Hi! I learned about Dataplug today at EuroSciPy from @macarronesc, looks very interesting.
One of the questions asked by the audience was "how to use this with non-AWS storage", and the answer was to write a custom plugin. I notice that Dataplug uses boto3, and I wonder if the project could benefit from obstore https://developmentseed.org/obstore/, which is cloud-agnostic, fast (thanks to Rust), and maybe fixes the pickle-ability issues?
|
class PickleableS3ClientProxy: |
|
""" |
|
A Pickleable S3 client proxy that can be pickled and unpickled. |
|
|
|
Dataplug requires having an S3 client that can be pickled and sent remotely to workers, so that |
|
remote workers can access S3 objects. This class is a proxy to an S3 client that can be pickled. |
See https://developmentseed.org/obstore/latest/alternatives/#obstore-vs-boto3
Hi! I learned about Dataplug today at EuroSciPy from @macarronesc, looks very interesting.
One of the questions asked by the audience was "how to use this with non-AWS storage", and the answer was to write a custom plugin. I notice that Dataplug uses
boto3, and I wonder if the project could benefit fromobstorehttps://developmentseed.org/obstore/, which is cloud-agnostic, fast (thanks to Rust), and maybe fixes the pickle-ability issues?dataplug/dataplug/storage/picklableS3.py
Lines 37 to 42 in 3f1eca1
See https://developmentseed.org/obstore/latest/alternatives/#obstore-vs-boto3