gs-wrap
wraps Google Cloud Storage API
for multi-threaded data manipulation including copying, reading, writing and
hashing.
Originally, we used our gsutilwrap,
a thin wrapper around gsutil
command-line interface, to simplify
the deployment and backup tasks related to Google Cloud Storage.
However, gsutilwrap
was prohibitively slow at copying many objects into
different destinations.
Therefore we developed gs-wrap
to accelerate these operations while keeping
it equally fast or faster than gsutilwrap
at other operations.
While the google-cloud-storage
library provided by Google offers sophisticated features and good performance,
its use cases and behavior differ from gsutil
.
Since we wanted the simplicity and usage patterns of gsutil
, we created
gs-wrap
, which wraps google-cloud-storage
in its core and with its
interface set to behave like gsutil
.
gs-wrap
is not the first Python library wrapping Google Cloud Storage API.
cloud-storage-client
takes a similar approach and aims to manage both Amazon's S3 and Google Cloud
Storage. Parts of it are also based on google-cloud-storage
, however the
library's behaviour differs from gsutil
which made it hard to use as an
in-place replacement for gsutilwrap
. Additionally, the library did not
offer all needed operations, for example copying to many destinations, reading,
writing and hashing.
The main strength of gs-wrap
is the ability to copy many objects from many
different paths to multiple destinations, while still mimicking gsutil
interface. A direct comparison of performance between gs-wrap
and
gsutilwrap
can be found in the section Benchmarks.
You need to create a Google Cloud Storage bucket to use this client library. Follow along with the official Google Cloud Storage documentation to learn how to create a bucket.
First a client for interacting with the Google Cloud Storage API needs to be
created. This one uses internally the Storage Client
from google-cloud-storage
.
One parameter can be passed to the client:
The Google Cloud Storage project which the client acts on behalf of. It will be passed when creating the internal client. If not passed, falls back to the default inferred from the locally authenticated Google Cloud SDK environment. Each project needs a separate client. Operations between two different projects are not supported.
import gswrap
client = gswrap.Client() # project is optional
Warning
Wildcards (*, **, ?, [chars], [char range]) are not supported by
Google Cloud Storage API and neither by gs-wrap
at the moment
[2019-01-16]. Reasons are that the gsutil
with wildcards can hardly be
equivalently reconstructed and that the toplevel search is extremely
inefficient. More information about gsutil
wildcards can be found here:
https://cloud.google.com/storage/docs/gsutil/addlhelp/WildcardNames
client.ls(gcs_url="gs://your-bucket/your-dir", recursive=False)
# gs://your-bucket/your-dir/your-subdir1/
# gs://your-bucket/your-dir/your-subdir2/
# gs://your-bucket/your-dir/file1
client.ls(gcs_url="gs://your-bucket/your-dir", recursive=True)
# gs://your-bucket/your-dir/your-subdir1/file1
# gs://your-bucket/your-dir/your-subdir1/file2
# gs://your-bucket/your-dir/your-subdir2/file1
# gs://your-bucket/your-dir/file1
If both the source and destination URL are cloud URLs from the same provider,
gsutil
copies data "in the cloud" (i.e. without downloading to and
uploading from the machine where you run gs-wrap
).
Note
client.cp() runs single-threaded by default. When multi-threading is activated, the maximum number of workers is the number of processors on the machine, multiplied by 5. This is the multi-threading default of the ThreadPoolExecuter from the concurrent.futures library.
# your-bucket before:
# gs://your-bucket/file1
client.cp(src="gs://your-bucket/file1",
dst="gs://your-bucket/your-dir/",
recursive=True)
# your-bucket after:
# gs://your-bucket/file1
# gs://your-bucket/your-dir/file1
# your-backup-bucket before:
# "empty"
client.cp(src="gs://your-bucket/file1",
dst="gs://your-backup-bucket/backup-file1",
recursive=False)
# your-backup-bucket after:
# gs://your-backup-bucket/backup-file1
# your-bucket before:
# "empty"
client.cp(src="gs://your-bucket/some-dir/",
dst="gs://your-bucket/another-dir/", recursive=False)
# google.api_core.exceptions.GoogleAPIError: No URLs matched
# your-bucket before:
# gs://your-bucket/some-dir/file1
# gs://your-bucket/some-dir/dir1/file11
# Destination URL without slash
client.cp(src="gs://your-bucket/some-dir/",
dst="gs://your-bucket/another-dir", recursive=True)
# your-bucket after:
# gs://your-bucket/another-dir/file1
# gs://your-bucket/another-dir/dir1/file11
# Destination URL with slash
client.cp(src="gs://your-bucket/some-dir/",
dst="gs://your-bucket/another-dir/", recursive=True)
# your-bucket after:
# gs://your-bucket/another-dir/some-dir/file1
# gs://your-bucket/another-dir/some-dir/dir1/file11
# Choose to copy multi-threaded. (default=False)
client.cp(src="gs://your-bucket/some-dir/",
dst="gs://your-bucket/another-dir", recursive=True, multithreaded=True)
# your-bucket after:
# gs://your-bucket/another-dir/file1
# gs://your-bucket/another-dir/dir1/file11
Note
recursive causes directories, buckets, and bucket subdirectories to be
copied recursively. If you upload from local disk to Google Cloud Storage
and set recursive to False
, gs-wrap
will raise an exception and inform you that no URL matched.
This mimicks the behaviour of gsutil
when no wildcards are used.
# Your local directory:
# /home/user/storage/file1
# /home/user/storage/file2
# your-bucket before:
# "empty"
client.cp(src="/home/user/storage/",
dst="gs://your-bucket/local/",
recursive=True)
# your-bucket after:
# gs://your-bucket/local/storage/file1
# gs://your-bucket/local/storage/file2
Note
recursive causes directories, buckets, and bucket subdirectories to be
copied recursively. If you upload from local disk to Google Cloud Storage
and set recursive to False
, gs-wrap
will raise an exception and inform you that no URL matched.
This mimicks the behaviour of gsutil
when no wildcards are used.
import os
# Current your-bucket:
# gs://your-bucket/file1
client.cp(
src="gs://your-bucket/file1",
dst="/home/user/storage/file1")
# Your local directory:
# /home/user/storage/file1
Note
All parameters can be used for any kind of cp
operation.
# Parameter: no_clobber example:
import os
# File content before: "hello"
os.stat("/home/user/storage/file1").st_mtime # 1537947563
client.cp(
src="gs://your-bucket/file1",
dst="/home/user/storage/file1",
no_clobber=True)
# no_clobber option stops from overwriting.
# File content after: "hello"
os.stat("/home/user/storage/file1").st_mtime # 1537947563
client.cp(
src="gs://your-bucket/file1",
dst="/home/user/storage/file1",
no_clobber=False)
# File content after: "hello world"
os.stat("/home/user/storage/file1").st_mtime # 1540889799
# Parameter: recursive and multi-threaded example:
# Your local directory:
# /home/user/storage/file1
# ...
# /home/user/storage/file1000
# your-bucket before:
# "empty"
# Execute normal recursive copy in multiple threads.
client.cp(src="/home/user/storage/",
dst="gs://your-bucket/local/",
recursive=True, multithreaded=True)
# your-bucket after:
# gs://your-bucket/local/storage/file1
# ...
# gs://your-bucket/local/storage/file1000
# Parameter: preserve_posix example:
# Your file before:
# /home/user/storage/file1
# e.g. file_mtime: 1547653413 equivalent to 2019-01-16 16:43:33
client.cp(src="/home/user/storage/file1",
dst="gs://your-backup-bucket/file1",
preserve_posix=False)
# your-backup-bucket after:
# gs://your-backup-bucket/file1 e.g. "no metadata file_mtime"
# Preserve the POSIX attributes. POSIX attributes are the metadata of a file.
client.cp(src="/home/user/storage/file1",
dst="gs://your-backup-bucket/file1",
preserve_posix=True)
# your-backup-bucket after:
# gs://your-backup-bucket/file1 e.g. file_mtime: 2019-01-16 16:43:33
sources_destinations = [
# Copy on Google Cloud Storage
('gs://your-bucket/your-dir/file',
'gs://your-bucket/backup-dir/file'),
# Copy from gcs to local
('gs://your-bucket/your-dir/file',
pathlib.Path('/home/user/storage/backup-file')),
# Copy from local to gcs
(pathlib.Path('/home/user/storage/new-file'),
'gs://your-bucket/your-dir/new-file'),
# Copy locally
(pathlib.Path('/home/user/storage/file'),
pathlib.Path('/home/user/storage/new-file'))]
client.cp_many_to_many(srcs_dsts=sources_destinations)
# your-bucket before:
# gs://your-bucket/file
client.rm(url="gs://your-bucket/file")
# your-bucket after:
# "empty"
# your-bucket before:
# gs://your-bucket/file1
# gs://your-bucket/your-dir/file2
# gs://your-bucket/your-dir/sub-dir/file3
client.rm(url="gs://your-bucket/your-dir", recursive=True)
# your-bucket after:
# gs://your-bucket/file1
client.write_text(url="gs://your-bucket/file",
text="Hello, I'm text",
encoding='utf-8')
client.read_text(url="gs://your-bucket/file",
encoding='utf-8')
# Hello I'm text
client.write_bytes(url="gs://your-bucket/data",
data="I'm important data".encode('utf-8'))
data = client.read_bytes(url="gs://your-bucket/data")
data.decode('utf-8')
# I'm important data
Note
POSIX attributes include meta information about a file. When copying a file locally or copying a file within Google Cloud Storage, the POSIX attributes are always preserved. On the other hand, when downloading or uploading file to Google Cloud Storage, the POSIX attributes is only preserved when preserve_posix is set to True.
file = pathlib.Path('/home/user/storage/file')
file.touch()
print(file.stat())
# os.stat_result(st_mode=33204, st_ino=19022665, st_dev=64769, st_nlink=1,
# st_uid=1000, st_gid=1000, st_size=0, st_atime=1544015997,
# st_mtime=1544015997, st_ctime=1544015997)
# Upload does not preserve POSIX attributes.
client.cp(src=pathlib.Path('/home/user/storage/file'),
dst="gs://your-bucket/file")
stats = client.stat(url="gs://your-bucket/file")
stats.creation_time # 2018-11-21 13:27:46.255000+00:00
stats.update_time # 2018-11-21 13:27:46.255000+00:00
stats.content_length # 1024 [bytes]
stats.storage_class # REGIONAL
stats.file_atime # None
stats.file_mtime # None
stats.posix_uid # None
stats.posix_gid # None
stats.posix_mode # None
stats.md5 # b'1B2M2Y8AsgTpgAmY7PhCfg=='
stats.crc32c # b'AAAAAA=='
# Upload with preserve_posix also copy POSIX attributes to blob.
# POSIX attributes are the metadata of a file.
# It also works for downloading.
client.cp(src=pathlib.Path('/home/user/storage/file'),
dst="gs://your-bucket/file", preserve_posix=True)
stats = client.stat(url="gs://your-bucket/file")
stats.creation_time # 2018-11-21 13:27:46.255000+00:00
stats.update_time # 2018-11-21 13:27:46.255000+00:00
stats.content_length # 1024 [bytes]
stats.storage_class # REGIONAL
stats.file_atime # 2018-11-21 13:27:46
stats.file_mtime # 2018-11-21 13:27:46
stats.posix_uid # 1000
stats.posix_gid # 1000
stats.posix_mode # 777
stats.md5 # b'1B2M2Y8AsgTpgAmY7PhCfg=='
stats.crc32c # b'AAAAAA=='
# Check modification time when copied with preserve_posix.
client.same_modtime(path='/home/user/storage/file',
url='gs://your-bucket/file')
# Check md5 hash to ensure content equality.
client.same_md5(path='/home/user/storage/file', url='gs://your-bucket/file')
# Retrieve hex digests of MD5 checksums for multiple URLs.
urls = ['gs://your-bucket/file1', 'gs://your-bucket/file2']
client.md5_hexdigests(urls=urls, multithreaded=False)
The documentation is available on readthedocs.
In order to use this library, you need to go through the following steps:
- Select or create a Cloud Platform project.
- Enable billing for your project.
- Enable the Google Cloud Storage API.
- Setup Authentication using the Google Cloud SDK.
- Install gs-wrap with pip:
pip3 install gs-wrap
- Check out the repository.
- In the repository root, create the virtual environment:
python3 -m venv venv3
- Activate the virtual environment:
source venv3/bin/activate
- Install the development dependencies:
pip3 install -e .[dev]
We use tox for testing and packaging the distribution. Assuming that the virtual environment has been activated and the development dependencies have been installed, run:
tox
We provide a set of pre-commit checks that lint and check code for formatting.
Namely, we use:
- yapf to check the formatting.
- The style of the docstrings is checked with pydocstyle.
- Static type analysis is performed with mypy.
- isort to sort your imports for you.
- Various linter checks are done with pylint.
- Doctests are executed using the Python doctest module.
- pyicontract-lint lints contracts in Python code defined with icontract library.
- twine to check the README for invalid markup which prevents it from rendering correctly on PyPI.
Run the pre-commit checks locally from an activated virtual environment with development dependencies:
./precommit.py
- The pre-commit script can also automatically format the code:
./precommit.py --overwrite
Assuming that the virtual environment has been activated, the development
dependencies have been installed and the PYTHONPATH
has been set to the
project directory, run the benchmarks with:
./benchmark/main.py *NAME OF YOUR GCS BUCKET*
Here are some of our benchmark results:
Benchmark list 10000 files:
TESTED | TIME | SPEEDUP |
gswrap | 3.22 s | - |
gsutilwrap | 3.98 s | 1.24 x |
Benchmark upload 10000 files:
TESTED | TIME | SPEEDUP |
gswrap | 45.12 s | - |
gsutilwrap | 34.85 s | 0.77 x |
Benchmark upload-many-to-many 500 files:
TESTED | TIME | SPEEDUP |
gswrap | 2.14 s | - |
gsutilwrap | 65.2 s | 30.49 x |
Benchmark download 10000 files:
TESTED | TIME | SPEEDUP |
gswrap | 43.92 s | - |
gsutilwrap | 43.01 s | 0.98 x |
Benchmark download-many-to-many 500 files:
TESTED | TIME | SPEEDUP |
gswrap | 5.85 s | - |
gsutilwrap | 62.93 s | 10.76 x |
Benchmark copy on remote 1000 files:
TESTED | TIME | SPEEDUP |
gswrap | 5.09 s | - |
gsutilwrap | 4.47 s | 0.88 x |
Benchmark copy-many-to-many-on-remote 500 files:
TESTED | TIME | SPEEDUP |
gswrap | 6.55 s | - |
gsutilwrap | 62.76 s | 9.57 x |
Benchmark remove 1000 files:
TESTED | TIME | SPEEDUP |
gswrap | 3.16 s | - |
gsutilwrap | 3.66 s | 1.16 x |
Benchmark read 100 files:
TESTED | TIME | SPEEDUP |
gswrap | 16.56 s | - |
gsutilwrap | 64.73 s | 3.91 x |
Benchmark write 30 files:
TESTED | TIME | SPEEDUP |
gswrap | 2.67 s | - |
gsutilwrap | 32.55 s | 12.17 x |
Benchmark stat 100 files:
TESTED | TIME | SPEEDUP |
gswrap | 6.39 s | - |
gsutilwrap | 48.15 s | 7.53 x |
All results of our benchmarks can be found here.
We follow Semantic Versioning. The version X.Y.Z indicates:
- X is the major version (backward-incompatible),
- Y is the minor version (backward-compatible), and
- Z is the patch version (backward-compatible bug fix).