Merge branch '2411-check-copyright'

[arvados.git] / sdk / python / arvados / commands / put.py

diff --git a/sdk/python/arvados/commands/put.py b/sdk/python/arvados/commands/put.py
index 66e122cd5e36551b6a441ea9a179e107839bfdfb..548f4b0948ae715393eb657a1693364e8b500639 100644 (file)
--- a/sdk/python/arvados/commands/put.py
+++ b/sdk/python/arvados/commands/put.py
@@ -1,8 +1,11 @@
-#!/usr/bin/env python
-
-# TODO:
-# --md5sum - display md5 of each file as read from disk
-
+# Copyright (C) The Arvados Authors. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from __future__ import division
+from future.utils import listitems, listvalues
+from builtins import str
+from builtins import object

 import argparse
 import arvados
 import arvados.collection
 import argparse
 import arvados
 import arvados.collection


@@ -23,6 +26,8 @@ import sys
 import tempfile
 import threading
 import time
 import tempfile
 import threading
 import time

+import traceback
+

 from apiclient import errors as apiclient_errors
 from arvados._version import __version__
 
 from apiclient import errors as apiclient_errors
 from arvados._version import __version__
 


@@ -38,7 +43,9 @@ upload_opts.add_argument('--version', action='version',
                          help='Print version and exit.')
 upload_opts.add_argument('paths', metavar='path', type=str, nargs='*',
                          help="""
                          help='Print version and exit.')
 upload_opts.add_argument('paths', metavar='path', type=str, nargs='*',
                          help="""

-Local file or directory. Default: read from standard input.
+Local file or directory. If path is a directory reference with a trailing
+slash, then just upload the directory's contents; otherwise upload the
+directory itself. Default: read from standard input.

 """)
 
 _group = upload_opts.add_mutually_exclusive_group()
 """)
 
 _group = upload_opts.add_mutually_exclusive_group()


@@ -52,6 +59,12 @@ Normalize the manifest by re-ordering files and streams after writing
 data.
 """)
 
 data.
 """)
 

+_group.add_argument('--dry-run', action='store_true', default=False,
+                    help="""
+Don't actually upload files, but only check if any file should be
+uploaded. Exit with code=2 when files are pending for upload.
+""")
+

 _group = upload_opts.add_mutually_exclusive_group()
 
 _group.add_argument('--as-stream', action='store_true', dest='stream',
 _group = upload_opts.add_mutually_exclusive_group()
 
 _group.add_argument('--as-stream', action='store_true', dest='stream',


@@ -127,6 +140,15 @@ physical storage devices (e.g., disks) should have a copy of each data
 block. Default is to use the server-provided default (if any) or 2.
 """)
 
 block. Default is to use the server-provided default (if any) or 2.
 """)
 

+upload_opts.add_argument('--threads', type=int, metavar='N', default=None,
+                         help="""
+Set the number of upload threads to be used. Take into account that
+using lots of threads will increase the RAM requirements. Default is
+to use 2 threads.
+On high latency installations, using a greater number will improve
+overall throughput.
+""")
+

 run_opts = argparse.ArgumentParser(add_help=False)
 
 run_opts.add_argument('--project-uuid', metavar='UUID', help="""
 run_opts = argparse.ArgumentParser(add_help=False)
 
 run_opts.add_argument('--project-uuid', metavar='UUID', help="""


@@ -168,6 +190,16 @@ _group.add_argument('--no-resume', action='store_false', dest='resume',
 Do not continue interrupted uploads from cached state.
 """)
 
 Do not continue interrupted uploads from cached state.
 """)
 

+_group = run_opts.add_mutually_exclusive_group()
+_group.add_argument('--follow-links', action='store_true', default=True,
+                    dest='follow_links', help="""
+Follow file and directory symlinks (default).
+""")
+_group.add_argument('--no-follow-links', action='store_false', dest='follow_links',
+                    help="""
+Do not follow file and directory symlinks.
+""")
+

 _group = run_opts.add_mutually_exclusive_group()
 _group.add_argument('--cache', action='store_true', dest='use_cache', default=True,
                     help="""
 _group = run_opts.add_mutually_exclusive_group()
 _group.add_argument('--cache', action='store_true', dest='use_cache', default=True,
                     help="""


@@ -188,7 +220,7 @@ def parse_arguments(arguments):
     if len(args.paths) == 0:
         args.paths = ['-']
 
     if len(args.paths) == 0:
         args.paths = ['-']
 

-    args.paths = map(lambda x: "-" if x == "/dev/stdin" else x, args.paths)
+    args.paths = ["-" if x == "/dev/stdin" else x for x in args.paths]

 
     if len(args.paths) != 1 or os.path.isdir(args.paths[0]):
         if args.filename:
 
     if len(args.paths) != 1 or os.path.isdir(args.paths[0]):
         if args.filename:


@@ -219,6 +251,10 @@ def parse_arguments(arguments):
     return args
 
 
     return args
 
 

+class PathDoesNotExistError(Exception):
+    pass
+
+

 class CollectionUpdateError(Exception):
     pass
 
 class CollectionUpdateError(Exception):
     pass
 


@@ -226,9 +262,30 @@ class CollectionUpdateError(Exception):
 class ResumeCacheConflict(Exception):
     pass
 
 class ResumeCacheConflict(Exception):
     pass
 

+

 class ArvPutArgumentConflict(Exception):
     pass
 
 class ArvPutArgumentConflict(Exception):
     pass
 

+
+class ArvPutUploadIsPending(Exception):
+    pass
+
+
+class ArvPutUploadNotPending(Exception):
+    pass
+
+
+class FileUploadList(list):
+    def __init__(self, dry_run=False):
+        list.__init__(self)
+        self.dry_run = dry_run
+
+    def append(self, other):
+        if self.dry_run:
+            raise ArvPutUploadIsPending()
+        super(FileUploadList, self).append(other)
+
+

 class ResumeCache(object):
     CACHE_DIR = '.cache/arvados/arv-put'
 
 class ResumeCache(object):
     CACHE_DIR = '.cache/arvados/arv-put'
 


@@ -240,13 +297,13 @@ class ResumeCache(object):
     @classmethod
     def make_path(cls, args):
         md5 = hashlib.md5()
     @classmethod
     def make_path(cls, args):
         md5 = hashlib.md5()

-        md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost'))
+        md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())

         realpaths = sorted(os.path.realpath(path) for path in args.paths)
         realpaths = sorted(os.path.realpath(path) for path in args.paths)

-        md5.update('\0'.join(realpaths))
+        md5.update(b'\0'.join([p.encode() for p in realpaths]))

         if any(os.path.isdir(path) for path in realpaths):
         if any(os.path.isdir(path) for path in realpaths):

-            md5.update("-1")
+            md5.update(b'-1')

         elif args.filename:
         elif args.filename:

-            md5.update(args.filename)
+            md5.update(args.filename.encode())

         return os.path.join(
             arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'),
             md5.hexdigest())
         return os.path.join(
             arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'),
             md5.hexdigest())


@@ -320,9 +377,11 @@ class ArvPutUploadJob(object):
 
     def __init__(self, paths, resume=True, use_cache=True, reporter=None,
                  bytes_expected=None, name=None, owner_uuid=None,
 
     def __init__(self, paths, resume=True, use_cache=True, reporter=None,
                  bytes_expected=None, name=None, owner_uuid=None,

-                 ensure_unique_name=False, num_retries=None, replication_desired=None,
-                 filename=None, update_time=20.0, update_collection=None,
-                 logger=logging.getLogger('arvados.arv_put')):
+                 ensure_unique_name=False, num_retries=None,
+                 put_threads=None, replication_desired=None,
+                 filename=None, update_time=60.0, update_collection=None,
+                 logger=logging.getLogger('arvados.arv_put'), dry_run=False,
+                 follow_links=True):

         self.paths = paths
         self.resume = resume
         self.use_cache = use_cache
         self.paths = paths
         self.resume = resume
         self.use_cache = use_cache


@@ -336,6 +395,7 @@ class ArvPutUploadJob(object):
         self.ensure_unique_name = ensure_unique_name
         self.num_retries = num_retries
         self.replication_desired = replication_desired
         self.ensure_unique_name = ensure_unique_name
         self.num_retries = num_retries
         self.replication_desired = replication_desired

+        self.put_threads = put_threads

         self.filename = filename
         self._state_lock = threading.Lock()
         self._state = None # Previous run state (file list & manifest)
         self.filename = filename
         self._state_lock = threading.Lock()
         self._state = None # Previous run state (file list & manifest)


@@ -344,16 +404,25 @@ class ArvPutUploadJob(object):
         self._collection_lock = threading.Lock()
         self._remote_collection = None # Collection being updated (if asked)
         self._local_collection = None # Collection from previous run manifest
         self._collection_lock = threading.Lock()
         self._remote_collection = None # Collection being updated (if asked)
         self._local_collection = None # Collection from previous run manifest

-        self._file_paths = [] # Files to be updated in remote collection
+        self._file_paths = set() # Files to be updated in remote collection

         self._stop_checkpointer = threading.Event()
         self._checkpointer = threading.Thread(target=self._update_task)
         self._stop_checkpointer = threading.Event()
         self._checkpointer = threading.Thread(target=self._update_task)

+        self._checkpointer.daemon = True

         self._update_task_time = update_time  # How many seconds wait between update runs
         self._update_task_time = update_time  # How many seconds wait between update runs

-        self._files_to_upload = []
+        self._files_to_upload = FileUploadList(dry_run=dry_run)
+        self._upload_started = False

         self.logger = logger
         self.logger = logger

+        self.dry_run = dry_run
+        self._checkpoint_before_quit = True
+        self.follow_links = follow_links

 
         if not self.use_cache and self.resume:
             raise ArvPutArgumentConflict('resume cannot be True when use_cache is False')
 
 
         if not self.use_cache and self.resume:
             raise ArvPutArgumentConflict('resume cannot be True when use_cache is False')
 

+        # Check for obvious dry-run responses
+        if self.dry_run and (not self.use_cache or not self.resume):
+            raise ArvPutUploadIsPending()
+

         # Load cached data if any and if needed
         self._setup_state(update_collection)
 
         # Load cached data if any and if needed
         self._setup_state(update_collection)
 


@@ -361,20 +430,32 @@ class ArvPutUploadJob(object):
         """
         Start supporting thread & file uploading
         """
         """
         Start supporting thread & file uploading
         """

-        self._checkpointer.daemon = True
-        self._checkpointer.start()
+        if not self.dry_run:
+            self._checkpointer.start()

         try:
             for path in self.paths:
                 # Test for stdin first, in case some file named '-' exist
                 if path == '-':
         try:
             for path in self.paths:
                 # Test for stdin first, in case some file named '-' exist
                 if path == '-':

+                    if self.dry_run:
+                        raise ArvPutUploadIsPending()

                     self._write_stdin(self.filename or 'stdin')
                     self._write_stdin(self.filename or 'stdin')

+                elif not os.path.exists(path):
+                     raise PathDoesNotExistError("file or directory '{}' does not exist.".format(path))

                 elif os.path.isdir(path):
                     # Use absolute paths on cache index so CWD doesn't interfere
                     # with the caching logic.
                 elif os.path.isdir(path):
                     # Use absolute paths on cache index so CWD doesn't interfere
                     # with the caching logic.

-                    prefixdir = path = os.path.abspath(path)
-                    if prefixdir != '/':
-                        prefixdir += '/'
-                    for root, dirs, files in os.walk(path):
+                    orig_path = path
+                    path = os.path.abspath(path)
+                    if orig_path[-1:] == os.sep:
+                        # When passing a directory reference with a trailing slash,
+                        # its contents should be uploaded directly to the collection's root.
+                        prefixdir = path
+                    else:
+                        # When passing a directory reference with no trailing slash,
+                        # upload the directory to the collection's root.
+                        prefixdir = os.path.dirname(path)
+                    prefixdir += os.sep
+                    for root, dirs, files in os.walk(path, followlinks=self.follow_links):

                         # Make os.walk()'s dir traversing order deterministic
                         dirs.sort()
                         files.sort()
                         # Make os.walk()'s dir traversing order deterministic
                         dirs.sort()
                         files.sort()


@@ -384,22 +465,49 @@ class ArvPutUploadJob(object):
                 else:
                     self._check_file(os.path.abspath(path),
                                      self.filename or os.path.basename(path))
                 else:
                     self._check_file(os.path.abspath(path),
                                      self.filename or os.path.basename(path))

+            # If dry-mode is on, and got up to this point, then we should notify that
+            # there aren't any file to upload.
+            if self.dry_run:
+                raise ArvPutUploadNotPending()
+            # Remove local_collection's files that don't exist locally anymore, so the
+            # bytes_written count is correct.
+            for f in self.collection_file_paths(self._local_collection,
+                                                path_prefix=""):
+                if f != 'stdin' and f != self.filename and not f in self._file_paths:
+                    self._local_collection.remove(f)

             # Update bytes_written from current local collection and
             # report initial progress.
             self._update()
             # Actual file upload
             # Update bytes_written from current local collection and
             # report initial progress.
             self._update()
             # Actual file upload

+            self._upload_started = True # Used by the update thread to start checkpointing

             self._upload_files()
             self._upload_files()

+        except (SystemExit, Exception) as e:
+            self._checkpoint_before_quit = False
+            # Log stack trace only when Ctrl-C isn't pressed (SIGINT)
+            # Note: We're expecting SystemExit instead of
+            # KeyboardInterrupt because we have a custom signal
+            # handler in place that raises SystemExit with the catched
+            # signal's code.
+            if isinstance(e, PathDoesNotExistError):
+                # We aren't interested in the traceback for this case
+                pass
+            elif not isinstance(e, SystemExit) or e.code != -2:
+                self.logger.warning("Abnormal termination:\n{}".format(
+                    traceback.format_exc()))
+            raise

         finally:
         finally:

-            # Stop the thread before doing anything else
-            self._stop_checkpointer.set()
-            self._checkpointer.join()
-            # Commit all pending blocks & one last _update()
-            self._local_collection.manifest_text()
-            self._update(final=True)
+            if not self.dry_run:
+                # Stop the thread before doing anything else
+                self._stop_checkpointer.set()
+                self._checkpointer.join()
+                if self._checkpoint_before_quit:
+                    # Commit all pending blocks & one last _update()
+                    self._local_collection.manifest_text()
+                    self._update(final=True)
+                    if save_collection:
+                        self.save_collection()

             if self.use_cache:
                 self._cache_file.close()
             if self.use_cache:
                 self._cache_file.close()

-            if save_collection:
-                self.save_collection()

 
     def save_collection(self):
         if self.update:
 
     def save_collection(self):
         if self.update:


@@ -437,7 +545,7 @@ class ArvPutUploadJob(object):
         Recursively get the total size of the collection
         """
         size = 0
         Recursively get the total size of the collection
         """
         size = 0

-        for item in collection.values():
+        for item in listvalues(collection):

             if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection):
                 size += self._collection_size(item)
             else:
             if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection):
                 size += self._collection_size(item)
             else:


@@ -449,24 +557,34 @@ class ArvPutUploadJob(object):
         Periodically called support task. File uploading is
         asynchronous so we poll status from the collection.
         """
         Periodically called support task. File uploading is
         asynchronous so we poll status from the collection.
         """

-        while not self._stop_checkpointer.wait(self._update_task_time):
+        while not self._stop_checkpointer.wait(1 if not self._upload_started else self._update_task_time):

             self._update()
 
     def _update(self, final=False):
         """
         Update cached manifest text and report progress.
         """
             self._update()
 
     def _update(self, final=False):
         """
         Update cached manifest text and report progress.
         """

-        with self._collection_lock:
-            self.bytes_written = self._collection_size(self._local_collection)
-            if self.use_cache:
-                # Update cache
-                with self._state_lock:
+        if self._upload_started:
+            with self._collection_lock:
+                self.bytes_written = self._collection_size(self._local_collection)
+                if self.use_cache:

                     if final:
                     if final:

-                        self._state['manifest'] = self._local_collection.manifest_text()
+                        manifest = self._local_collection.manifest_text()

                     else:
                         # Get the manifest text without comitting pending blocks
                     else:
                         # Get the manifest text without comitting pending blocks

-                        self._state['manifest'] = self._local_collection._get_manifest_text(".", strip=False, normalize=False, only_committed=True)
-                self._save_state()
+                        manifest = self._local_collection.manifest_text(strip=False,
+                                                                        normalize=False,
+                                                                        only_committed=True)
+                    # Update cache
+                    with self._state_lock:
+                        self._state['manifest'] = manifest
+            if self.use_cache:
+                try:
+                    self._save_state()
+                except Exception as e:
+                    self.logger.error("Unexpected error trying to save cache file: {}".format(e))
+        else:
+            self.bytes_written = self.bytes_skipped

         # Call the reporter, if any
         self.report_progress()
 
         # Call the reporter, if any
         self.report_progress()
 


@@ -475,17 +593,22 @@ class ArvPutUploadJob(object):
             self.reporter(self.bytes_written, self.bytes_expected)
 
     def _write_stdin(self, filename):
             self.reporter(self.bytes_written, self.bytes_expected)
 
     def _write_stdin(self, filename):

-        output = self._local_collection.open(filename, 'w')
+        output = self._local_collection.open(filename, 'wb')

         self._write(sys.stdin, output)
         output.close()
 
     def _check_file(self, source, filename):
         self._write(sys.stdin, output)
         output.close()
 
     def _check_file(self, source, filename):

-        """Check if this file needs to be uploaded"""
+        """
+        Check if this file needs to be uploaded
+        """
+        # Ignore symlinks when requested
+        if (not self.follow_links) and os.path.islink(source):
+            return

         resume_offset = 0
         should_upload = False
         new_file_in_cache = False
         # Record file path for updating the remote collection before exiting
         resume_offset = 0
         should_upload = False
         new_file_in_cache = False
         # Record file path for updating the remote collection before exiting

-        self._file_paths.append(filename)
+        self._file_paths.add(filename)

 
         with self._state_lock:
             # If no previous cached data on this file, store it for an eventual
 
         with self._state_lock:
             # If no previous cached data on this file, store it for an eventual


@@ -541,17 +664,17 @@ class ArvPutUploadJob(object):
 
     def _upload_files(self):
         for source, resume_offset, filename in self._files_to_upload:
 
     def _upload_files(self):
         for source, resume_offset, filename in self._files_to_upload:

-            with open(source, 'r') as source_fd:
+            with open(source, 'rb') as source_fd:

                 with self._state_lock:
                     self._state['files'][source]['mtime'] = os.path.getmtime(source)
                     self._state['files'][source]['size'] = os.path.getsize(source)
                 if resume_offset > 0:
                     # Start upload where we left off
                 with self._state_lock:
                     self._state['files'][source]['mtime'] = os.path.getmtime(source)
                     self._state['files'][source]['size'] = os.path.getsize(source)
                 if resume_offset > 0:
                     # Start upload where we left off

-                    output = self._local_collection.open(filename, 'a')
+                    output = self._local_collection.open(filename, 'ab')

                     source_fd.seek(resume_offset)
                 else:
                     # Start from scratch
                     source_fd.seek(resume_offset)
                 else:
                     # Start from scratch

-                    output = self._local_collection.open(filename, 'w')
+                    output = self._local_collection.open(filename, 'wb')

                 self._write(source_fd, output)
                 output.close(flush=False)
 
                 self._write(source_fd, output)
                 output.close(flush=False)
 


@@ -585,19 +708,21 @@ class ArvPutUploadJob(object):
         if self.use_cache:
             # Set up cache file name from input paths.
             md5 = hashlib.md5()
         if self.use_cache:
             # Set up cache file name from input paths.
             md5 = hashlib.md5()

-            md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost'))
+            md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())

             realpaths = sorted(os.path.realpath(path) for path in self.paths)
             realpaths = sorted(os.path.realpath(path) for path in self.paths)

-            md5.update('\0'.join(realpaths))
+            md5.update(b'\0'.join([p.encode() for p in realpaths]))

             if self.filename:
             if self.filename:

-                md5.update(self.filename)
+                md5.update(self.filename.encode())

             cache_filename = md5.hexdigest()
             cache_filepath = os.path.join(
                 arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'),
                 cache_filename)
             cache_filename = md5.hexdigest()
             cache_filepath = os.path.join(
                 arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'),
                 cache_filename)

-            if self.resume:
+            if self.resume and os.path.exists(cache_filepath):
+                self.logger.info("Resuming upload from cache file {}".format(cache_filepath))

                 self._cache_file = open(cache_filepath, 'a+')
             else:
                 # --no-resume means start with a empty cache file.
                 self._cache_file = open(cache_filepath, 'a+')
             else:
                 # --no-resume means start with a empty cache file.

+                self.logger.info("Creating new cache file at {}".format(cache_filepath))

                 self._cache_file = open(cache_filepath, 'w+')
             self._cache_filename = self._cache_file.name
             self._lock_file(self._cache_file)
                 self._cache_file = open(cache_filepath, 'w+')
             self._cache_filename = self._cache_file.name
             self._lock_file(self._cache_file)


@@ -614,10 +739,22 @@ class ArvPutUploadJob(object):
                     # Cache file empty, set up new cache
                     self._state = copy.deepcopy(self.EMPTY_STATE)
             else:
                     # Cache file empty, set up new cache
                     self._state = copy.deepcopy(self.EMPTY_STATE)
             else:

+                self.logger.info("No cache usage requested for this run.")

                 # No cache file, set empty state
                 self._state = copy.deepcopy(self.EMPTY_STATE)
             # Load the previous manifest so we can check if files were modified remotely.
                 # No cache file, set empty state
                 self._state = copy.deepcopy(self.EMPTY_STATE)
             # Load the previous manifest so we can check if files were modified remotely.

-            self._local_collection = arvados.collection.Collection(self._state['manifest'], replication_desired=self.replication_desired)
+            self._local_collection = arvados.collection.Collection(self._state['manifest'], replication_desired=self.replication_desired, put_threads=self.put_threads)
+
+    def collection_file_paths(self, col, path_prefix='.'):
+        """Return a list of file paths by recursively go through the entire collection `col`"""
+        file_paths = []
+        for name, item in listitems(col):
+            if isinstance(item, arvados.arvfile.ArvadosFile):
+                file_paths.append(os.path.join(path_prefix, name))
+            elif isinstance(item, arvados.collection.Subcollection):
+                new_prefix = os.path.join(path_prefix, name)
+                file_paths += self.collection_file_paths(item, path_prefix=new_prefix)
+        return file_paths

 
     def _lock_file(self, fileobj):
         try:
 
     def _lock_file(self, fileobj):
         try:


@@ -629,17 +766,20 @@ class ArvPutUploadJob(object):
         """
         Atomically save current state into cache.
         """
         """
         Atomically save current state into cache.
         """

+        with self._state_lock:
+            # We're not using copy.deepcopy() here because it's a lot slower
+            # than json.dumps(), and we're already needing JSON format to be
+            # saved on disk.
+            state = json.dumps(self._state)

         try:
         try:

-            with self._state_lock:
-                state = copy.deepcopy(self._state)
-            new_cache_fd, new_cache_name = tempfile.mkstemp(
-                dir=os.path.dirname(self._cache_filename))
-            self._lock_file(new_cache_fd)
-            new_cache = os.fdopen(new_cache_fd, 'r+')
-            json.dump(state, new_cache)
+            new_cache = tempfile.NamedTemporaryFile(
+                mode='w+',
+                dir=os.path.dirname(self._cache_filename), delete=False)
+            self._lock_file(new_cache)
+            new_cache.write(state)

             new_cache.flush()
             os.fsync(new_cache)
             new_cache.flush()
             os.fsync(new_cache)

-            os.rename(new_cache_name, self._cache_filename)
+            os.rename(new_cache.name, self._cache_filename)

         except (IOError, OSError, ResumeCacheConflict) as error:
             self.logger.error("There was a problem while saving the cache file: {}".format(error))
             try:
         except (IOError, OSError, ResumeCacheConflict) as error:
             self.logger.error("There was a problem while saving the cache file: {}".format(error))
             try:


@@ -657,7 +797,14 @@ class ArvPutUploadJob(object):
         return self._my_collection().manifest_locator()
 
     def portable_data_hash(self):
         return self._my_collection().manifest_locator()
 
     def portable_data_hash(self):

-        return self._my_collection().portable_data_hash()
+        pdh = self._my_collection().portable_data_hash()
+        m = self._my_collection().stripped_manifest().encode()
+        local_pdh = '{}+{}'.format(hashlib.md5(m).hexdigest(), len(m))
+        if pdh != local_pdh:
+            logger.warning("\n".join([
+                "arv-put: API server provided PDH differs from local manifest.",
+                "         This should not happen; showing API server version."]))
+        return pdh

 
     def manifest_text(self, stream_name=".", strip=False, normalize=False):
         return self._my_collection().manifest_text(stream_name, strip, normalize)
 
     def manifest_text(self, stream_name=".", strip=False, normalize=False):
         return self._my_collection().manifest_text(stream_name, strip, normalize)


@@ -678,7 +825,7 @@ class ArvPutUploadJob(object):
                     locators.append(loc)
                 return locators
         elif isinstance(item, arvados.collection.Collection):
                     locators.append(loc)
                 return locators
         elif isinstance(item, arvados.collection.Collection):

-            l = [self._datablocks_on_item(x) for x in item.values()]
+            l = [self._datablocks_on_item(x) for x in listvalues(item)]

             # Fast list flattener method taken from:
             # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python
             return [loc for sublist in l for loc in sublist]
             # Fast list flattener method taken from:
             # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python
             return [loc for sublist in l for loc in sublist]


@@ -693,14 +840,20 @@ class ArvPutUploadJob(object):
         return datablocks
 
 
         return datablocks
 
 

-def expected_bytes_for(pathlist):
+def expected_bytes_for(pathlist, follow_links=True):

     # Walk the given directory trees and stat files, adding up file sizes,
     # so we can display progress as percent
     bytesum = 0
     for path in pathlist:
         if os.path.isdir(path):
     # Walk the given directory trees and stat files, adding up file sizes,
     # so we can display progress as percent
     bytesum = 0
     for path in pathlist:
         if os.path.isdir(path):

-            for filename in arvados.util.listdir_recursive(path):
-                bytesum += os.path.getsize(os.path.join(path, filename))
+            for root, dirs, files in os.walk(path, followlinks=follow_links):
+                # Sum file sizes
+                for f in files:
+                    filepath = os.path.join(root, f)
+                    # Ignore symlinked files when requested
+                    if (not follow_links) and os.path.islink(filepath):
+                        continue
+                    bytesum += os.path.getsize(filepath)

         elif not os.path.isfile(path):
             return None
         else:
         elif not os.path.isfile(path):
             return None
         else:


@@ -744,6 +897,7 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
     global api_client
 
     logger = logging.getLogger('arvados.arv_put')
     global api_client
 
     logger = logging.getLogger('arvados.arv_put')

+    logger.setLevel(logging.INFO)

     args = parse_arguments(arguments)
     status = 0
     if api_client is None:
     args = parse_arguments(arguments)
     status = 0
     if api_client is None:


@@ -754,6 +908,9 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
         if args.stream or args.raw:
             logger.error("Cannot use --name with --stream or --raw")
             sys.exit(1)
         if args.stream or args.raw:
             logger.error("Cannot use --name with --stream or --raw")
             sys.exit(1)

+        elif args.update_collection:
+            logger.error("Cannot use --name with --update-collection")
+            sys.exit(1)

         collection_name = args.name
     else:
         collection_name = "Saved at {} by {}@{}".format(
         collection_name = args.name
     else:
         collection_name = "Saved at {} by {}@{}".format(


@@ -780,7 +937,11 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
     else:
         reporter = None
 
     else:
         reporter = None
 

-    bytes_expected = expected_bytes_for(args.paths)
+    # If this is used by a human, and there's at least one directory to be
+    # uploaded, the expected bytes calculation can take a moment.
+    if args.progress and any([os.path.isdir(f) for f in args.paths]):
+        logger.info("Calculating upload size, this could take some time...")
+    bytes_expected = expected_bytes_for(args.paths, follow_links=args.follow_links)

 
     try:
         writer = ArvPutUploadJob(paths = args.paths,
 
     try:
         writer = ArvPutUploadJob(paths = args.paths,


@@ -791,11 +952,14 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
                                  bytes_expected = bytes_expected,
                                  num_retries = args.retries,
                                  replication_desired = args.replication,
                                  bytes_expected = bytes_expected,
                                  num_retries = args.retries,
                                  replication_desired = args.replication,

+                                 put_threads = args.threads,

                                  name = collection_name,
                                  owner_uuid = project_uuid,
                                  ensure_unique_name = True,
                                  update_collection = args.update_collection,
                                  name = collection_name,
                                  owner_uuid = project_uuid,
                                  ensure_unique_name = True,
                                  update_collection = args.update_collection,

-                                 logger=logger)
+                                 logger=logger,
+                                 dry_run=args.dry_run,
+                                 follow_links=args.follow_links)

     except ResumeCacheConflict:
         logger.error("\n".join([
             "arv-put: Another process is already uploading this data.",
     except ResumeCacheConflict:
         logger.error("\n".join([
             "arv-put: Another process is already uploading this data.",


@@ -805,18 +969,25 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
         logger.error("\n".join([
             "arv-put: %s" % str(error)]))
         sys.exit(1)
         logger.error("\n".join([
             "arv-put: %s" % str(error)]))
         sys.exit(1)

+    except ArvPutUploadIsPending:
+        # Dry run check successful, return proper exit code.
+        sys.exit(2)
+    except ArvPutUploadNotPending:
+        # No files pending for upload
+        sys.exit(0)

 
     # Install our signal handler for each code in CAUGHT_SIGNALS, and save
     # the originals.
     orig_signal_handlers = {sigcode: signal.signal(sigcode, exit_signal_handler)
                             for sigcode in CAUGHT_SIGNALS}
 
 
     # Install our signal handler for each code in CAUGHT_SIGNALS, and save
     # the originals.
     orig_signal_handlers = {sigcode: signal.signal(sigcode, exit_signal_handler)
                             for sigcode in CAUGHT_SIGNALS}
 

-    if not args.update_collection and args.resume and writer.bytes_written > 0:
+    if not args.dry_run and not args.update_collection and args.resume and writer.bytes_written > 0:

         logger.warning("\n".join([
             "arv-put: Resuming previous upload from last checkpoint.",
             "         Use the --no-resume option to start over."]))
 
         logger.warning("\n".join([
             "arv-put: Resuming previous upload from last checkpoint.",
             "         Use the --no-resume option to start over."]))
 

-    writer.report_progress()
+    if not args.dry_run:
+        writer.report_progress()

     output = None
     try:
         writer.start(save_collection=not(args.stream or args.raw))
     output = None
     try:
         writer.start(save_collection=not(args.stream or args.raw))


@@ -824,9 +995,19 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
         logger.error("\n".join([
             "arv-put: %s" % str(error)]))
         sys.exit(1)
         logger.error("\n".join([
             "arv-put: %s" % str(error)]))
         sys.exit(1)

+    except ArvPutUploadIsPending:
+        # Dry run check successful, return proper exit code.
+        sys.exit(2)
+    except ArvPutUploadNotPending:
+        # No files pending for upload
+        sys.exit(0)
+    except PathDoesNotExistError as error:
+        logger.error("\n".join([
+            "arv-put: %s" % str(error)]))
+        sys.exit(1)

 
     if args.progress:  # Print newline to split stderr from stdout for humans.
 
     if args.progress:  # Print newline to split stderr from stdout for humans.

-        logger.error("\n")
+        logger.info("\n")

 
     if args.stream:
         if args.normalize:
 
     if args.stream:
         if args.normalize:


@@ -859,7 +1040,7 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
         if not output.endswith('\n'):
             stdout.write('\n')
 
         if not output.endswith('\n'):
             stdout.write('\n')
 

-    for sigcode, orig_handler in orig_signal_handlers.items():
+    for sigcode, orig_handler in listitems(orig_signal_handlers):

         signal.signal(sigcode, orig_handler)
 
     if status != 0:
         signal.signal(sigcode, orig_handler)
 
     if status != 0: