Scripts to download files from google storage based on sha1 sums

upload_to_google_storage.py

Index: upload_to_google_storage.py
diff --git a/upload_to_google_storage.py b/upload_to_google_storage.py
new file mode 100755
index 0000000000000000000000000000000000000000..6a68fef4afdc7cc9dd98f7a98b92f14594ba797b
--- /dev/null
+++ b/upload_to_google_storage.py
@@ -0,0 +1,235 @@
+#!/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 hashlib
+import optparse
+import os
+import Queue
+import re
+import sys
+import threading
+import time
+
+from download_from_google_storage import check_bucket_permissions
+from download_from_google_storage import get_sha1
+from download_from_google_storage import Gsutil
+from download_from_google_storage import printer_worker
+
+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 get_md5(filename):
+  md5_calculator = hashlib.md5()
+  with open(filename, 'rb') as f:
+    while True:
+      chunk = f.read(1024*1024)
+      if not chunk:
+        break
+      md5_calculator.update(chunk)
+  return md5_calculator.hexdigest()
+
+
+def get_md5_cached(filename):
+  """Don't calculate the MD5 if we can find a .md5 file."""
+  # See if we can find an existing MD5 sum stored in a file.
+  if os.path.exists('%s.md5' % filename):
+    with open('%s.md5' % filename) as f:

[M-A Ruel, 2013/03/08 02:07:00]

I prefer 'rb' for consistency with the other calls

[Ryan Tseng, 2013/03/08 02:34:37]

Done.

+      md5_match = re.search('([a-z0-9]{32})', f.read())
+      if md5_match:
+        return md5_match.group(1)
+  else:
+    md5_hash = get_md5(filename)
+    with open('%s.md5' % filename, 'w') as f:
+      f.write(md5_hash)
+    return md5_hash
+
+
+def _upload_worker(
+    thread_num, q, base_url, gsutil, md5_lock, force,

[M-A Ruel, 2013/03/08 02:07:00]

s/q/upload_queue/

[Ryan Tseng, 2013/03/08 02:34:37]

Done.

+    use_md5, stdout_queue, ret_codes):
+  while True:
+    filename, sha1_sum = q.get()
+    if not filename:
+      break
+    file_url = '%s/%s' % (base_url, sha1_sum)
+    if gsutil.check_call('ls', file_url)[0] == 0 and not force:
+      # 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.
+        with md5_lock:
+          if use_md5:
+            local_md5 = get_md5_cached(filename)
+          else:
+            local_md5 = get_md5(filename)
+        if local_md5 == remote_md5:
+          stdout_queue.put(
+              '%d> File %s already exists at %s and MD5 matches, exiting' %
+              (thread_num, filename, file_url))
+          continue
+    stdout_queue.put('%d> Uploading %s to %s' % (
+        thread_num, filename, file_url))
+    code, _, err = gsutil.check_call('cp', '-q', filename, file_url)
+    if code != 0:
+      ret_codes.put(
+          (code,
+           'Encountered error on uploading %s to %s\n%s' %
+           (filename, file_url, err)))
+      continue
+
+
+def get_targets(args, parser, use_null_terminator):
+  if not args:
+    parser.error('Missing target.')
+
+  if len(args) == 1 and args[0] == '-':
+    # Take stdin as a newline or null seperated list of files.
+    if use_null_terminator:
+      return sys.stdin.read().split('\0')
+    else:
+      return sys.stdin.read().splitlines()
+  else:
+    return args
+
+
+def upload_to_google_storage(
+    input_filenames, base_url, gsutil, force,
+    use_md5, num_threads, skip_hashing):
+  # We only want one MD5 calculation happening at a time to avoid HD thrashing.
+  md5_lock = threading.Lock()
+
+  # Start up all the worker threads.
+  all_threads = []
+  ret_codes = Queue.Queue()
+  ret_codes.put((0, None))
+  upload_queue = Queue.Queue()
+  upload_timer = time.time()
+  stdout_queue = Queue.Queue()
+  for thread_num in range(num_threads):
+    t = threading.Thread(
+        target=_upload_worker,
+        args=[thread_num, upload_queue, base_url, gsutil.clone(), md5_lock,
+              force, use_md5, stdout_queue, ret_codes])
+    t.daemon = True
+    t.start()
+    all_threads.append(t)
+
+  # We want to hash everything in a single thread since its faster.
+  # The bottleneck is in disk IO, not CPU.
+  hashing_start = 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 skip_hashing:
+      print 'Found hash for %s, skipping.' % filename
+      with open(filename + '.sha1', 'rb') as f:
+        upload_queue.put((filename, f.read()))

[M-A Ruel, 2013/03/08 02:07:00]

you should limit the read size so the script doesn't end up reading a 1mb .sha1
file by error. Not that it'll likely happen but still.

And by "verification", I mean ensuring it's 40 hex characters.

[Ryan Tseng, 2013/03/08 02:34:37]

Oops, yes it should definitely verify that its a 40 char hex string.  It should
also throw an error and exit since something is almost certainly wrong.

+      continue
+    print 'Calculating hash for %s...' % filename,
+    sha1_sum = get_sha1(filename)
+    with open(filename + '.sha1', 'wb') as f:
+      f.write(sha1_sum)
+    print 'done'
+    upload_queue.put((filename, sha1_sum))
+  hashing_duration = time.time() - hashing_start
+
+  # Wait for everything to finish.
+  for _ in all_threads:
+    upload_queue.put((None, None))  # To mark the end of the work queue.
+  printer_thread = threading.Thread(target=printer_worker, args=[stdout_queue])

[M-A Ruel, 2013/03/08 02:07:00]

Start it earlier.

[Ryan Tseng, 2013/03/08 02:34:37]

Done.

+  printer_thread.daemon = True
+  printer_thread.start()
+  for t in all_threads:
+    t.join()
+  stdout_queue.put(None)
+  printer_thread.join()
+
+  # Print timing information.
+  print 'Hashing %s files took %1f seconds' % (
+      len(input_filenames), hashing_duration)
+  print 'Uploading took %1f seconds' % (time.time() - upload_timer)
+
+  # See if we ran into any errors.
+  max_ret_code = 0
+  for ret_code, message in ret_codes.queue:
+    max_ret_code = max(ret_code, max_ret_code)
+    if message:
+      print >> sys.stderr, message
+
+  if not max_ret_code:
+    print 'Success!'
+
+  return max_ret_code
+
+
+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',
+                    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',
+                    help='Skip hashing if .sha1 file exists.')
+  parser.add_option('-0', '--use_null_terminator', action='store_true',
+                    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()
+
+  # Enumerate our inputs.
+  input_filenames = get_targets(args, parser, options.use_null_terminator)
+
+  # 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 we have a valid bucket with valid permissions.
+  base_url, code = check_bucket_permissions(options.bucket, gsutil)
+  if code:
+    return code
+
+  return upload_to_google_storage(
+      input_filenames, base_url, gsutil, options.force, options.use_md5,
+      options.num_threads, options.skip_hashing)
+
+
+if __name__ == '__main__':
+  sys.exit(main(sys.argv))