# Copyright 2022 Google LLC

#

# Licensed under the Apache License, Version 2.0 (the "License");

# you may not use this file except in compliance with the License.

# You may obtain a copy of the License at

#

# http://www.apache.org/licenses/LICENSE-2.0

#

# Unless required by applicable law or agreed to in writing, software

# distributed under the License is distributed on an "AS IS" BASIS,

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

# See the License for the specific language governing permissions and

# limitations under the License.

"""Concurrent media operations."""

import concurrent.futures

import io

import inspect

import os

import warnings

import pickle

import copyreg

import struct

import base64

import functools

from pathlib import Path

from google.api_core import exceptions

from google.cloud.storage import Client

from google.cloud.storage import Blob

from google.cloud.storage.blob import _get_host_name

from google.cloud.storage.blob import _quote

from google.cloud.storage.constants import _DEFAULT_TIMEOUT

from google.cloud.storage.retry import DEFAULT_RETRY

import google_crc32c

from google.cloud.storage._media.requests.upload import XMLMPUContainer

from google.cloud.storage._media.requests.upload import XMLMPUPart

from google.cloud.storage.exceptions import DataCorruption, InvalidPathError

TM_DEFAULT_CHUNK_SIZE = 32 * 1024 * 1024

DEFAULT_MAX_WORKERS = 8

MAX_CRC32C_ZERO_ARRAY_SIZE = 4 * 1024 * 1024

METADATA_HEADER_TRANSLATION = {

"cacheControl": "Cache-Control",

"contentDisposition": "Content-Disposition",

"contentEncoding": "Content-Encoding",

"contentLanguage": "Content-Language",

"customTime": "x-goog-custom-time",

"storageClass": "x-goog-storage-class",

}

# Constants to be passed in as `worker_type`.

PROCESS = "process"

THREAD = "thread"

DOWNLOAD_CRC32C_MISMATCH_TEMPLATE = """\

Checksum mismatch while downloading:

{}

The object metadata indicated a crc32c checksum of:

{}

but the actual crc32c checksum of the downloaded contents was:

{}

"""

_cached_clients = {}

def _deprecate_threads_param(func):

@functools.wraps(func)

def convert_threads_or_raise(*args, **kwargs):

binding = inspect.signature(func).bind(*args, **kwargs)

threads = binding.arguments.get("threads")

if threads:

worker_type = binding.arguments.get("worker_type")

max_workers = binding.arguments.get("max_workers")

if worker_type or max_workers: # Parameter conflict

raise ValueError(

"The `threads` parameter is deprecated and conflicts with its replacement parameters, `worker_type` and `max_workers`."

)

# No conflict, so issue a warning and set worker_type and max_workers.

warnings.warn(

"The `threads` parameter is deprecated. Please use `worker_type` and `max_workers` parameters instead."

)

args = binding.args

kwargs = binding.kwargs

kwargs["worker_type"] = THREAD

kwargs["max_workers"] = threads

return func(*args, **kwargs)

else:

return func(*args, **kwargs)

return convert_threads_or_raise

@_deprecate_threads_param

def upload_many(

file_blob_pairs,

skip_if_exists=False,

upload_kwargs=None,

threads=None,

deadline=None,

raise_exception=False,

worker_type=PROCESS,

max_workers=DEFAULT_MAX_WORKERS,

):

"""Upload many files concurrently via a worker pool.

:type file_blob_pairs: List(Tuple(IOBase or str, 'google.cloud.storage.blob.Blob'))

:param file_blob_pairs:

A list of tuples of a file or filename and a blob. Each file will be

uploaded to the corresponding blob by using APIs identical to

`blob.upload_from_file()` or `blob.upload_from_filename()` as

appropriate.

File handlers are only supported if worker_type is set to THREAD.

If worker_type is set to PROCESS, please use filenames only.

:type skip_if_exists: bool

:param skip_if_exists:

If True, blobs that already have a live version will not be overwritten.

This is accomplished by setting `if_generation_match = 0` on uploads.

Uploads so skipped will result in a 412 Precondition Failed response

code, which will be included in the return value but not raised

as an exception regardless of the value of raise_exception.

:type upload_kwargs: dict

:param upload_kwargs:

A dictionary of keyword arguments to pass to the upload method. Refer

to the documentation for `blob.upload_from_file()` or

`blob.upload_from_filename()` for more information. The dict is directly

passed into the upload methods and is not validated by this function.

:type threads: int

:param threads:

***DEPRECATED*** Sets `worker_type` to THREAD and `max_workers` to the

number specified. If `worker_type` or `max_workers` are set explicitly,

this parameter should be set to None. Please use `worker_type` and

`max_workers` instead of this parameter.

:type deadline: int

:param deadline:

The number of seconds to wait for all threads to resolve. If the

deadline is reached, all threads will be terminated regardless of their

progress and `concurrent.futures.TimeoutError` will be raised. This can

be left as the default of `None` (no deadline) for most use cases.

:type raise_exception: bool

:param raise_exception:

If True, instead of adding exceptions to the list of return values,

instead they will be raised. Note that encountering an exception on one

operation will not prevent other operations from starting. Exceptions

are only processed and potentially raised after all operations are

complete in success or failure.

If skip_if_exists is True, 412 Precondition Failed responses are

considered part of normal operation and are not raised as an exception.

:type worker_type: str

:param worker_type:

The worker type to use; one of `google.cloud.storage.transfer_manager.PROCESS`

or `google.cloud.storage.transfer_manager.THREAD`.

Although the exact performance impact depends on the use case, in most

situations the PROCESS worker type will use more system resources (both

memory and CPU) and result in faster operations than THREAD workers.

Because the subprocesses of the PROCESS worker type can't access memory

from the main process, Client objects have to be serialized and then

recreated in each subprocess. The serialization of the Client object

for use in subprocesses is an approximation and may not capture every

detail of the Client object, especially if the Client was modified after

its initial creation or if `Client._http` was modified in any way.

THREAD worker types are observed to be relatively efficient for

operations with many small files, but not for operations with large

files. PROCESS workers are recommended for large file operations.

PROCESS workers do not support writing to file handlers. Please refer

to files by filename only when using PROCESS workers.

:type max_workers: int

:param max_workers:

The maximum number of workers to create to handle the workload.

With PROCESS workers, a larger number of workers will consume more

system resources (memory and CPU) at once.

How many workers is optimal depends heavily on the specific use case,

and the default is a conservative number that should work okay in most

cases without consuming excessive resources.

:raises: :exc:`concurrent.futures.TimeoutError` if deadline is exceeded.

:rtype: list

:returns: A list of results corresponding to, in order, each item in the

input list. If an exception was received, it will be the result

for that operation. Otherwise, the return value from the successful

upload method is used (which will be None).

"""

if upload_kwargs is None:

upload_kwargs = {}

if skip_if_exists:

upload_kwargs = upload_kwargs.copy()

upload_kwargs["if_generation_match"] = 0

upload_kwargs["command"] = "tm.upload_many"

pool_class, needs_pickling = _get_pool_class_and_requirements(worker_type)

with pool_class(max_workers=max_workers) as executor:

futures = []

for path_or_file, blob in file_blob_pairs:

# File objects are only supported by the THREAD worker because they can't

# be pickled.

if needs_pickling and not isinstance(path_or_file, str):

raise ValueError(

"Passing in a file object is only supported by the THREAD worker type. Please either select THREAD workers, or pass in filenames only."

)

futures.append(

executor.submit(

_call_method_on_maybe_pickled_blob,

_pickle_client(blob) if needs_pickling else blob,

(

"_handle_filename_and_upload"

if isinstance(path_or_file, str)

else "_prep_and_do_upload"

),

path_or_file,

**upload_kwargs,

)

)

concurrent.futures.wait(

futures, timeout=deadline, return_when=concurrent.futures.ALL_COMPLETED

)

results = []

for future in futures:

exp = future.exception()

# If raise_exception is False, don't call future.result()

if exp and not raise_exception:

results.append(exp)

# If skip_if_exists and the exception is PreconditionFailed, do same.

elif exp and skip_if_exists and isinstance(exp, exceptions.PreconditionFailed):

results.append(exp)

# Get the real result. If there was an exception not handled above,

# this will raise it.

else:

results.append(future.result())

return results

def _resolve_path(target_dir, blob_path):

if os.name == "nt" and ":" in blob_path:

raise InvalidPathError(f"{blob_path} cannot be downloaded into {target_dir}")

target_dir = Path(target_dir)

blob_path = Path(blob_path)

# blob_path.anchor will be '/' if `blob_path` is full path else it'll empty.

# This is useful to concatnate target_dir = /local/target , and blob_path =

# /usr/local/mybin into /local/target/usr/local/mybin

concatenated_path = target_dir / blob_path.relative_to(blob_path.anchor)

return concatenated_path.resolve()

@_deprecate_threads_param

def download_many(

blob_file_pairs,

download_kwargs=None,

threads=None,

deadline=None,

raise_exception=False,

worker_type=PROCESS,

max_workers=DEFAULT_MAX_WORKERS,

*,

skip_if_exists=False,

):

"""Download many blobs concurrently via a worker pool.

:type blob_file_pairs: List(Tuple('google.cloud.storage.blob.Blob', IOBase or str))

:param blob_file_pairs:

A list of tuples of blob and a file or filename. Each blob will be downloaded to the corresponding blob by using APIs identical to blob.download_to_file() or blob.download_to_filename() as appropriate.

Note that blob.download_to_filename() does not delete the destination file if the download fails.

File handlers are only supported if worker_type is set to THREAD.

If worker_type is set to PROCESS, please use filenames only.

:type download_kwargs: dict

:param download_kwargs:

A dictionary of keyword arguments to pass to the download method. Refer

to the documentation for `blob.download_to_file()` or

`blob.download_to_filename()` for more information. The dict is directly

passed into the download methods and is not validated by this function.

:type threads: int

:param threads:

***DEPRECATED*** Sets `worker_type` to THREAD and `max_workers` to the

number specified. If `worker_type` or `max_workers` are set explicitly,

this parameter should be set to None. Please use `worker_type` and

`max_workers` instead of this parameter.

:type deadline: int

:param deadline:

The number of seconds to wait for all threads to resolve. If the

deadline is reached, all threads will be terminated regardless of their

progress and `concurrent.futures.TimeoutError` will be raised. This can

be left as the default of `None` (no deadline) for most use cases.

:type raise_exception: bool

:param raise_exception:

If True, instead of adding exceptions to the list of return values,

instead they will be raised. Note that encountering an exception on one

operation will not prevent other operations from starting. Exceptions

are only processed and potentially raised after all operations are

complete in success or failure.

:type worker_type: str

:param worker_type:

The worker type to use; one of `google.cloud.storage.transfer_manager.PROCESS`

or `google.cloud.storage.transfer_manager.THREAD`.

Although the exact performance impact depends on the use case, in most

situations the PROCESS worker type will use more system resources (both

memory and CPU) and result in faster operations than THREAD workers.

Because the subprocesses of the PROCESS worker type can't access memory

from the main process, Client objects have to be serialized and then

recreated in each subprocess. The serialization of the Client object

for use in subprocesses is an approximation and may not capture every

detail of the Client object, especially if the Client was modified after

its initial creation or if `Client._http` was modified in any way.

THREAD worker types are observed to be relatively efficient for

operations with many small files, but not for operations with large

files. PROCESS workers are recommended for large file operations.

PROCESS workers do not support writing to file handlers. Please refer

to files by filename only when using PROCESS workers.

:type max_workers: int

:param max_workers:

The maximum number of workers to create to handle the workload.

With PROCESS workers, a larger number of workers will consume more

system resources (memory and CPU) at once.

How many workers is optimal depends heavily on the specific use case,

and the default is a conservative number that should work okay in most

cases without consuming excessive resources.

:type skip_if_exists: bool

:param skip_if_exists:

Before downloading each blob, check if the file for the filename exists;

if it does, skip that blob.

:raises: :exc:`concurrent.futures.TimeoutError` if deadline is exceeded.

:rtype: list

:returns: A list of results corresponding to, in order, each item in the

input list. If an exception was received, it will be the result

for that operation. Otherwise, the return value from the successful

download method is used (which will be None).

"""

if download_kwargs is None:

download_kwargs = {}

download_kwargs["command"] = "tm.download_many"

pool_class, needs_pickling = _get_pool_class_and_requirements(worker_type)

with pool_class(max_workers=max_workers) as executor:

futures = []

for blob, path_or_file in blob_file_pairs:

# File objects are only supported by the THREAD worker because they can't

# be pickled.

if needs_pickling and not isinstance(path_or_file, str):

raise ValueError(

"Passing in a file object is only supported by the THREAD worker type. Please either select THREAD workers, or pass in filenames only."

)

if skip_if_exists and isinstance(path_or_file, str):

if os.path.isfile(path_or_file):

continue

futures.append(

executor.submit(

_call_method_on_maybe_pickled_blob,

_pickle_client(blob) if needs_pickling else blob,

(

"_handle_filename_and_download"

if isinstance(path_or_file, str)

else "_prep_and_do_download"

),

path_or_file,

**download_kwargs,

)

)

concurrent.futures.wait(

futures, timeout=deadline, return_when=concurrent.futures.ALL_COMPLETED

)

results = []

for future in futures:

# If raise_exception is False, don't call future.result()

if not raise_exception:

exp = future.exception()

if exp:

results.append(exp)

continue

# Get the real result. If there was an exception, this will raise it.

results.append(future.result())

return results

@_deprecate_threads_param

def upload_many_from_filenames(

bucket,

filenames,

source_directory="",

blob_name_prefix="",

skip_if_exists=False,

blob_constructor_kwargs=None,

upload_kwargs=None,

threads=None,

deadline=None,

raise_exception=False,

worker_type=PROCESS,

max_workers=DEFAULT_MAX_WORKERS,

*,

additional_blob_attributes=None,

):

"""Upload many files concurrently by their filenames.

The destination blobs are automatically created, with blob names based on

the source filenames and the blob_name_prefix.

For example, if the `filenames` include "images/icon.jpg",

`source_directory` is "/home/myuser/", and `blob_name_prefix` is "myfiles/",

then the file at "/home/myuser/images/icon.jpg" will be uploaded to a blob

named "myfiles/images/icon.jpg".

:type bucket: :class:`google.cloud.storage.bucket.Bucket`

:param bucket:

The bucket which will contain the uploaded blobs.

:type filenames: list(str)

:param filenames:

A list of filenames to be uploaded. This may include part of the path.

The file will be accessed at the full path of `source_directory` +

`filename`.

:type source_directory: str

:param source_directory:

A string that will be prepended (with `os.path.join()`) to each filename

in the input list, in order to find the source file for each blob.

Unlike the filename itself, the source_directory does not affect the

name of the uploaded blob.

For instance, if the source_directory is "/tmp/img/" and a filename is

"0001.jpg", with an empty blob_name_prefix, then the file uploaded will

be "/tmp/img/0001.jpg" and the destination blob will be "0001.jpg".

This parameter can be an empty string.

Note that this parameter allows directory traversal (e.g. "/", "../")

and is not intended for unsanitized end user input.

:type blob_name_prefix: str

:param blob_name_prefix:

A string that will be prepended to each filename in the input list, in

order to determine the name of the destination blob. Unlike the filename

itself, the prefix string does not affect the location the library will

look for the source data on the local filesystem.

For instance, if the source_directory is "/tmp/img/", the

blob_name_prefix is "myuser/mystuff-" and a filename is "0001.jpg" then

the file uploaded will be "/tmp/img/0001.jpg" and the destination blob

will be "myuser/mystuff-0001.jpg".

The blob_name_prefix can be blank (an empty string).

:type skip_if_exists: bool

:param skip_if_exists:

If True, blobs that already have a live version will not be overwritten.

This is accomplished by setting `if_generation_match = 0` on uploads.

Uploads so skipped will result in a 412 Precondition Failed response

code, which will be included in the return value, but not raised

as an exception regardless of the value of raise_exception.

:type blob_constructor_kwargs: dict

:param blob_constructor_kwargs:

A dictionary of keyword arguments to pass to the blob constructor. Refer

to the documentation for `blob.Blob()` for more information. The dict is

directly passed into the constructor and is not validated by this

function. `name` and `bucket` keyword arguments are reserved by this

function and will result in an error if passed in here.

:type upload_kwargs: dict

:param upload_kwargs:

A dictionary of keyword arguments to pass to the upload method. Refer

to the documentation for `blob.upload_from_file()` or

`blob.upload_from_filename()` for more information. The dict is directly

passed into the upload methods and is not validated by this function.

:type threads: int

:param threads:

***DEPRECATED*** Sets `worker_type` to THREAD and `max_workers` to the

number specified. If `worker_type` or `max_workers` are set explicitly,

this parameter should be set to None. Please use `worker_type` and

`max_workers` instead of this parameter.

:type deadline: int

:param deadline:

The number of seconds to wait for all threads to resolve. If the

deadline is reached, all threads will be terminated regardless of their

progress and `concurrent.futures.TimeoutError` will be raised. This can

be left as the default of `None` (no deadline) for most use cases.

:type raise_exception: bool

:param raise_exception:

If True, instead of adding exceptions to the list of return values,

instead they will be raised. Note that encountering an exception on one

operation will not prevent other operations from starting. Exceptions

are only processed and potentially raised after all operations are

complete in success or failure.

If skip_if_exists is True, 412 Precondition Failed responses are

considered part of normal operation and are not raised as an exception.

:type worker_type: str

:param worker_type:

The worker type to use; one of `google.cloud.storage.transfer_manager.PROCESS`

or `google.cloud.storage.transfer_manager.THREAD`.

Although the exact performance impact depends on the use case, in most

situations the PROCESS worker type will use more system resources (both

memory and CPU) and result in faster operations than THREAD workers.

Because the subprocesses of the PROCESS worker type can't access memory

from the main process, Client objects have to be serialized and then

recreated in each subprocess. The serialization of the Client object

for use in subprocesses is an approximation and may not capture every

detail of the Client object, especially if the Client was modified after

its initial creation or if `Client._http` was modified in any way.

THREAD worker types are observed to be relatively efficient for

operations with many small files, but not for operations with large

files. PROCESS workers are recommended for large file operations.

:type max_workers: int

:param max_workers:

The maximum number of workers to create to handle the workload.

With PROCESS workers, a larger number of workers will consume more

system resources (memory and CPU) at once.

How many workers is optimal depends heavily on the specific use case,

and the default is a conservative number that should work okay in most

cases without consuming excessive resources.

:type additional_blob_attributes: dict

:param additional_blob_attributes:

A dictionary of blob attribute names and values. This allows the

configuration of blobs beyond what is possible with

blob_constructor_kwargs. For instance, {"cache_control": "no-cache"}

would set the cache_control attribute of each blob to "no-cache".

As with blob_constructor_kwargs, this affects the creation of every

blob identically. To fine-tune each blob individually, use `upload_many`

and create the blobs as desired before passing them in.

:raises: :exc:`concurrent.futures.TimeoutError` if deadline is exceeded.

:rtype: list

:returns: A list of results corresponding to, in order, each item in the

input list. If an exception was received, it will be the result

for that operation. Otherwise, the return value from the successful

upload method is used (which will be None).

"""

if blob_constructor_kwargs is None:

blob_constructor_kwargs = {}

if additional_blob_attributes is None:

additional_blob_attributes = {}

file_blob_pairs = []

for filename in filenames:

path = os.path.join(source_directory, filename)

blob_name = blob_name_prefix + filename

blob = bucket.blob(blob_name, **blob_constructor_kwargs)

for prop, value in additional_blob_attributes.items():

setattr(blob, prop, value)

file_blob_pairs.append((path, blob))

return upload_many(

file_blob_pairs,

skip_if_exists=skip_if_exists,

upload_kwargs=upload_kwargs,

deadline=deadline,

raise_exception=raise_exception,

worker_type=worker_type,

max_workers=max_workers,

)

@_deprecate_threads_param

def download_many_to_path(

bucket,

blob_names,

destination_directory="",

blob_name_prefix="",

download_kwargs=None,

threads=None,

deadline=None,

create_directories=True,

raise_exception=False,

worker_type=PROCESS,

max_workers=DEFAULT_MAX_WORKERS,

*,

skip_if_exists=False,

):

"""Download many files concurrently by their blob names.

The destination files are automatically created, with paths based on the

source `blob_names` and the `destination_directory`.

The destination files are not automatically deleted if their downloads fail,

so please check the return value of this function for any exceptions, or

enable `raise_exception=True`, and process the files accordingly.

For example, if the `blob_names` include "icon.jpg", `destination_directory`

is "/home/myuser/", and `blob_name_prefix` is "images/", then the blob named

"images/icon.jpg" will be downloaded to a file named

"/home/myuser/icon.jpg".

Note1: if the path after combining `blob_name` and `destination_directory`

resolves outside `destination_directory` a warning will be issued and the

that particular blob will NOT be downloaded. This may happen in scenarios

where `blob_name` contains "../"

For example,

consider `destination_directory` is "downloads/gcs_blobs" and

`blob_name` is '../hello.blob'. This blob will not be downloaded

because the final resolved path would be "downloads/hello.blob"

To give further examples, the following blobs will not be downloaded because

it "escapes" the "destination_directory"

"../../local/target", # skips download

"../escape.txt", # skips download

"go/four/levels/deep/../../../../../somefile1", # skips download

"go/four/levels/deep/../some_dir/../../../../../invalid/path1" # skips download

however the following blobs will be downloaded because the final resolved

destination_directory is still child of given destination_directory

"data/../sibling.txt",

"dir/./file.txt",

"go/four/levels/deep/../somefile2",

"go/four/levels/deep/../some_dir/valid/path1",

"go/four/levels/deep/../some_dir/../../../../valid/path2",

It is adviced to use other APIs such as `transfer_manager.download_many` or

`Blob.download_to_filename` or `Blob.download_to_file` to download such blobs.

Note2:

The resolved download_directory will always be relative to user provided

`destination_directory`. For example,

a `blob_name` "/etc/passwd" will be downloaded into

"destination_directory/etc/passwd" instead of "/etc/passwd"

Similarly,

"/tmp/my_fav_blob" downloads to "destination_directory/tmp/my_fav_blob"

:type bucket: :class:`google.cloud.storage.bucket.Bucket`

:param bucket:

The bucket which contains the blobs to be downloaded

:type blob_names: list(str)

:param blob_names:

A list of blobs to be downloaded. The blob name in this string will be

used to determine the destination file path as well.

The full name to the blob must be blob_name_prefix + blob_name. The

blob_name is separate from the blob_name_prefix because the blob_name

will also determine the name of the destination blob. Any shared part of

the blob names that need not be part of the destination path should be

included in the blob_name_prefix.

:type destination_directory: str

:param destination_directory:

A string that will be prepended to each blob_name in the input list, in

order to determine the destination path for that blob.

For instance, if the destination_directory string is "/tmp/img" and a

blob_name is "0001.jpg", with an empty blob_name_prefix, then the source

blob "0001.jpg" will be downloaded to destination "/tmp/img/0001.jpg" .

This parameter can be an empty string.

Note directory traversal may be possible as long as the final

(e.g. "/", "../") resolved path is inside "destination_directory".

See examples above.

:type blob_name_prefix: str

:param blob_name_prefix:

A string that will be prepended to each blob_name in the input list, in

order to determine the name of the source blob. Unlike the blob_name

itself, the prefix string does not affect the destination path on the

local filesystem. For instance, if the destination_directory is

"/tmp/img/", the blob_name_prefix is "myuser/mystuff-" and a blob_name

is "0001.jpg" then the source blob "myuser/mystuff-0001.jpg" will be

downloaded to "/tmp/img/0001.jpg". The blob_name_prefix can be blank

(an empty string).

:type download_kwargs: dict

:param download_kwargs:

A dictionary of keyword arguments to pass to the download method. Refer

to the documentation for `blob.download_to_file()` or

`blob.download_to_filename()` for more information. The dict is directly

passed into the download methods and is not validated by this function.

:type threads: int

:param threads:

***DEPRECATED*** Sets `worker_type` to THREAD and `max_workers` to the

number specified. If `worker_type` or `max_workers` are set explicitly,

this parameter should be set to None. Please use `worker_type` and

`max_workers` instead of this parameter.

:type deadline: int

:param deadline:

The number of seconds to wait for all threads to resolve. If the

deadline is reached, all threads will be terminated regardless of their

progress and `concurrent.futures.TimeoutError` will be raised. This can

be left as the default of `None` (no deadline) for most use cases.

:type create_directories: bool

:param create_directories:

If True, recursively create any directories that do not exist. For

instance, if downloading object "images/img001.png", create the

directory "images" before downloading.

:type raise_exception: bool

:param raise_exception:

If True, instead of adding exceptions to the list of return values,

instead they will be raised. Note that encountering an exception on one

operation will not prevent other operations from starting. Exceptions

are only processed and potentially raised after all operations are

complete in success or failure. If skip_if_exists is True, 412

Precondition Failed responses are considered part of normal operation

and are not raised as an exception.

:type worker_type: str

:param worker_type:

The worker type to use; one of `google.cloud.storage.transfer_manager.PROCESS`

or `google.cloud.storage.transfer_manager.THREAD`.

Although the exact performance impact depends on the use case, in most

situations the PROCESS worker type will use more system resources (both

memory and CPU) and result in faster operations than THREAD workers.

Because the subprocesses of the PROCESS worker type can't access memory

from the main process, Client objects have to be serialized and then

recreated in each subprocess. The serialization of the Client object

for use in subprocesses is an approximation and may not capture every

detail of the Client object, especially if the Client was modified after

its initial creation or if `Client._http` was modified in any way.

THREAD worker types are observed to be relatively efficient for

operations with many small files, but not for operations with large

files. PROCESS workers are recommended for large file operations.

:type max_workers: int

:param max_workers:

The maximum number of workers to create to handle the workload.

With PROCESS workers, a larger number of workers will consume more

system resources (memory and CPU) at once.

How many workers is optimal depends heavily on the specific use case,

and the default is a conservative number that should work okay in most

cases without consuming excessive resources.

:type skip_if_exists: bool

:param skip_if_exists:

Before downloading each blob, check if the file for the filename exists;

if it does, skip that blob. This only works for filenames.

:raises: :exc:`concurrent.futures.TimeoutError` if deadline is exceeded.

:rtype: List[None|Exception|UserWarning]

:returns: A list of results corresponding to, in order, each item in the

input list. If an exception was received or a download was skipped

(e.g., due to existing file or path traversal), it will be the result

for that operation (as an Exception or UserWarning, respectively).

Otherwise, the result will be None for a successful download.

"""

results = [None] * len(blob_names)

blob_file_pairs = []

indices_to_process = []

for i, blob_name in enumerate(blob_names):

full_blob_name = blob_name_prefix + blob_name

try:

resolved_path = _resolve_path(destination_directory, blob_name)

except InvalidPathError as e:

msg = f"The blob {blob_name} will **NOT** be downloaded. {e}"

warnings.warn(msg)

results[i] = UserWarning(msg)

continue

if not resolved_path.parent.is_relative_to(

Path(destination_directory).resolve()

):

msg = (

f"The blob {blob_name} will **NOT** be downloaded. "

f"The resolved destination_directory - {resolved_path.parent} - is either invalid or "

f"escapes user provided {Path(destination_directory).resolve()} . Please download this file separately using `download_to_filename`"

)

warnings.warn(msg)

results[i] = UserWarning(msg)

continue

resolved_path = str(resolved_path)

if skip_if_exists and os.path.isfile(resolved_path):

msg = f"The blob {blob_name} is skipped because destination file already exists"

results[i] = UserWarning(msg)

continue

if create_directories:

directory, _ = os.path.split(resolved_path)

os.makedirs(directory, exist_ok=True)

blob_file_pairs.append((bucket.blob(full_blob_name), resolved_path))

indices_to_process.append(i)

many_results = download_many(

blob_file_pairs,

download_kwargs=download_kwargs,

deadline=deadline,

raise_exception=raise_exception,

worker_type=worker_type,

max_workers=max_workers,

skip_if_exists=False, # skip_if_exists is handled in the loop above

)

for meta_index, result in zip(indices_to_process, many_results):

results[meta_index] = result

return results

def download_chunks_concurrently(

blob,

filename,

chunk_size=TM_DEFAULT_CHUNK_SIZE,

download_kwargs=None,

deadline=None,

worker_type=PROCESS,

max_workers=DEFAULT_MAX_WORKERS,

*,

crc32c_checksum=True,

):

"""Download a single file in chunks, concurrently.

In some environments, using this feature with mutiple processes will result

in faster downloads of large files.

Using this feature with multiple threads is unlikely to improve download

performance under normal circumstances due to Python interpreter threading

behavior. The default is therefore to use processes instead of threads.

:type blob: :class:`google.cloud.storage.blob.Blob`

:param blob:

The blob to be downloaded.

:type filename: str

:param filename:

The destination filename or path.

:type chunk_size: int

:param chunk_size:

The size in bytes of each chunk to send. The optimal chunk size for

maximum throughput may vary depending on the exact network environment

and size of the blob.

:type download_kwargs: dict

:param download_kwargs:

A dictionary of keyword arguments to pass to the download method. Refer

to the documentation for `blob.download_to_file()` or

`blob.download_to_filename()` for more information. The dict is directly

passed into the download methods and is not validated by this function.

Keyword arguments "start" and "end" which are not supported and will

cause a ValueError if present. The key "checksum" is also not supported

in `download_kwargs`, but see the argument `crc32c_checksum` (which does

not go in `download_kwargs`) below.

:type deadline: int

:param deadline:

The number of seconds to wait for all threads to resolve. If the

deadline is reached, all threads will be terminated regardless of their

progress and `concurrent.futures.TimeoutError` will be raised. This can

be left as the default of `None` (no deadline) for most use cases.

:type worker_type: str

:param worker_type:

The worker type to use; one of `google.cloud.storage.transfer_manager.PROCESS`

or `google.cloud.storage.transfer_manager.THREAD`.

Although the exact performance impact depends on the use case, in most

situations the PROCESS worker type will use more system resources (both

memory and CPU) and result in faster operations than THREAD workers.

Because the subprocesses of the PROCESS worker type can't access memory

from the main process, Client objects have to be serialized and then

recreated in each subprocess. The serialization of the Client object

for use in subprocesses is an approximation and may not capture every

detail of the Client object, especially if the Client was modified after

its initial creation or if `Client._http` was modified in any way.

THREAD worker types are observed to be relatively efficient for

operations with many small files, but not for operations with large

files. PROCESS workers are recommended for large file operations.

:type max_workers: int

:param max_workers:

The maximum number of workers to create to handle the workload.

With PROCESS workers, a larger number of workers will consume more

system resources (memory and CPU) at once.

How many workers is optimal depends heavily on the specific use case,

and the default is a conservative number that should work okay in most

cases without consuming excessive resources.

:type crc32c_checksum: bool

:param crc32c_checksum:

Whether to compute a checksum for the resulting object, using the crc32c

algorithm. As the checksums for each chunk must be combined using a

feature of crc32c that is not available for md5, md5 is not supported.

:raises:

:exc:`concurrent.futures.TimeoutError`

if deadline is exceeded.

:exc:`google.cloud.storage._media.common.DataCorruption`

if the download's checksum doesn't agree with server-computed

checksum. The `google.cloud.storage._media` exception is used here for

consistency with other download methods despite the exception

originating elsewhere.

"""

client = blob.client

if download_kwargs is None:

download_kwargs = {}

if "start" in download_kwargs or "end" in download_kwargs:

raise ValueError(

"Download arguments 'start' and 'end' are not supported by download_chunks_concurrently."

)

if "checksum" in download_kwargs:

raise ValueError(

"'checksum' is in download_kwargs, but is not supported because sliced downloads have a different checksum mechanism from regular downloads. Use the 'crc32c_checksum' argument on download_chunks_concurrently instead."

)

download_kwargs = download_kwargs.copy()

download_kwargs["checksum"] = None

download_kwargs["command"] = "tm.download_sharded"

# We must know the size and the generation of the blob.

if not blob.size or not blob.generation:

blob.reload()

pool_class, needs_pickling = _get_pool_class_and_requirements(worker_type)

# Pickle the blob ahead of time (just once, not once per chunk) if needed.

maybe_pickled_blob = _pickle_client(blob) if needs_pickling else blob

futures = []

# Create and/or truncate the destination file to prepare for sparse writing.

with open(filename, "wb") as _:

pass

with pool_class(max_workers=max_workers) as executor: