Mercurial > public > sg101

annotate core/management/commands/ssl_images.py @ 1005:9fc12bbc8c81

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Tweaks for news v2.0 with old news pending stories present.
author Brian Neal <bgneal@gmail.com>
date Thu, 26 Nov 2015 23:16:07 -0600
parents 65b2bc9cb3cc
children fc528d4509b0
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@987 23 import lxml.html
bgneal@863 24 import markdown.inlinepatterns
bgneal@881 25 from PIL import Image
bgneal@979 26 import requests
bgneal@859 27
bgneal@860 28 from comments.models import Comment
bgneal@860 29 from forums.models import Post
bgneal@979 30 from core.download import download_file
bgneal@979 31 from core.functions import remove_file
bgneal@881 32 from core.s3 import S3Bucket
bgneal@987 33 from news.models import Story
bgneal@860 34
bgneal@860 35
bgneal@859 36 LOGFILE = os.path.join(settings.PROJECT_PATH, 'logs', 'ssl_images.log')
bgneal@859 37 logger = logging.getLogger(__name__)
bgneal@859 38
bgneal@871 39 IMAGE_LINK_RE = re.compile(markdown.inlinepatterns.IMAGE_LINK_RE,
bgneal@871 40 re.DOTALL | re.UNICODE)
bgneal@871 41 IMAGE_REF_RE = re.compile(markdown.inlinepatterns.IMAGE_REFERENCE_RE,
bgneal@871 42 re.DOTALL | re.UNICODE)
bgneal@863 43
bgneal@868 44 SG101_HOSTS = set(['www.surfguitar101.com', 'surfguitar101.com'])
bgneal@963 45 WHITELIST_HOSTS = set(settings.USER_IMAGES_SOURCES)
bgneal@987 46 MODEL_CHOICES = ['comments', 'posts', 'news']
bgneal@866 47
bgneal@881 48 PHOTO_MAX_SIZE = (660, 720)
bgneal@979 49 PHOTO_BASE_URL = settings.HOT_LINK_PHOTOS_BASE_URL
bgneal@979 50 PHOTO_BUCKET_NAME = settings.HOT_LINK_PHOTOS_BUCKET
bgneal@881 51
bgneal@899 52 CACHE_FILENAME = 'ssl_images_cache.json'
bgneal@899 53
bgneal@863 54 quit_flag = False
bgneal@881 55 bucket = None
bgneal@881 56 url_cache = {}
bgneal@899 57 bad_hosts = set()
bgneal@980 58 request_timeout = None
bgneal@863 59
bgneal@863 60
bgneal@863 61 def signal_handler(signum, frame):
bgneal@863 62 """SIGINT signal handler"""
bgneal@863 63 global quit_flag
bgneal@863 64 quit_flag = True
bgneal@863 65
bgneal@859 66
bgneal@859 67 def _setup_logging():
bgneal@859 68 logger.setLevel(logging.DEBUG)
bgneal@859 69 logger.propagate = False
bgneal@859 70 handler = logging.FileHandler(filename=LOGFILE, encoding='utf-8')
bgneal@859 71 formatter = logging.Formatter('%(asctime)s %(levelname)s %(message)s')
bgneal@859 72 handler.setFormatter(formatter)
bgneal@859 73 logger.addHandler(handler)
bgneal@859 74
bgneal@981 75 requests_log = logging.getLogger("requests.packages.urllib3")
bgneal@981 76 requests_log.setLevel(logging.INFO)
bgneal@981 77 requests_log.propagate = True
bgneal@981 78 requests_log.addHandler(handler)
bgneal@981 79
bgneal@982 80 dl_log = logging.getLogger("core.download")
bgneal@982 81 dl_log.setLevel(logging.INFO)
bgneal@982 82 dl_log.propagate = True
bgneal@982 83 dl_log.addHandler(handler)
bgneal@982 84
bgneal@859 85
bgneal@979 86 def resize_image(img_path):
bgneal@979 87 """Resizes the image found at img_path if necessary.
bgneal@979 88
bgneal@979 89 Returns True if the image was resized or resizing wasn't necessary.
bgneal@979 90 Returns False if the image could not be read or processed.
bgneal@881 91 """
bgneal@979 92 try:
bgneal@979 93 image = Image.open(img_path)
bgneal@979 94 except IOError as ex:
bgneal@979 95 logger.error("Error opening %s: %s", img_path, ex)
bgneal@979 96 return False
bgneal@881 97
bgneal@881 98 if image.size > PHOTO_MAX_SIZE:
bgneal@881 99 logger.info('Resizing from %s to %s', image.size, PHOTO_MAX_SIZE)
bgneal@981 100 try:
bgneal@981 101 image.thumbnail(PHOTO_MAX_SIZE, Image.ANTIALIAS)
bgneal@981 102 image.save(img_path)
bgneal@981 103 except IOError as ex:
bgneal@981 104 logger.error("Error resizing image from %s: %s", img_path, ex)
bgneal@981 105 return False
bgneal@881 106
bgneal@979 107 return True
bgneal@979 108
bgneal@881 109
bgneal@882 110 def gen_key():
bgneal@882 111 """Return a random key."""
bgneal@882 112 return base64.b64encode(uuid.uuid4().bytes, '-_').rstrip('=')
bgneal@882 113
bgneal@882 114
bgneal@881 115 def upload_image(img_path):
bgneal@881 116 """Upload image file located at img_path to our S3 bucket.
bgneal@881 117
bgneal@881 118 Returns the URL of the image in the bucket or None if an error occurs.
bgneal@881 119 """
bgneal@881 120 logger.info("upload_image starting")
bgneal@881 121 # Make a unique name for the image in the bucket
bgneal@881 122 ext = os.path.splitext(img_path)[1]
bgneal@882 123 file_key = gen_key() + ext
bgneal@881 124 try:
bgneal@881 125 return bucket.upload_from_filename(file_key, img_path, public=True)
bgneal@881 126 except IOError as ex:
bgneal@881 127 logger.error("Error uploading file: %s", ex)
bgneal@881 128 return None
bgneal@881 129
bgneal@881 130
bgneal@888 131 def convert_to_ssl(parsed_url):
bgneal@888 132 """Top-level function for moving an image to SSL."""
bgneal@888 133
bgneal@888 134 src = parsed_url.geturl()
bgneal@888 135
bgneal@899 136 if parsed_url.hostname in bad_hosts:
bgneal@899 137 logger.info("Host known to be bad, skipping: %s", src)
bgneal@899 138 return None
bgneal@899 139
bgneal@899 140 # Check the cache
bgneal@897 141 try:
bgneal@897 142 new_url = url_cache[src]
bgneal@897 143 except KeyError:
bgneal@897 144 # cache miss, try to get the file
bgneal@899 145 new_url = save_image_to_cloud(parsed_url)
bgneal@897 146 url_cache[src] = new_url
bgneal@897 147 else:
bgneal@897 148 if new_url:
bgneal@897 149 logger.info("Found URL in cache: %s => %s", src, new_url)
bgneal@897 150 else:
bgneal@897 151 logger.info("URL known to be bad, skipping: %s", src)
bgneal@888 152
bgneal@889 153 return new_url
bgneal@888 154
bgneal@888 155
bgneal@899 156 def save_image_to_cloud(parsed_url):
bgneal@881 157 """Downloads an image at a given source URL. Uploads it to cloud storage.
bgneal@881 158
bgneal@881 159 Returns the new URL or None if unsuccessful.
bgneal@881 160 """
bgneal@979 161 url = parsed_url.geturl()
bgneal@979 162 fn = None
bgneal@979 163 try:
bgneal@980 164 fn = download_file(url, timeout=request_timeout)
bgneal@979 165 except requests.ConnectionError as ex:
bgneal@979 166 logger.error("ConnectionError, ignoring host %s", parsed_url.hostname)
bgneal@979 167 bad_hosts.add(parsed_url.hostname)
bgneal@979 168 except requests.RequestException as ex:
bgneal@979 169 logger.error("%s", ex)
bgneal@979 170 except Exception as ex:
bgneal@979 171 logger.exception("%s", ex)
bgneal@979 172
bgneal@881 173 if fn:
bgneal@979 174 with remove_file(fn):
bgneal@979 175 if resize_image(fn):
bgneal@979 176 return upload_image(fn)
bgneal@881 177 return None
bgneal@868 178
bgneal@868 179
bgneal@866 180 def replace_image_markup(match):
bgneal@870 181 src_parts = match.group(8).split()
bgneal@868 182 if src_parts:
bgneal@868 183 src = src_parts[0]
bgneal@868 184 if src[0] == "<" and src[-1] == ">":
bgneal@868 185 src = src[1:-1]
bgneal@868 186 else:
bgneal@868 187 src = ''
bgneal@868 188
bgneal@868 189 title = ''
bgneal@868 190 if len(src_parts) > 1:
bgneal@868 191 title = " ".join(src_parts[1:])
bgneal@870 192 alt = match.group(1)
bgneal@868 193
bgneal@871 194 new_src = None
bgneal@868 195 if src:
bgneal@986 196 try:
bgneal@986 197 r = urlparse.urlparse(src)
bgneal@986 198 except ValueError:
bgneal@986 199 return u'{bad image}'
bgneal@986 200
bgneal@871 201 if r.hostname in SG101_HOSTS:
bgneal@871 202 new_src = r.path # convert to relative path
bgneal@871 203 elif r.scheme == 'http':
bgneal@888 204 # Try a few things to get this on ssl:
bgneal@888 205 new_src = convert_to_ssl(r)
bgneal@868 206 elif r.scheme == 'https':
bgneal@963 207 if r.hostname in WHITELIST_HOSTS:
bgneal@963 208 new_src = src # already in whitelist
bgneal@963 209 else:
bgneal@963 210 new_src = convert_to_ssl(r)
bgneal@868 211
bgneal@868 212 if new_src:
bgneal@868 213 if title:
bgneal@871 214 s = u'![{alt}]({src} {title})'.format(alt=alt, src=new_src, title=title)
bgneal@868 215 else:
bgneal@868 216 s = u'![{alt}]({src})'.format(alt=alt, src=new_src)
bgneal@868 217 else:
bgneal@868 218 # something's messed up, convert to a link using original src
bgneal@868 219 s = u'[{alt}]({src})'.format(alt=alt, src=src)
bgneal@868 220
bgneal@868 221 return s
bgneal@860 222
bgneal@860 223
bgneal@887 224 def warn_if_image_refs(text, model_name, pk):
bgneal@887 225 """Search text for Markdown image reference markup.
bgneal@887 226
bgneal@887 227 We aren't expecting these, but we will log something if we see any.
bgneal@887 228 """
bgneal@887 229 if IMAGE_REF_RE.search(text):
bgneal@887 230 logger.warning("Image reference found in %s pk = #%d", model_name, pk)
bgneal@887 231
bgneal@887 232
bgneal@866 233 def process_post(text):
bgneal@863 234 """Process the post object:
bgneal@863 235
bgneal@863 236 A regex substitution is run on the post's text field. This fixes up image
bgneal@863 237 links, getting rid of plain old http sources; either converting to https
bgneal@863 238 or relative style links (if the link is to SG101).
bgneal@863 239
bgneal@863 240 """
bgneal@866 241 return IMAGE_LINK_RE.sub(replace_image_markup, text)
bgneal@863 242
bgneal@863 243
bgneal@987 244 def process_html(html):
bgneal@987 245 """Process the html fragment, converting to https where needed."""
bgneal@987 246 s = html.strip()
bgneal@987 247 if not s:
bgneal@987 248 return s
bgneal@987 249
bgneal@987 250 changed = False
bgneal@987 251 root = lxml.html.fragment_fromstring(s, create_parent=True)
bgneal@987 252 for img in root.iter('img'):
bgneal@987 253 src = img.get('src')
bgneal@987 254 src = src.strip() if src else ''
bgneal@987 255 if src:
bgneal@987 256 try:
bgneal@987 257 r = urlparse.urlparse(src)
bgneal@987 258 except ValueError:
bgneal@987 259 logger.warning("Bad url? Should not happen; skipping...")
bgneal@987 260 continue
bgneal@987 261
bgneal@987 262 new_src = None
bgneal@987 263 if r.hostname in SG101_HOSTS:
bgneal@987 264 new_src = r.path # convert to relative path
bgneal@987 265 elif ((r.scheme == 'http') or
bgneal@987 266 (r.scheme == 'https' and r.hostname not in WHITELIST_HOSTS)):
bgneal@987 267 new_src = convert_to_ssl(r)
bgneal@987 268 if not new_src:
bgneal@987 269 # failed to convert to https; convert to a link
bgneal@987 270 tail = img.tail
bgneal@987 271 img.clear()
bgneal@987 272 img.tag = 'a'
bgneal@987 273 img.set('href', src)
bgneal@987 274 img.text = 'Image'
bgneal@987 275 img.tail = tail
bgneal@987 276 changed = True
bgneal@987 277
bgneal@987 278 if new_src:
bgneal@987 279 img.set('src', new_src)
bgneal@987 280 changed = True
bgneal@987 281
bgneal@987 282 if changed:
bgneal@987 283 result = lxml.html.tostring(root, encoding='utf-8')
bgneal@988 284 result = result[5:-6] # strip off parent div we added
bgneal@988 285 return result.decode('utf-8')
bgneal@987 286 return html
bgneal@987 287
bgneal@987 288
bgneal@894 289 def html_check(html):
bgneal@894 290 """Return True if the given HTML fragment has <img> tags with src attributes
bgneal@894 291 that use http, and False otherwise.
bgneal@894 292 """
bgneal@894 293 if not html:
bgneal@894 294 return False
bgneal@894 295
bgneal@894 296 root = etree.HTML(html)
bgneal@894 297 for img in root.iter('img'):
bgneal@894 298 src = img.get('src')
bgneal@894 299 if src and src.lower().startswith('http:'):
bgneal@894 300 return True
bgneal@894 301 return False
bgneal@894 302
bgneal@894 303
bgneal@859 304 class Command(NoArgsCommand):
bgneal@859 305 help = "Rewrite forum posts and comments to not use http for images"
bgneal@859 306 option_list = NoArgsCommand.option_list + (
bgneal@866 307 make_option('-m', '--model',
bgneal@866 308 choices=MODEL_CHOICES,
bgneal@866 309 help="which model to update; must be one of {{{}}}".format(
bgneal@866 310 ', '.join(MODEL_CHOICES))),
bgneal@860 311 make_option('-i', '--i',
bgneal@859 312 type='int',
bgneal@863 313 help="optional first slice index; the i in [i:j]"),
bgneal@860 314 make_option('-j', '--j',
bgneal@859 315 type='int',
bgneal@863 316 help="optional second slice index; the j in [i:j]"),
bgneal@898 317 make_option('-t', '--timeout',
bgneal@980 318 type='float',
bgneal@979 319 help="optional socket timeout (secs)",
bgneal@980 320 default=30.0),
bgneal@859 321 )
bgneal@859 322
bgneal@859 323 def handle_noargs(self, **options):
bgneal@895 324 time_started = datetime.datetime.now()
bgneal@859 325 _setup_logging()
bgneal@860 326 logger.info("Starting; arguments received: %s", options)
bgneal@859 327
bgneal@866 328 if options['model'] not in MODEL_CHOICES:
bgneal@866 329 raise CommandError('Please choose a --model option')
bgneal@859 330
bgneal@866 331 if options['model'] == 'comments':
bgneal@860 332 qs = Comment.objects.all()
bgneal@987 333 text_attrs = ['comment']
bgneal@881 334 model_name = 'Comment'
bgneal@987 335 elif options['model'] == 'posts':
bgneal@987 336 qs = Post.objects.all()
bgneal@987 337 text_attrs = ['body']
bgneal@987 338 model_name = 'Post'
bgneal@860 339 else:
bgneal@987 340 qs = Story.objects.all()
bgneal@987 341 text_attrs = ['short_text', 'long_text']
bgneal@987 342 model_name = 'Story'
bgneal@987 343
bgneal@987 344 html_based = options['model'] == 'news'
bgneal@860 345
bgneal@860 346 i, j = options['i'], options['j']
bgneal@860 347
bgneal@860 348 if i is not None and i < 0:
bgneal@860 349 raise CommandError("-i must be >= 0")
bgneal@860 350 if j is not None and j < 0:
bgneal@860 351 raise CommandError("-j must be >= 0")
bgneal@860 352 if j is not None and i is not None and j <= i:
bgneal@860 353 raise CommandError("-j must be > -i")
bgneal@860 354
bgneal@860 355 if i is not None and j is not None:
bgneal@860 356 qs = qs[i:j]
bgneal@860 357 elif i is not None and j is None:
bgneal@860 358 qs = qs[i:]
bgneal@860 359 elif i is None and j is not None:
bgneal@860 360 qs = qs[:j]
bgneal@860 361
bgneal@881 362 # Set global socket timeout
bgneal@980 363 global request_timeout
bgneal@980 364 request_timeout = options.get('timeout')
bgneal@980 365 logger.info("Using socket timeout of %4.2f", request_timeout)
bgneal@881 366
bgneal@863 367 # Install signal handler for ctrl-c
bgneal@863 368 signal.signal(signal.SIGINT, signal_handler)
bgneal@863 369
bgneal@881 370 # Create bucket to upload photos
bgneal@881 371 global bucket
bgneal@881 372 bucket = S3Bucket(access_key=settings.USER_PHOTOS_ACCESS_KEY,
bgneal@881 373 secret_key=settings.USER_PHOTOS_SECRET_KEY,
bgneal@881 374 base_url=PHOTO_BASE_URL,
bgneal@881 375 bucket_name=PHOTO_BUCKET_NAME)
bgneal@887 376
bgneal@899 377 # Load cached info from previous runs
bgneal@899 378 load_cache()
bgneal@899 379
bgneal@887 380 if i is None:
bgneal@887 381 i = 0
bgneal@887 382
bgneal@895 383 count = 0
bgneal@881 384 for n, model in enumerate(qs.iterator()):
bgneal@863 385 if quit_flag:
bgneal@863 386 logger.warning("SIGINT received, exiting")
bgneal@881 387 break
bgneal@881 388 logger.info("Processing %s #%d (pk = %d)", model_name, n + i, model.pk)
bgneal@987 389 save_flag = False
bgneal@987 390 for text_attr in text_attrs:
bgneal@987 391 txt = getattr(model, text_attr)
bgneal@987 392
bgneal@987 393 if html_based:
bgneal@987 394 new_txt = process_html(txt)
bgneal@987 395 else:
bgneal@987 396 new_txt = process_post(txt)
bgneal@987 397 warn_if_image_refs(txt, model_name, model.pk)
bgneal@987 398
bgneal@987 399 if txt != new_txt:
bgneal@987 400 logger.info("Content changed on %s #%d (pk = %d)",
bgneal@987 401 model_name, n + i, model.pk)
bgneal@987 402 logger.debug(u"original: %s", txt)
bgneal@987 403 logger.debug(u"changed: %s", new_txt)
bgneal@987 404 setattr(model, text_attr, new_txt)
bgneal@987 405 save_flag = True
bgneal@987 406 elif not html_based and html_check(model.html):
bgneal@987 407 # Check for content generated with older smiley code that used
bgneal@987 408 # absolute URLs for the smiley images. If True, then just save
bgneal@987 409 # the model again to force updated HTML to be created.
bgneal@987 410 logger.info("Older Smiley HTML detected, forcing a save")
bgneal@987 411 save_flag = True
bgneal@987 412
bgneal@987 413 if save_flag:
bgneal@894 414 model.save()
bgneal@895 415 count += 1
bgneal@860 416
bgneal@895 417 time_finished = datetime.datetime.now()
bgneal@895 418 elapsed = time_finished - time_started
bgneal@895 419 logger.info("ssl_images exiting; number of objects: %d; elapsed: %s",
bgneal@895 420 count, elapsed)
bgneal@897 421
bgneal@897 422 http_images = len(url_cache)
bgneal@897 423 https_images = sum(1 for v in url_cache.itervalues() if v)
bgneal@897 424 bad_images = http_images - https_images
bgneal@897 425 if http_images > 0:
bgneal@897 426 pct_saved = float(https_images) / http_images * 100.0
bgneal@897 427 else:
bgneal@897 428 pct_saved = 0.0
bgneal@897 429
bgneal@897 430 logger.info("Summary: http: %d; https: %d; lost: %d; saved: %3.1f %%",
bgneal@897 431 http_images, https_images, bad_images, pct_saved)
bgneal@899 432
bgneal@899 433 save_cache()
bgneal@899 434 logger.info("ssl_images done")
bgneal@899 435
bgneal@899 436
bgneal@899 437 def load_cache():
bgneal@899 438 """Load cache from previous runs."""
bgneal@899 439 logger.info("Loading cached information")
bgneal@899 440 try:
bgneal@899 441 with open(CACHE_FILENAME, 'r') as fp:
bgneal@899 442 d = json.load(fp)
bgneal@899 443 except IOError as ex:
bgneal@899 444 logger.error("Cache file (%s) IOError: %s", CACHE_FILENAME, ex)
bgneal@899 445 return
bgneal@899 446 except ValueError:
bgneal@899 447 logger.error("Mangled cache file: %s", CACHE_FILENAME)
bgneal@899 448 return
bgneal@899 449
bgneal@899 450 global bad_hosts, url_cache
bgneal@899 451 try:
bgneal@899 452 bad_hosts = set(d['bad_hosts'])
bgneal@899 453 url_cache = d['url_cache']
bgneal@899 454 except KeyError:
bgneal@899 455 logger.error("Malformed cache file: %s", CACHE_FILENAME)
bgneal@899 456
bgneal@899 457
bgneal@899 458 def save_cache():
bgneal@899 459 """Save our cache to a file for future runs."""
bgneal@899 460 logger.info("Saving cached information")
bgneal@899 461 d = {'bad_hosts': list(bad_hosts), 'url_cache': url_cache}
bgneal@899 462 with open(CACHE_FILENAME, 'w') as fp:
bgneal@899 463 json.dump(d, fp, indent=4)