def test_submissions_between_order(self):
all_subs = list(submissions_between(self.r,
self.sr,
highest_timestamp=time.time(),
verbosity=self.verbosity))
for i in range(len(all_subs) - 1):
self.assertGreaterEqual(all_subs[i].created_utc,
all_subs[i + 1].created_utc)
sr_obj = self.r.get_subreddit(self.sr)
all_subs_sr_object = list(
submissions_between(self.r,
sr_obj,
verbosity=self.verbosity)
)
self.assertEqual(all_subs, all_subs_sr_object)
all_subs_reversed = list(submissions_between(self.r,
sr_obj,
newest_first=False,
verbosity=self.verbosity))
self.assertEqual(all_subs, list(reversed(all_subs_reversed)))
def test_submissions_between_raises_correctly(self):
with self.assertRaises(PRAWException):
list(
submissions_between(self.r,
self.sr,
extra_cloudsearch_fields={'self': 'yes'},
verbosity=self.verbosity))
def test_submissions_between_special_subreddit(self):
for submission in submissions_between(
self.r, "all", # so it doesn't take forever
highest_timestamp=1121000000,
verbosity=self.verbosity):
self.assertLessEqual(submission.created_utc, 1121000000)
self.assertGreaterEqual(submission.created_utc, 0)
def main(*args):
"""
Main loop of scraping iteration.
Arguments:
terms is a string of search terms
subreddit is a string representing subreddit to search in
Returns:
None (write to file instead)
"""
if len(args) == 1:
terms = args[0]
subreddit = 'all'
elif len(args) == 2:
terms = args[0]
subreddit = args[1]
else:
print ('Usage: python reddit_scrape.py "<search terms>" '
'[<subreddit>]')
return
output_file = "__".join(['_'.join(terms.split()),
subreddit]) + '.json'
# initialize Reddit session (see praw.ini for OAuth details)
r = praw.Reddit(user_agent='Full Scraper by '
'/u/<your_username_here> version 1.0',
site_name='my_app')
# gather refresh token from praw.ini
try:
r.refresh_access_information()
except OAuthAppRequired:
print "Check your OAuth login information (see ./praw.ini)."
# get list of search results with the given terms
search_results = list(submissions_between(
r,
subreddit,
extra_cloudsearch_fields={'selftext': terms}
)
)
# get comments for each submission above
# and pull the ones that match the search
search_results_with_comments = get_comments(search_results)
# get list of cleaned search results
cleaned_results = clean_results(search_results_with_comments,
terms)
# write results to file in directory ./output/
write_results(cleaned_results, terms, output_file)
def get_past_comments(lowest_timestamp = None, highest_timestamp = None, from_oldest = False):
"""gets comments that were made before the given time"""
if from_oldest:
highest_timestamp = find_oldest()
subs = submissions_between(r, "soccer", highest_timestamp = highest_timestamp,
lowest_timestamp = lowest_timestamp)
comments = (sub.comments for sub in subs)
for sub_comments in comments:
for comment in sub_comments:
if not isinstance(comment, MoreComments):
yield Comment(comment)
elif isinstance(comment, MoreComments):
comments = comment.comments
def get_subreddit_stats(sublist, subreddit_stats):
for sub in sublist:
if str(sub) in subreddit_stats:
continue
downvoted_submissions = 0
total_submissions = 0
for submission in submissions_between(r,
sub,
lowest_timestamp=lowest_timestamp,
highest_timestamp=highest_timestamp,
verbosity=0):
assert submission.created_utc <= highest_timestamp and submission.created_utc >= lowest_timestamp
if submission.score <= 0:
downvoted_submissions += 1
total_submissions += 1
subreddit_stats[sub] = {"downvoted_submissions": downvoted_submissions,
"total_submissions": total_submissions}
subreddit_stats.sync()
print sub, total_submissions, downvoted_submissions, downvoted_submissions / float(total_submissions) if total_submissions else 0
def get_flair_stats():
user = os.environ['REDDIT_USERNAME']
user_agent = 'Calculating ignored bugs by {}'.format(user)
r = praw.Reddit(user_agent)
flair_stats = defaultdict(lambda: defaultdict(lambda: 0))
for s in submissions_between(r, 'bugs', lowest_timestamp=1400000000):
created = datetime.utcfromtimestamp(s.created_utc)
month = (created.year, created.month)
# They started to add flairs since Janury, 2015
if month < (2015, 2):
break
# Current month has incomplete data
if month == (date.today().year, date.today().month):
continue
# Submissions without flairs seems to be mainly duplicate submissions removed by mods
# They are not viewable in the interface, so we aren't counting them
if not s.link_flair_text:
print "IGNORING POST WITHOUT A FLAIR", s.permalink, s.title
continue
flair_stats[month][s.link_flair_text] += 1
return flair_stats
def test_submissions_between_with_filters(self):
all_subs = list(submissions_between(self.r,
self.sr,
verbosity=self.verbosity))
t1 = 1420000000
t2 = 1441111111
t1_t2_subs = list(submissions_between(self.r,
self.sr,
lowest_timestamp=t1,
highest_timestamp=t2,
verbosity=self.verbosity))
def filter_subs(subs,
lowest_timestamp=0,
highest_timestamp=10**10,
criterion=None):
filtered = [s for s in subs
if s.created_utc <= highest_timestamp and
s.created_utc >= lowest_timestamp and
(criterion is None or criterion(s))]
# make sure we never accidentally craft a bad test case
self.assertGreater(len(filtered), 0)
return filtered
t1_t2_subs_canon = filter_subs(all_subs, t1, t2)
self.assertEqual(t1_t2_subs, t1_t2_subs_canon)
self_subs = list(
submissions_between(self.r,
self.sr,
extra_cloudsearch_fields={"self": "1"},
verbosity=self.verbosity)
)
self_subs_canon = filter_subs(all_subs,
criterion=lambda s: s.is_self)
self.assertEqual(self_subs, self_subs_canon)
def wa_criterion(s):
return not s.is_self and \
urlparse(s.url).netloc == "web.archive.org"
wa_cs_fields = {"self": "0",
"site": "web.archive.org"}
subs_wa = list(
submissions_between(self.r,
self.sr,
extra_cloudsearch_fields=wa_cs_fields,
verbosity=self.verbosity)
)
subs_wa_canon = filter_subs(all_subs, criterion=wa_criterion)
self.assertEqual(subs_wa, subs_wa_canon)
patu_cs_fields = {"self": "1",
"author": "PyAPITestUser2",
"title": "test"}
def patu_criterion(s):
return s.is_self and \
s.author.name == "PyAPITestUser2" and\
"test" in s.title.lower()
subs_patu = list(
submissions_between(self.r,
self.sr,
extra_cloudsearch_fields=patu_cs_fields,
verbosity=self.verbosity)
)
subs_patu_canon = filter_subs(all_subs, criterion=patu_criterion)
self.assertEqual(subs_patu, subs_patu_canon)
def test_submissions_between_raises_correctly(self):
with self.assertRaises(PRAWException):
list(submissions_between(self.r,
self.sr,
extra_cloudsearch_fields={'self': 'yes'},
verbosity=self.verbosity))
def scrape():
reddit_obj = praw.Reddit(user_agent="Fake Images Scraper")
psbattle = reddit_obj.get_subreddit(
"photoshopbattles") #subreddit of interest
#submissions = psbattle.get_top_from_year(limit=None)#, params={"after" : "t3_4cd8qr"})
submissions = submissions_between(
reddit_obj, psbattle, highest_timestamp=1416475603.0
) #leave lowest_timestamp and highest_timestamp blank to get all submissions
img_count = 0
count_submissions = 1
img_count_train = 0
img_count_test = 0
link_file_train = open(LINK_FILE_TRAIN, "w")
link_file_test = open(LINK_FILE_TEST, "w")
while img_count < NUM_LINKS_TO_GET: #TODO need this?
for submission in submissions: #go through each submission
try:
if img_count > NUM_LINKS_TO_GET:
break
print "next submission: ", submission
if not submission:
print "Not a submission"
else:
print "submission id: ", submission.id
print "submission timestamp: ", submission.created_utc
#decide if images from this submission will be for training or test
link_file = None
testing = False #are we adding images to testing or training
#add first images to test (avoid divide by 0)
#then add to test if we have more than 5x number of training images as testing
#NOTE: add all images from a submission to train or test
#otherwise classifier might have very similar images in training dataset
if img_count_test == 0 or img_count_train / img_count_test >= 5:
testing = True
link_file = link_file_test
else:
link_file = link_file_train
#get all comments for now (note: may cause dataset imbalance?
#also takes longer because more API calls)
submission.replace_more_comments(
limit=None, threshold=0) #limit=None for all comments
comments = [comment for comment in submission.comments if \
not isinstance(comment, praw.objects.MoreComments) and comment.is_root][:NUM_IMGS_PER_COMMENT] #look for at most 10 images a comment
count_comments = 1
for comment in comments: #each (root) comment (containing image)
#if we've made it this far assume image is original
links = find_links(
comment.body) #get links (presumably of images)
#this link is valid so download image at this link
for link in links:
if "](" in link: #correct mistake made by regex: sometimes get http://imgur.com...](http://imgur.com...)
link = link[:link.index(
"]"
)] #get only up to and not including the ]
link = link.replace(
")", ""
) #because sometimes Imgur links have trailing )
link_file.write(link + " ")
link_file.write(submission.id + " ")
link_file.write(str(count_comments) + " ")
link_file.write(str(img_count + 1) + "\n")
img_count += 1
if testing:
img_count_test += 1
else:
img_count_train += 1
count_comments += 1 #count comment
count_submissions += 1 #count this submission as a new one
print("%d valid comments on submission %d. Now %d image links total: %d train, %d test" \
% (count_comments - 1, count_submissions - 1, img_count, img_count_train, img_count_test))
except Exception as e:
print "exception: ", e
break
#finish up
link_file_train.close()
link_file_test.close()
print("%d image links scraped in total" % img_count)
def scrape():
reddit_obj = praw.Reddit(user_agent = "Fake Images Scraper")
psbattle = reddit_obj.get_subreddit("photoshopbattles") #subreddit of interest
#submissions = psbattle.get_top_from_year(limit=None)#, params={"after" : "t3_4cd8qr"})
submissions = submissions_between(reddit_obj, psbattle, highest_timestamp = 1416475603.0) #leave lowest_timestamp and highest_timestamp blank to get all submissions
img_count = 0
count_submissions = 1
img_count_train = 0
img_count_test = 0
link_file_train = open(LINK_FILE_TRAIN, "w")
link_file_test = open(LINK_FILE_TEST, "w")
while img_count < NUM_LINKS_TO_GET: #TODO need this?
for submission in submissions: #go through each submission
try:
if img_count > NUM_LINKS_TO_GET:
break
print "next submission: ", submission
if not submission:
print "Not a submission"
else:
print "submission id: ", submission.id
print "submission timestamp: ", submission.created_utc
#decide if images from this submission will be for training or test
link_file = None
testing = False #are we adding images to testing or training
#add first images to test (avoid divide by 0)
#then add to test if we have more than 5x number of training images as testing
#NOTE: add all images from a submission to train or test
#otherwise classifier might have very similar images in training dataset
if img_count_test == 0 or img_count_train / img_count_test >= 5:
testing = True
link_file = link_file_test
else:
link_file = link_file_train
#get all comments for now (note: may cause dataset imbalance?
#also takes longer because more API calls)
submission.replace_more_comments(limit=None, threshold = 0) #limit=None for all comments
comments = [comment for comment in submission.comments if \
not isinstance(comment, praw.objects.MoreComments) and comment.is_root][:NUM_IMGS_PER_COMMENT] #look for at most 10 images a comment
count_comments = 1
for comment in comments: #each (root) comment (containing image)
#if we've made it this far assume image is original
links = find_links(comment.body) #get links (presumably of images)
#this link is valid so download image at this link
for link in links:
if "](" in link: #correct mistake made by regex: sometimes get http://imgur.com...](http://imgur.com...)
link = link[:link.index("]")] #get only up to and not including the ]
link = link.replace(")","") #because sometimes Imgur links have trailing )
link_file.write(link + " ")
link_file.write(submission.id + " ")
link_file.write(str(count_comments) + " ")
link_file.write(str(img_count + 1) + "\n")
img_count += 1
if testing:
img_count_test += 1
else:
img_count_train += 1
count_comments += 1 #count comment
count_submissions += 1 #count this submission as a new one
print("%d valid comments on submission %d. Now %d image links total: %d train, %d test" \
% (count_comments - 1, count_submissions - 1, img_count, img_count_train, img_count_test))
except Exception as e:
print "exception: ", e
break
#finish up
link_file_train.close()
link_file_test.close()
print("%d image links scraped in total" % img_count)
