Allow #clear! to accept block for conditional deletion

[hwo411]

• rescue NotFound error if the file was already deleted

Enables conditional cache clearing and also doesn't raise if some other process just deleted the object you want to delete.

[hwo411]

Btw I realized that all clear! methods in shrine storages have different arguments. Does it make sense to unify the interface of these methods in shrine?

[rosskevin]

I was just looking to do a similar thing (delete older than x), thanks for putting this up @hwo411. @renchap @janko-m do you have thoughts on unifying the clear! args or a preferred signature?

[janko]

Since #clear! method is not used in Shrine, I didn't think about unifying it. I think the most flexible is a block-based API, the one S3 storage uses and the one you implemented here, so definitely 👍on this PR. I'm thinking of adding block-based API to FileSystem storage as well.

[janko]

Hmm, but this implementation removes batching deletes, and now each will will be deleted individually. You can still elegantly keep batching with File::List#all and lazy Enumerators:

def clear!(&block)
  prefix = "#{@prefix}/" if @prefix
  files = get_bucket.files prefix: prefix

  all_files = files.all.lazy.select(&block)

  batch_delete(all_files)
end

def batch_delete(files)
  files.each_slice(100) do |file_batch|
    file_batch.each(&:delete) # can we batch deletes with google-cloud-storage gem?
  end
end

I now realized that batching deletes must have been removed when migrating from google-api-client gem to google-cloud-storage gem. I'm wondering if it's still possible to have batching.

[hwo411]

I didn't find how to do batch deletes while I was working on this PR, looks like there is no such option for Google Cloud.

[renchap]

Thanks for this PR, I will review it shortly.
For the batched delete issue, I opened a feature request in google-cloud-storage: googleapis/google-cloud-ruby#2914

[renchap]

Thanks for the PR, merged!
I will open an issue for the batch delete to be implemented once google-cloud-storage supports it.