Support for Google Cloud Storage connector about hdfs3 HOT 3 CLOSED

Perhaps there should be a separate small library to interact with Google Storage? Libraries like hdfs3 and s3fs are really designed with just their one storage system in mind. They provide similar enough interfaces that it's easy enough to connect them to dask.distributed. Perhaps what you want would require another gsfs project?

It looks like Google Storage is trying hard to piggyback on tools like boto and HDFS clients by mimicking interfaces (see https://cloud.google.com/storage/docs/gspythonlibrary). I'm not particularly surprised that their efforts at mimicking these interfaces isn't perfectly smooth, however, it might be worth seeing if there is some way to coerce Google's efforts with boto to work with boto3, at which point s3fs would be a good candidate (According to the comments in this SO post this appears to be incomplete).

@broxtronix, you're a heavy user of GCS. What would you use to pull out a byte range from a file on GS? Has anyone heard of them working to support boto3?

Alternatively, you could stick with the HDFS interface layer if you were to convince the libhdfs3 project to support whatever odd dialect GS is using, at which point this Python wrapper library would probably work.

from hdfs3.

It seems that GCS already has something giving file-like objects for interaction https://cloud.google.com/appengine/docs/python/googlecloudstorageclient/functions#open

It would not be difficult to write a distcp using distributed from S3 or this GCS to HDFS3 for further use.

from hdfs3.

I have written some proof of concept code, and may be making a separate library to handle GCS.

from hdfs3.