This gist describes how to undelete files in GCS while retaining most of the metadata. I couldn't find any examples of how to do this other than in the UI. GCS doesn't seem to use delete markers like AWS. If anyone knows a better way please say so.

undelete_gcs_objects.py

# This gist describes how to undelete files in GCS while retaining most of the metadata.
# GCS doesn't seem to use delete markers like AWS. If anyone knows a better way please say so in the comments.

# I couldn't find any examples of how to do this other than in the UI.
# GCS documentation doesn't seem to include instructions for restoring versioned objects in python, but I found
# that this method works. Confusingly they use terms like version, generation, and revision interchangably in
# the documentation. I couldn't understand how object deletion works on versioned buckets.

import tempfile
from google.cloud import storage

# Set your bucket settings here

# bucket_name = 'my_bucket'
# directory_name = 'this/is/my/path/'
client = storage.Client()
bucket = client.get_bucket(bucket_name)

# list all deleted objects in the directory of interest
deleted_blobs = [
  blob for blob in bucket.list_blobs(prefix=directory_name, versions=True)
  if blob.time_deleted
]

for blob in deleted_blobs:
  # Copy old file revision into a temporary file
  tf = tempfile.NamedTemporaryFile(prefix="", suffix=blob.name)
  blob.download_to_file(tf)
  tf.seek(0)

  # Create a new blob generation
  new_blob = bucket.blob(blob_name=blob.name)
  # Copy old properties over
  new_blob._properties = blob._properties.copy()

  # Clear the ones that mark file generation/deletion/revision/version of the old object
  new_blob._properties.pop('timeDeleted')
  new_blob._properties.pop('generation')
  new_blob._properties.pop('metageneration')

  # Upload the old version as another new version
  new_blob.upload_from_file(tf, rewind=True)