Scripts to download files from google storage based on sha1 sums

upload_to_google_storage.py

+#!/usr/bin/env python
+# Copyright (c) 2012 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Uploads files to Google Storage content addressed."""
+
+import optparse
+import os
+import Queue
+import re
+import sys
+import threading
+import time
+
+from common import Gsutil

[M-A Ruel, 2013/02/25 15:15:06]

Replace with:
import common

[Ryan Tseng, 2013/02/27 02:06:56]

Done.

+from common import GetSHA1
+from common import GetMD5
+
+GSUTIL_DEFAULT_PATH = os.path.join(
+    os.path.dirname(os.path.abspath(__file__)),
+    'third_party', 'gsutil', 'gsutil')
+
+USAGE_STRING = """%prog [options] target [target2 ...].
+Target is the file intended to be uploaded to Google Storage.
+If target is "-", then a list of files will be taken from standard input
+
+This script will generate a file (original filename).sha1 containing the
+sha1 sum of the uploaded file.
+It is recommended that the .sha1 file is checked into the repository,
+the original file removed from the repository, and a hook added to the
+DEPS file to call download_from_google_storage.py.
+
+Example usages
+--------------
+
+Scan the current directory and upload all files larger than 1MB:
+find . -name .svn -prune -o -size +1000k -type f -print0 | %prog -0 -
+"""
+
+
+def _upload_worker(thread_num, q, base_url, gsutil, options, md5_lock):
+  while True:
+    try:
+      filename, sha1_sum = q.get_nowait()
+      file_url = '%s/%s' % (base_url, sha1_sum)
+      if gsutil.check_call('ls', file_url)[0] == 0 and not options.force:

[M-A Ruel, 2013/02/25 15:15:06]

I don't see gsutil being defined anywhere, did you test this code?

[Ryan Tseng, 2013/02/27 02:06:56]

The gsutil object is initialized in main() and passed into the worker thread as
an argument. 

+        # File exists, check MD5 hash.
+        _, out, _ = gsutil.check_call('ls', '-L', file_url)
+        etag_match = re.search('ETag:\s+([a-z0-9]{32})', out)
+        if etag_match:
+          remote_md5 = etag_match.group(1)
+          # Calculate the MD5 checksum to match it to Google Storage's ETag.
+          local_md5 = GetMD5(filename, md5_lock, options.use_md5)
+          if local_md5 == remote_md5:
+            print ('File %s already exists at %s and MD5 matches, exiting' %
+                   (filename, file_url))
+            continue
+      print 'Uploading %s to %s' % (filename, file_url)
+      code = gsutil.call('cp', '-q', filename, file_url)
+      if code != 0:
+        print >> sys.stderr, gsutil.stderr
+        continue
+    except Queue.Empty:
+      return
+
+
+def main(args):
+  parser = optparse.OptionParser(USAGE_STRING)
+  parser.add_option('-b', '--bucket',
+                    help='Google Storage bucket to upload to.')
+  parser.add_option('-e', '--boto', help='Specify a custom boto file.')
+  parser.add_option('-f', '--force', action='store_true',
+                    help='Force upload even if remote file exists.')
+  parser.add_option('-g', '--gsutil_path', default=GSUTIL_DEFAULT_PATH,
+                    help='Path to the gsutil script.')
+  parser.add_option('-m', '--use_md5', action='store_true', default=False,

[M-A Ruel, 2013/02/25 15:15:06]

Remove default=False everywhere, it's unnecessary.

[Ryan Tseng, 2013/02/27 02:06:56]

Done.

+                    help='Generate MD5 files when scanning, and don\'t check '
+                    'the MD5 checksum if a .md5 file is found.')
+  parser.add_option('-t', '--num_threads', default=1, type='int',
+                    help='Number of uploader threads to run.')
+  parser.add_option('-s', '--skip_hashing', action='store_true', default=False,
+                    help='Skip hashing if .sha1 file exists.')
+  parser.add_option('-0', '--use_null_terminator', action='store_true',
+                    default=False, help='Use \\0 instead of \\n when parsing '
+                    'the file list from stdin.  This is useful if the input '
+                    'is coming from "find ... -print0".')
+  (options, args) = parser.parse_args()
+
+  if len(args) < 1:

[M-A Ruel, 2013/02/25 15:15:06]

if not args:

[Ryan Tseng, 2013/02/27 02:06:56]

Done.

+    parser.error('Missing target.')
+  elif len(args) == 1 and args[0] == '-':
+    # Take stdin as a newline or null seperated list of files.
+    if options.use_null_terminator:
+      input_filenames = sys.stdin.read().split('\0')
+    else:
+      input_filenames = sys.stdin.read().splitlines()
+  else:
+    input_filenames = args
+
+  if not options.bucket:
+    parser.error('Missing bucket.  Specify bucket with --bucket.')
+  base_url = 'gs://%s' % options.bucket
+
+  # Make sure we can find a working instance of gsutil.
+  if os.path.exists(GSUTIL_DEFAULT_PATH):
+    gsutil = Gsutil(GSUTIL_DEFAULT_PATH)
+  else:
+    print >> sys.stderr, ('gsutil not found in %s, bad depot_tools checkout?' %
+                          GSUTIL_DEFAULT_PATH)
+    return 1
+
+  # Check if we have permissions to the Google Storage bucket.

[M-A Ruel, 2013/02/25 15:15:06]

Can you split the rest of this code into its separate function? I'm a big fan of
splitting argument processing from actual "main processing".

[Ryan Tseng, 2013/02/27 02:06:56]

Done.

+  code, _, ls_err = gsutil.check_call('ls', base_url)
+  if code == 403:
+    code, _, _ = gsutil.call('config')
+    if code != 0:
+      print >> sys.stderr, 'Error while authenticating to %s.' % base_url
+      return 403
+  elif code == 404:
+    print >> sys.stderr, '%s not found.' % base_url
+    return 404
+  elif code != 0:
+    print >> sys.stderr, ls_err
+    return code
+
+  # We want to hash everything in a single thread since its faster.
+  # The bottleneck is in disk IO, not CPU.
+  upload_queue = Queue.Queue()
+  hash_timer = time.time()
+  for filename in input_filenames:
+    if not os.path.exists(filename):
+      print 'Error: %s not found, skipping.' % filename
+      continue
+    if os.path.exists('%s.sha1' % filename) and options.skip_hashing:
+      print 'Found hash for %s, skipping.' % filename
+      upload_queue.put((filename, open('%s.sha1' % filename).read()))
+      continue
+    print 'Calculating hash for %s...' % filename,
+    sha1_sum = GetSHA1(filename)
+    with open(filename + '.sha1', 'wb') as f:
+      f.write(sha1_sum)
+    print 'done'
+    upload_queue.put((filename, sha1_sum))
+  hash_time = time.time() - hash_timer
+
+  # Start up all the worker threads.
+  all_threads = []
+
+  # We only want one MD5 calculation happening at a time.
+  md5_lock = threading.Lock()
+  upload_timer = time.time()
+
+  for thread_num in range(options.num_threads):
+    t = threading.Thread(target=_upload_worker, args=[thread_num,

[M-A Ruel, 2013/02/25 15:15:06]

Argument alignement
Start the threads before enqueuing.

[Ryan Tseng, 2013/02/27 02:06:56]

Done.

+        upload_queue, base_url, gsutil.clone(), options, md5_lock])
+    t.daemon = True
+    t.start()
+    all_threads.append(t)
+
+  # Wait for everything to finish.
+  for t in all_threads:
+    t.join()
+
+  print 'Success.'
+  print 'Hashing %s files took %1f seconds' % (len(input_filenames), hash_time)
+  print 'Uploading took %1f seconds' % (time.time() - upload_timer)
+  return 0
+
+
+if __name__ == '__main__':
+  sys.exit(main(sys.argv))