What we are trying to say is that FileSystems will have a convenient and
nice interface that does the right thing yet each FileSystem will have an
arbitrary and/or overly complex interface. If we have users only relying on
using FileSystems, why do we need that complexity within FileSystem
implementations?

On Thu, Sep 20, 2018 at 12:46 PM Tim Robertson <[email protected]>
wrote:

> Thanks everyone.
>
> @chamikaramj <https://github.com/chamikaramj> - if we are to correctly
> support an OVERWRITE_EXISTING_FILES flag I think we'd either have to pass
> the flag down into the filesystem implementations themselves, or do like
> the HDFS one does now and always throw FileAlreadyExistsException (note a
> user reports that issue today on user@ with AvroIO and LocalFileSystem)
> when the target is in place. The later would require some of them to do a
> "does target exist" check before attempting the move. Do you have a
> preference on that?
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <https://github.com/apache/beam/pull/6289#issuecomment-423309448>, or mute
> the thread
> <https://github.com/notifications/unsubscribe-auth/AJnK7OIo-UAHqH6UokdDFhpr5hkV6BM2ks5uc_CagaJpZM4WQHrB>
> .
>


[ Full content available at: https://github.com/apache/beam/pull/6289 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to