How to run Google gsutil using Python

Question

After installing and configuring Google Cloud SDK gsutil command can be run by simply typing its name and the argument(-s) using Windows cmd.

Here is the example:

"C:\Program Files (x86)\Google\Cloud SDK\google-cloud-sdk\bin\gcloud" version

But the same command fails if run using Python subprocess. With subprocess's shell argument set to True the ImportError occurs:

import subprocess

cmd = '"C:/Program Files (x86)/Google/Cloud SDK/google-cloud-sdk/bin/gsutil" version'

p = subprocess.Popen(cmd, shell=True)

.....

ImportError: No module named site

With subprocess's shell argument set to False then the WindowsError: [Error 2] The system cannot find the file specified occurs:

p = subprocess.Popen(cmd, shell=False)

Is there a way to run gsutil on Windows using Python?

Solution

Note that the proper and official way to interact with Google Cloud Storage is to make use of the Google Cloud Client Library for Python and not running the gsutil command through subprocess.Popen. If you are not setting up merely some tests I would suggest you to follow from the beginning this way if there is not any technological constrain that makes this way impracticable.

You can check at the following links the relative Overview and Documentation. A small example taken from the Documentation can be the following:

from google.cloud import storage

client = storage.Client()
bucket = client.get_bucket('<your-bucket-name>')
blob = bucket.blob('my-test-file.txt')
blob.upload_from_string('this is test content!')

You can find a further example here using google-cloud-python with the Datastore and Cloud Storage to manage expenses.

Comments

[HoosierDaddy]

The python API does not allow use of the -m option for parallelism, as far as I know. So there are reasons for using subprocess and the gsutil command.

[tfad334]

@UricSou: You can share client instances across threads because the storage client uses the requests library. Just create client instances after multiprocessing.Pool.

[CpILL]

Also, the python API is dead slow comapred to the commandline

[MAC]

@GalloCedrone how to upload folder using above approach?