found URL in cache: %s => %s", src, new_url)
bgneal@897 148 else:
bgneal@897 149 logger.info("URL known to be bad, skipping: %s", src)
bgneal@888 150
bgneal@889 151 return new_url
bgneal@888 152
bgneal@888 153
bgneal@899 154 def save_image_to_cloud(parsed_url):
bgneal@881 155 """Downloads an image at a given source URL. Uploads it to cloud storage.
bgneal@881 156
bgneal@881 157 Returns the new URL or None if unsuccessful.
bgneal@881 158 """
bgneal@979 159 url = parsed_url.geturl()
bgneal@979 160 fn = None
bgneal@979 161 try:
bgneal@980 162 fn = download_file(url, timeout=request_timeout)
bgneal@979 163 except requests.ConnectionError as ex:
bgneal@979 164 logger.error("ConnectionError, ignoring host %s", parsed_url.hostname)
bgneal@979 165 bad_hosts.add(parsed_url.hostname)
bgneal@979 166 except requests.RequestException as ex:
bgneal@979 167 logger.error("%s", ex)
bgneal@979 168 except Exception as ex:
bgneal@979 169 logger.exception("%s", ex)
bgneal@979 170
bgneal@881 171 if fn:
bgneal@979 172 with remove_file(fn):
bgneal@979 173 if resize_image(fn):
bgneal@979 174 return upload_image(fn)
bgneal@881 175 return None
bgneal@868 176
bgneal@868 177
bgneal@866 178 def replace_image_markup(match):
bgneal@870 179 src_parts = match.group(8).split()
bgneal@868 180 if src_parts:
bgneal@868 181 src = src_parts[0]
bgneal@868 182 if src[0] == "<" and src[-1] == ">":
bgneal@868 183 src = src[1:-1]
bgneal@868 184 else:
bgneal@868 185 src = ''
bgneal@868 186
bgneal@868 187 title = ''
bgneal@868 188 if len(src_parts) > 1:
bgneal@868 189 title = " ".join(src_parts[1:])
bgneal@870 190 alt = match.group(1)
bgneal@868 191
bgneal@871 192 new_src = None
bgneal@868 193 if src:
bgneal@986 194 try:
bgneal@986 195 r = urlparse.urlparse(src)
bgneal@986 196 except ValueError:
bgneal@986 197 return u'{bad image}'
bgneal@986 198
bgneal@871 199 if r.hostname in SG101_HOSTS:
bgneal@871 200 new_src = r.path # convert to relative path
bgneal@871 201 elif r.scheme == 'http':
bgneal@888 202 # Try a few things to get this on ssl:
bgneal@888 203 new_src = convert_to_ssl(r)
bgneal@868 204 elif r.scheme == 'https':
bgneal@963 205 if r.hostname in WHITELIST_HOSTS:
bgneal@963 206 new_src = src # already in whitelist
bgneal@963 207 else:
bgneal@963 208 new_src = convert_to_ssl(r)
bgneal@868 209
bgneal@868 210 if new_src:
bgneal@868 211 if title:
bgneal@871 212 s = u'![{alt}]({src} {title})'.format(alt=alt, src=new_src, title=title)
bgneal@868 213 else:
bgneal@868 214 s = u'![{alt}]({src})'.format(alt=alt, src=new_src)
bgneal@868 215 else:
bgneal@868 216 # something's messed up, convert to a link using original src
bgneal@868 217 s = u'[{alt}]({src})'.format(alt=alt, src=src)
bgneal@868 218
bgneal@868 219 return s
bgneal@860 220
bgneal@860 221
bgneal@887 222 def warn_if_image_refs(text, model_name, pk):
bgneal@887 223 """Search text for Markdown image reference markup.
bgneal@887 224
bgneal@887 225 We aren't expecting these, but we will log something if we see any.
bgneal@887 226 """
bgneal@887 227 if IMAGE_REF_RE.search(text):
bgneal@887 228 logger.warning("Image reference found in %s pk = #%d", model_name, pk)
bgneal@887 229
bgneal@887 230
bgneal@866 231 def process_post(text):
bgneal@863 232 """Process the post object:
bgneal@863 233
bgneal@863 234 A regex substitution is run on the post's text field. This fixes up image
bgneal@863 235 links, getting rid of plain old http sources; either converting to https
bgneal@863 236 or relative style links (if the link is to SG101).
bgneal@863 237
bgneal@863 238 """
bgneal@866 239 return IMAGE_LINK_RE.sub(replace_image_markup, text)
bgneal@863 240
bgneal@863 241
bgneal@894 242 def html_check(html):
bgneal@894 243 """Return True if the given HTML fragment has <img> tags with src attributes
bgneal@894 244 that use http, and False otherwise.
bgneal@894 245 """
bgneal@894 246 if not html:
bgneal@894 247 return False
bgneal@894 248
bgneal@894 249 root = etree.HTML(html)
bgneal@894 250 for img in root.iter('img'):
bgneal@894 251 src = img.get('src')
bgneal@894 252 if src and src.lower().startswith('http:'):
bgneal@894 253 return True
bgneal@894 254 return False
bgneal@894 255
bgneal@894 256
bgneal@859 257 class Command(NoArgsCommand):
bgneal@859 258 help = "Rewrite forum posts and comments to not use http for images"
bgneal@859 259 option_list = NoArgsCommand.option_list + (
bgneal@866 260 make_option('-m', '--model',
bgneal@866 261 choices=MODEL_CHOICES,
bgneal@866 262 help="which model to update; must be one of {{{}}}".format(
bgneal@866 263 ', '.join(MODEL_CHOICES))),
bgneal@860 264 make_option('-i', '--i',
bgneal@859 265 type='int',
bgneal@863 266 help="optional first slice index; the i in [i:j]"),
bgneal@860 267 make_option('-j', '--j',
bgneal@859 268 type='int',
bgneal@863 269 help="optional second slice index; the j in [i:j]"),
bgneal@898 270 make_option('-t', '--timeout',
bgneal@980 271 type='float',
bgneal@979 272 help="optional socket timeout (secs)",
bgneal@980 273 default=30.0),
bgneal@859 274 )
bgneal@859 275
bgneal@859 276 def handle_noargs(self, **options):
bgneal@895 277 time_started = datetime.datetime.now()
bgneal@859 278 _setup_logging()
bgneal@860 279 logger.info("Starting; arguments received: %s", options)
bgneal@859 280
bgneal@866 281 if options['model'] not in MODEL_CHOICES:
bgneal@866 282 raise CommandError('Please choose a --model option')
bgneal@859 283
bgneal@866 284 if options['model'] == 'comments':
bgneal@860 285 qs = Comment.objects.all()
bgneal@866 286 text_attr = 'comment'
bgneal@881 287 model_name = 'Comment'
bgneal@860 288 else:
bgneal@860 289 qs = Post.objects.all()
bgneal@866 290 text_attr = 'body'
bgneal@881 291 model_name = 'Post'
bgneal@860 292
bgneal@860 293 i, j = options['i'], options['j']
bgneal@860 294
bgneal@860 295 if i is not None and i < 0:
bgneal@860 296 raise CommandError("-i must be >= 0")
bgneal@860 297 if j is not None and j < 0:
bgneal@860 298 raise CommandError("-j must be >= 0")
bgneal@860 299 if j is not None and i is not None and j <= i:
bgneal@860 300 raise CommandError("-j must be > -i")
bgneal@860 301
bgneal@860 302 if i is not None and j is not None:
bgneal@860 303 qs = qs[i:j]
bgneal@860 304 elif i is not None and j is None:
bgneal@860 305 qs = qs[i:]
bgneal@860 306 elif i is None and j is not None:
bgneal@860 307 qs = qs[:j]
bgneal@860 308
bgneal@881 309 # Set global socket timeout
bgneal@980 310 global request_timeout
bgneal@980 311 request_timeout = options.get('timeout')
bgneal@980 312 logger.info("Using socket timeout of %4.2f", request_timeout)
bgneal@881 313
bgneal@863 314 # Install signal handler for ctrl-c
bgneal@863 315 signal.signal(signal.SIGINT, signal_handler)
bgneal@863 316
bgneal@881 317 # Create bucket to upload photos
bgneal@881 318 global bucket
bgneal@881 319 bucket = S3Bucket(access_key=settings.USER_PHOTOS_ACCESS_KEY,
bgneal@881 320 secret_key=settings.USER_PHOTOS_SECRET_KEY,
bgneal@881 321 base_url=PHOTO_BASE_URL,
bgneal@881 322 bucket_name=PHOTO_BUCKET_NAME)
bgneal@887 323
bgneal@899 324 # Load cached info from previous runs
bgneal@899 325 load_cache()
bgneal@899 326
bgneal@887 327 if i is None:
bgneal@887 328 i = 0
bgneal@887 329
bgneal@895 330 count = 0
bgneal@881 331 for n, model in enumerate(qs.iterator()):
bgneal@863 332 if quit_flag:
bgneal@863 333 logger.warning("SIGINT received, exiting")
bgneal@881 334 break
bgneal@881 335 logger.info("Processing %s #%d (pk = %d)", model_name, n + i, model.pk)
bgneal@866 336 txt = getattr(model, text_attr)
bgneal@887 337 warn_if_image_refs(txt, model_name, model.pk)
bgneal@866 338 new_txt = process_post(txt)
bgneal@881 339 if txt != new_txt:
bgneal@889 340 logger.info("Content changed on %s #%d (pk = %d)",
bgneal@887 341 model_name, n + i, model.pk)
bgneal@881 342 logger.debug("original: %s", txt)
bgneal@881 343 logger.debug("changed: %s", new_txt)
bgneal@887 344 setattr(model, text_attr, new_txt)
bgneal@887 345 model.save()
bgneal@894 346 elif html_check(model.html):
bgneal@894 347 # Check for content generated with older smiley code that used
bgneal@894 348 # absolute URLs for the smiley images. If True, then just save
bgneal@894 349 # the model again to force updated HTML to be created.
bgneal@894 350 logger.info("Older Smiley HTML detected, forcing a save")
bgneal@894 351 model.save()
bgneal@895 352 count += 1
bgneal@860 353
bgneal@895 354 time_finished = datetime.datetime.now()
bgneal@895 355 elapsed = time_finished - time_started
bgneal@895 356 logger.info("ssl_images exiting; number of objects: %d; elapsed: %s",
bgneal@895 357 count, elapsed)
bgneal@897 358
bgneal@897 359 http_images = len(url_cache)
bgneal@897 360 https_images = sum(1 for v in url_cache.itervalues() if v)
bgneal@897 361 bad_images = http_images - https_images
bgneal@897 362 if http_images > 0:
bgneal@897 363 pct_saved = float(https_images) / http_images * 100.0
bgneal@897 364 else:
bgneal@897 365 pct_saved = 0.0
bgneal@897 366
bgneal@897 367 logger.info("Summary: http: %d; https: %d; lost: %d; saved: %3.1f %%",
bgneal@897 368 http_images, https_images, bad_images, pct_saved)
bgneal@899 369
bgneal@899 370 save_cache()
bgneal@899 371 logger.info("ssl_images done")
bgneal@899 372
bgneal@899 373
bgneal@899 374 def load_cache():
bgneal@899 375 """Load cache from previous runs."""
bgneal@899 376 logger.info("Loading cached information")
bgneal@899 377 try:
bgneal@899 378 with open(CACHE_FILENAME, 'r') as fp:
bgneal@899 379 d = json.load(fp)
bgneal@899 380 except IOError as ex:
bgneal@899 381 logger.error("Cache file (%s) IOError: %s", CACHE_FILENAME, ex)
bgneal@899 382 return
bgneal@899 383 except ValueError:
bgneal@899 384 logger.error("Mangled cache file: %s", CACHE_FILENAME)
bgneal@899 385 return
bgneal@899 386
bgneal@899 387 global bad_hosts, url_cache
bgneal@899 388 try:
bgneal@899 389 bad_hosts = set(d['bad_hosts'])
bgneal@899 390 url_cache = d['url_cache']
bgneal@899 391 except KeyError:
bgneal@899 392 logger.error("Malformed cache file: %s", CACHE_FILENAME)
bgneal@899 393
bgneal@899 394
bgneal@899 395 def save_cache():
bgneal@899 396 """Save our cache to a file for future runs."""
bgneal@899 397 logger.info("Saving cached information")
bgneal@899 398 d = {'bad_hosts': list(bad_hosts), 'url_cache': url_cache}
bgneal@899 399 with open(CACHE_FILENAME, 'w') as fp:
bgneal@899 400 json.dump(d, fp, indent=4)