Improve download_list

bitter/VERSION

@@ -1 +1 @@
-0.9.3
+0.9.5

bitter/cli.py

@@ -42,10 +42,58 @@ def main(ctx, verbose, logging_level, config, credentials):
       utils.copy_credentials_to_config(credentials, config)
 
 
-@main.group()
+@main.group(invoke_without_command=True)
 @click.pass_context
 def credentials(ctx):
-    pass
+    wq = crawlers.TwitterQueue.from_config(conffile=bconf.CONFIG_FILE)
+    for worker in wq.queue:
+        print('#'*20)
+        try:
+            resp = worker.client.application.rate_limit_status()
+            print(worker.name)
+        except Exception as ex:
+            print('{}: AUTHENTICATION ERROR: {}'.format(worker.name, ex) )
+
+
+@credentials.command('limits')
+@click.option('--all', type=bool, default=False, required=False,
+              help=('Print all limits. By default, it only limits that '
+                    'have been consumed will be shown.'))
+@click.argument('url', required=False)
+@click.pass_context
+def get_limits(ctx, all, url):
+    wq = crawlers.TwitterQueue.from_config(conffile=bconf.CONFIG_FILE)
+    total = {}
+    for worker in wq.queue:
+        resp = worker.client.application.rate_limit_status()
+        print('#'*20)
+        print(worker.name)
+        if url:
+            limit = 'NOT FOUND'
+            print('URL is: {}'.format(url))
+            cat = url.split('/')[1]
+            if cat in resp['resources']:
+                limit = resp['resources'][cat].get(url, None) or resp['resources'][cat]
+            else:
+                print('Cat {} not found'.format(cat))
+                continue
+            for k in limit:
+                total[k] = total.get(k, 0) + limit[k]
+            print('{}: {}'.format(url, limit))
+            continue
+        nres = {}
+        if not all:
+            for res, urls in resp['resources'].items():
+                nurls = {}
+                for u, limits in urls.items():
+                    if limits['limit'] != limits['remaining']:
+                        nurls[u] = limits
+                if nurls:
+                    nres[res] = nurls
+            resp = nres
+        print(json.dumps(resp, indent=2))
+    if url:
+        print('Total for {}: {}'.format(url, total))
 
 @credentials.command('add')
 @click.option('--consumer_key', default=None)
@@ -98,11 +146,17 @@ def get_tweets(ctx, tweetsfile, folder, update, retry, delimiter, header, quotec
         click.echo('Cancelling')
         return
     wq = crawlers.TwitterQueue.from_config(conffile=bconf.CONFIG_FILE)
-    for i in utils.download_file(wq, tweetsfile, folder, delimiter=delimiter,
+
-                                 batch_method=utils.tweet_download_batch,
+    status = tqdm('Queried')
-                                 header=header, quotechar=quotechar,
+    failed = 0
-                                 column=column, update=update, retry_failed=retry):
+    for tid, obj in utils.download_file(wq, tweetsfile, folder, delimiter=delimiter,
-        pass
+                                        batch_method=utils.tweet_download_batch,
+                                        header=header, quotechar=quotechar,
+                                        column=column, update=update, retry_failed=retry):
+        status.update(1)
+        if not obj:
+            failed += 1
+            status.set_description('Failed: %s. Queried' % failed, refresh=True)
 
 @tweet.command('search')
 @click.argument('query')
@@ -383,34 +437,6 @@ def reset_extractor(ctx):
     session = make_session(db)
     session.query(ExtractorEntry).filter(ExtractorEntry.pending==True).update({'pending':False})
 
-@main.command('limits')
-@click.argument('url', required=False)
-@click.pass_context
-def get_limits(ctx, url):
-    wq = crawlers.TwitterQueue.from_config(conffile=bconf.CONFIG_FILE)
-    total = {}
-    for worker in wq.queue:
-        resp = worker.client.application.rate_limit_status()
-        print('#'*20)
-        print(worker.name)
-        if url:
-            limit = 'NOT FOUND'
-            print('URL is: {}'.format(url))
-            cat = url.split('/')[1]
-            if cat in resp['resources']:
-                limit = resp['resources'][cat].get(url, None) or resp['resources'][cat]
-            else:
-                print('Cat {} not found'.format(cat))
-                continue
-            for k in limit:
-                total[k] = total.get(k, 0) + limit[k]
-            print('{}: {}'.format(url, limit))
-        else:
-            print(json.dumps(resp, indent=2))
-    if url:
-        print('Total for {}: {}'.format(url, total))
-
-
 
 @main.command(context_settings=dict(ignore_unknown_options=True, allow_extra_args=False),
               help='''Issue a call to an endpoint of the Twitter API.''')

bitter/utils.py

@@ -13,6 +13,11 @@ import sqlalchemy
 import os
 import multiprocessing
 from multiprocessing.pool import ThreadPool
+from multiprocessing import Queue
+
+import queue
+import threading
+from select import select
 
 from functools import partial
 
@@ -22,6 +27,7 @@ from itertools import islice, chain
 from contextlib import contextmanager
 
 from collections import Counter
+from random import choice
 
 from builtins import map, filter
 
@@ -53,7 +59,7 @@ def chunk(iterable, n):
 def parallel(func, source, chunksize=1, numcpus=multiprocessing.cpu_count()):
     source = chunk(source, chunksize)
     p = ThreadPool(numcpus*2)
-    results = p.imap_unordered(func, source, chunksize=int(1000/numcpus))
+    results = p.imap_unordered(func, source)
     for i in chain.from_iterable(results):
         yield i
 
@@ -507,7 +513,8 @@ def id_failed(oid, folder):
 
 def tweet_download_batch(wq, batch):
     tweets = wq.statuses.lookup(_id=",".join(batch), map=True)['id']
-    return tweets.items()
+    for tid, tweet in tweets.items():
+        yield tid, tweet
 
 def user_download_batch(wq, batch):
     screen_names = []
@@ -547,45 +554,81 @@ def user_download_batch(wq, batch):
         yield (name, None)
 
 
-def download_list(wq, lst, folder, update=False, retry_failed=False, ignore_fails=True,
+def dump_result(oid, obj, folder, ignore_fails=True):
+    if obj:
+        try:
+            write_json(obj, folder=folder, oid=oid)
+            failed = fail_file(oid, folder)
+            if os.path.exists(failed):
+                os.remove(failed)
+        except Exception as ex:
+            logger.error('%s: %s' % (oid, ex))
+            if not ignore_fails:
+                raise
+    else:
+        logger.info('Object not recovered: {}'.format(oid))
+        with open(fail_file(oid, folder), 'w') as f:
+            print('Object not found', file=f)
+
+def download_list(wq, lst, folder, update=False, retry_failed=False, ignore_fails=False,
                   batch_method=tweet_download_batch):
-    def filter_lines(line):
+
-        # print('Checking {}'.format(line))
+    done = Queue()
-        oid = line[0]
+
-        if (cached_id(oid, folder) and not update) or (id_failed(oid, folder) and not retry_failed):
+    down = Queue()
-            yield None
+
-        else:
+
-            yield str(oid)
+    def filter_list(lst, done, down):
-
+        print('filtering')
-    def print_result(res):
+        for oid in lst:
-        for oid, obj in res:
+            # print('Checking {}'.format(line))
-          if obj:
+            cached = cached_id(oid, folder)
-              try:
+            if (cached and not update):
-                  write_json(obj, folder=folder, oid=oid)
+                done.put((oid, cached))
-                  failed = fail_file(oid, folder)
+            elif (id_failed(oid, folder) and not retry_failed):
-                  if os.path.exists(failed):
+                done.put((oid, None))
-                      os.remove(failed)
+            else:
-                  yield 1
+                down.put(oid)
-              except Exception as ex:
+        down.put(None)
-                  logger.error('%s: %s' % (oid, ex))
+
-                  if not ignore_fails:
+    def download_results(batch_method, down, done):
-                      raise
+        def gen():
-          else:
+            while True:
-              logger.info('Object not recovered: {}'.format(oid))
+                r = down.get()
-              with open(fail_file(oid, folder), 'w') as f:
+                if not r:
-                  print('Object not found', file=f)
+                    return
-              yield -1
+                yield r
-
+
-    objects_to_crawl = filter(lambda x: x is not None, tqdm(parallel(filter_lines, lst), desc='Total objects'))
+        for t in parallel(batch_method, gen(), 100):
-    batch_method = partial(batch_method, wq)
+            done.put(t)
-    objects = parallel(batch_method, objects_to_crawl, 100)
+
-    failed = 0
+    def batch(*args, **kwargs):
-    pbar = tqdm(parallel(print_result, objects), desc='Queried')
+        return batch_method(wq, *args, **kwargs)
-    for res in pbar:
+
-        if res < 0:
+    tc = threading.Thread(target=filter_list, args=(lst, done, down), daemon=True)
-            failed += 1
+    tc.start()
-            pbar.set_description('Failed: %s. Queried' % failed, refresh=True)
+    td = threading.Thread(target=download_results, args=(batch, down, done), daemon=True)
-        yield res
+    td.start()
+
+    def check_threads(ts, done):
+        for t in ts:
+            t.join()
+        done.put(None)
+
+    wait = threading.Thread(target=check_threads, args=([tc, td], done), daemon=True)
+    wait.start()
+
+    while True:
+        rec = done.get()
+
+        if rec is None:
+            break
+
+        oid, obj = rec
+        dump_result(oid, obj, folder, ignore_fails)
+        yield rec
+
+    wait.join()
 
 
 def download_file(wq, csvfile, folder, column=0, delimiter=',',
@@ -595,8 +638,14 @@ def download_file(wq, csvfile, folder, column=0, delimiter=',',
         csvreader = csv.reader(f, delimiter=str(delimiter), quotechar=str(quotechar))
         if header:
             next(csvreader)
-        tweets = map(lambda row: row[0].strip(), csvreader)
+
-        for res in download_list(wq, tweets, folder, batch_method=batch_method,
+        def reader(r):
+            for row in csvreader:
+                if len(row) > column:
+                    yield row[column].strip()
+
+
+        for res in download_list(wq, reader(csvreader), folder, batch_method=batch_method,
                                  **kwargs):
             yield res