Scripts to download files from google storage based on sha1 sums

build/download_from_google_storage.py

Index: build/download_from_google_storage.py
diff --git a/build/download_from_google_storage.py b/build/download_from_google_storage.py
new file mode 100755
index 0000000000000000000000000000000000000000..784ee802038d38b2f682fafd91a6620a49351e0f
--- /dev/null
+++ b/build/download_from_google_storage.py
@@ -0,0 +1,267 @@
+#!/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.
+
+"""Script to download files from Google Storage."""
+
+
+import hashlib
+import optparse
+import os
+import Queue
+import re
+import subprocess
+import sys
+import tempfile
+import threading
+import time
+import zipfile
+
+# TODO(hinoka): This is currently incorrect. Should find a better default.

[M-A Ruel, 2013/01/17 19:53:17]

Why is it incorrect.

+GSUTIL_DEFAULT_PATH = os.path.join(
+    os.path.dirname(os.path.normpath(__file__)),

[M-A Ruel, 2013/01/17 19:53:17]

You want abspath or realpath, not normpath.

+    '..', '..', 'third_party', 'gsutil', 'gsutil')
+
+
+class Gsutil():

[M-A Ruel, 2013/01/17 19:53:17]

object

+  def __init__(self, path, boto_path=None, timeout=None):

[M-A Ruel, 2013/01/17 19:53:17]

Are the default values necessary?

+    if not os.path.exists(path):
+      raise OSError('GSUtil not found in %s' % path)
+    self.path = path
+
+    self.timeout = timeout
+    self.boto_path = boto_path
+
+  def call(self, *args):
+    def _thread_main():
+      thr = threading.current_thread()
+      env = os.environ.copy()

[M-A Ruel, 2013/01/17 19:53:17]

You can make this in the main thread instead of the worker thread.

+      if self.boto_path is not None:
+        env['AWS_CREDENTIAL_FILE'] = self.boto_path
+      thr.status = subprocess.call((sys.executable, self.path) + args, env=env)

[M-A Ruel, 2013/01/17 19:53:17]

subprocess2.check_call(timeout=) works well and is unit tested.

+    t = threading.Thread(target=_thread_main)
+    t.start()
+    t.join(self.timeout)
+    if thr.isAlive():
+      raise RuntimeError('%s %s timed out after %d seconds.' % (
+              self.path, ' '.join(args), self.timeout))
+    return thr.status
+
+  def check_call(self, *args):
+    def _thread_main():
+      thr = threading.current_thread()
+      env = os.environ.copy()
+      if self.boto_path is not None:
+        env['AWS_CREDENTIAL_FILE'] = self.boto_path
+      p = subprocess.Popen((sys.executable, self.path) + args,
+                           stdout=subprocess.PIPE,
+                           stderr=subprocess.PIPE,
+                           env=env)
+      code = p.wait()
+      out, err = p.communicate()
+      thr.status = (code, out, err)
+
+    thr = threading.Thread(target=_thread_main)
+    thr.start()
+    thr.join(self.timeout)
+    if thr.isAlive():
+      raise RuntimeError('%s %s timed out after %d seconds.' % (
+              self.path, ' '.join(args),  self.timeout))
+    code, out, err = thr.status
+    status_code_match = re.search('status=([0-9]+)', err)
+    if status_code_match:
+      return int(status_code_match.groups(1))
+    elif ('You are attempting to access protected data with '
+          'no configured credentials.' in err):
+      return (403, out, err)
+    elif 'No such object' in err:
+      return (404, out, err)
+    else:
+      return (code, out, err)
+
+  def clone(self):
+    return Gsutil(self.path, self.boto_path, self.timeout)
+
+
+def CheckSHA1(sha1_sum, filename):
+  sha1 = hashlib.sha1()
+  with open(filename, 'rb') as f:
+    while True:
+      # Read in 1mb chunks, so it doesn't all have to be loaded into memory.
+      chunk = f.read(1024*1024)
+      if not chunk:
+        break
+      sha1.update(chunk)
+  return sha1_sum == sha1.hexdigest()
+
+
+def _downloader_worker_thread(thread_num, q, options, base_url, gsutil):
+  while True:
+    try:
+      input_sha1_sum, output_filename = q.get_nowait()

[M-A Ruel, 2013/01/17 19:53:17]

Please move the except: up here and move the rest out of the try/except.

+      if os.path.exists(output_filename) and not options.force:
+        if CheckSHA1(input_sha1_sum, output_filename):
+          print 'File %s exists and SHA1 sum (%s) matches. Skipping.' % (
+              output_filename , input_sha1_sum)
+          continue
+      # Check if file exists.
+      file_url = '%s/%s' % (base_url, input_sha1_sum)
+      if gsutil.check_call('ls', file_url) != 0:
+        print >>sys.stderr, 'File %s for %s does not exist, skipping.' % (

[M-A Ruel, 2013/01/17 19:53:17]

You are outputing from a thread? I'd recommend against that.

+            file_url, output_filename)
+        continue
+      # Fetch the file.
+      print 'Downloading %s to %s...' % (file_url, output_filename)
+      code, out, err = gsutil.call('cp', '-q', file_url, output_filename)
+      if code != 0:
+        print >>sys.stderr, gsutil.stderr
+        return code
+    except Queue.Empty:
+      return
+
+
+def main(args):
+  usage = ('usage: %prog [options] target\nTarget must be:\n'
+           '(default) a sha1 sum ([A-Za-z0-9]{40}).\n(-s or --sha1_file) a '
+           '.sha1 file, containing a sha1 sum on the first line. (-d or '
+           '--directory) A directory to scan for .sha1 files. ')
+  parser = optparse.OptionParser(usage)
+  parser.add_option('-o', '--output', default=None,
+                    help='Specify the output file name. Defaults to:\n'
+                         '(a) Given a SHA1 hash, the name is the SHA1 hash.\n'
+                         '(b) Given a .sha1 file or directory, the name will '
+                         'match (.*).sha1.')
+  parser.add_option('-b', '--bucket', default='chrome-artifacts',
+                    help='Google Storage bucket to fetch from.')
+  parser.add_option('-e', '--boto', default=None,
+                    help='Specify a custom boto file.')
+  parser.add_option('-c', '--no_resume', action='store_true', default=False,
+                    help='Resume download if file is partially downloaded.')
+  parser.add_option('-f', '--force', action='store_true', default=False,
+                    help='Force download even if local file exists.')
+  parser.add_option('-r', '--recursive', action='store_true', default=False,
+                    help='Scan folders recursively for .sha1 files. '
+                         'Must be used with -d/--directory')
+  parser.add_option('-t', '--num_threads', default=1, type='int',
+                    help='Number of downloader threads to run.')
+  parser.add_option('-d', '--directory', action='store_true', default=False,
+                    help='The target is a directory.  '
+                         'Cannot be used with -s/--sha1_file.')
+  parser.add_option('-s', '--sha1_file', action='store_true', default=False,
+                    help='The target is a file containing a sha1 sum.  '

[M-A Ruel, 2013/01/17 19:53:17]

What's the goal of this flag? Is it necessary/useful?

+                         'Cannot be used with -d/--directory.')
+  # This file should be stored in tools/deps_scripts/ and we want the path to
+  # third_party/gsutil/gsutil
+  parser.add_option('-g', '--gsutil_path', default=GSUTIL_DEFAULT_PATH,
+          help='Path to the gsutil script.')
+
+  (options, args) = parser.parse_args()
+  if len(args) < 1:
+    parser.error('Missing target.')
+  elif len(args) > 1:

[M-A Ruel, 2013/01/17 19:53:17]

s/elif/if/ on all of these

+    # TODO(hinoka): Multi target support.

[M-A Ruel, 2013/01/17 19:53:17]

Not necessary.

+    parser.error('Too many targets.')
+  elif options.sha1_file and options.directory:
+    parser.error('Both --directory and --sha1_file are specified, '
+                 'can only specify one.')
+  elif options.recursive and not options.directory:
+    parser.error('--recursive specified but --directory not specified.')
+  elif options.output and options.directory:
+    parser.error('--directory is specified, so --output has no effect.')
+  else:
+    input_filename = args[0]
+
+  # Set output filename if not specified.
+  if not options.output and not options.directory:
+    if not options.sha1_file:
+      # Target is a sha1 sum, so output filename would also be the sha1 sum.
+      options.output = input_filename
+    elif options.sha1_file:
+      # Target is a .sha1 file.
+      if not input_filename.endswith('.sha1'):
+        parser.error('--sha1_file is specified, but the input filename '
+                     'does not end with .sha1, and no --output is specified. '
+                     'Either make sure the input filename has a .sha1 '
+                     'extension, or specify --output.')
+      options.output = input_filename[:-5]
+    else:
+      raise Exception('Unreachable state.')

[M-A Ruel, 2013/01/17 19:53:17]

raise NotImplementedError()

+
+  # Check if output file already exists.
+  if not options.directory and not options.force and not options.no_resume:
+    if os.path.exists(options.output):
+      parser.error('Output file %s exists and --no_resume is specified.'
+                   % options.output)
+
+
+  base_url = 'gs://%s' % options.bucket
+
+  # Make sure we can find a working instance of gsutil.
+  if os.path.exists(options.gsutil_path):
+    gsutil = Gsutil(options.gsutil_path, boto_path=options.boto)
+  else:
+    for path in os.environ["PATH"].split(os.pathsep):
+      if os.path.exists(path) and 'gsutil' in os.listdir(path):
+        gsutil = Gsutil(os.path.join(path, 'gsutil'), boto_path=options.boto)
+
+  # Check if we have permissions to the Google Storage bucket.

[M-A Ruel, 2013/01/17 19:53:17]

Since the argument parsing code is non trivial, which includes finding gsutil,
I'd like you to split the actual work into another function. You can name it
"download_from_google_storage"

+  code, ls_out, 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, exiting' % 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
+
+  # Enumerate our work queue.
+  work_queue = Queue.Queue()
+  work_queue_size = 0
+  if options.directory:
+    if options.recursive:
+      for root, dirs, files in os.walk(input_filename):
+        if '.svn' in dirs:

[M-A Ruel, 2013/01/17 19:53:17]

And .git?

+          dirs.remove('.svn')
+        if not options.recursive:
+          for item in dirs:
+            dirs.remove(item)
+        for filename in files:
+          full_path = os.path.join(root, filename)
+          if full_path.endswith('.sha1'):
+            with open(full_path) as f:
+              sha1_match = re.search('([A-Za-z0-9]{40})', f.read(1024))

[M-A Ruel, 2013/01/17 19:53:17]

match(r'^...$', ...) ?

+            if sha1_match:

[M-A Ruel, 2013/01/17 19:53:17]

Silently drop otherwise?

+              work_queue.put((sha1_match.groups(1)[0],
+                              full_path.replace('.sha1', '')))
+              work_queue_size += 1
+  else:
+    work_queue.put((input_filename, options.output))
+    work_queue_size += 1
+
+  # Start up all the worker threads.

[M-A Ruel, 2013/01/17 19:53:17]

Why not start up before the enumeration?

+  all_threads = []
+  download_timer = time.time()
+  for thread_num in range(options.num_threads):
+    t = threading.Thread(target=_downloader_worker_thread, args=[thread_num,
+        work_queue, options, base_url, gsutil.clone()])
+    t.daemon = True
+    t.start()
+    all_threads.append(t)
+
+  # Wait for all downloads to finish.
+  for t in all_threads:
+    t.join()
+
+  print 'Success.'
+  print 'Downloading %d files took %1f second(s)' % (
+      work_queue_size, time.time() - download_timer)
+  return 0
+
+
+if __name__ == '__main__':
+  sys.exit(main(sys.argv))