Blob.from_string() should be able to parse all valid gcs uri

[pPanda-beta]

Blob.from_string is terminating parsing of object names at hash(#) character.

Test Code (minimalistic)

from google.cloud import storage
blob = storage.Blob.from_string('gs://my-bucket/my/problamatic/object#name')
print(blob.name)
# my/problamatic/object

Expected

Object name should be my/problamatic/object#name

Actual

Object name found is my/problamatic/object

Temporary Workaround

import re
from google.cloud import storage
GS_PATTERN = re.compile(r"gs://(?P<bucket_name>[^/]+)/(?P<object_name>.+)")

m = GS_PATTERN.match('gs://my-bucket/my/problamatic/object#name')
blob = storage.Bucket(name=m.group('bucket_name'), client=None).blob(m.group('object_name'))
print(blob.name)
# my/problamatic/object#name

Version

google_cloud_storage-2.10.0-py2.py3-none-any.whl

Suggestion

urlsplit is not a good utility for parsing gcs uris. I have seen problems with % char, trailing ?, ... etc with urlsplit. Most of those characters are valid in context of gcs object names.

python-storage/google/cloud/storage/blob.py

Line 406 in 663edc5

scheme, netloc, path, query, frag = urlsplit(uri)

[cojenco]

Thanks for filing. I understand how this could cause a parsing issue. I will work with the team to address next steps.

Please note that we strongly recommended users to avoid # and control characters in object names, see more details in the official documentation.

[frankyn]

Adding findings from Node.js storage after our conversation @cojenco.

Node.js does not yet support a similar method to this but has a FR for it:

However, the client does support gcs URI's in File#copy() and move(). It does support # as part of the object name. Here's an example:

(async () => {
  const {Storage} = require('@google-cloud/storage');

  const bucketName = "bucket-name";

  const storage = new Storage();

  const file = storage.bucket(bucketName).file('object-name');
  const [newFile] = await file.copy("gs://bucket-name/new-year-new-me#random");
  const [metadata] = await newFile.getMetadata();
  console.log(metadata); // Logs metadata and contains object name with `#`
})();