remove_all should not return error if object doesn't exist.

[damooo]

Currently remove_all returns error if object doesn't exist. Where as delete does not.

Here is the line:

https://github.com/datafuselabs/opendal/blob/c4ada149d68d4822bd7b5403009e2efaff6739b8/src/types/operator/operator.rs#L734

[Xuanwo]

Thank you for the report! We have identified this as a bug and it will be addressed through #1585.

[damooo]

Also, it depends on scan as in here to get flat list.

Backends like fs doesn't have scan capability. Instead can we fallback to walk dir tree and delete for such backends which only has list capability?

@Xuanwo ?

[damooo]

Object entries seems retrieved trough only scan here?

https://github.com/datafuselabs/opendal/blob/aeb751d65336e854885fff16a1d43bc25656bea5/src/types/operator/operator.rs#L740

[Xuanwo]

We did this at CompleteLayer

[damooo]

Oh, great then.

And next comment seems more plausible though, as ordering of entries may not guaranteed, for child to come first in fs case.

Also it probably can fail unnecessarily, if a backend supports scan, but has true directories.
Then deleting directory before a child, if it's entry came first in scan, can cause error.

[damooo]

Is there anything i am missing regarding above question regarding ordering? Evaluating remove_all if it is suited to purge remnant directories in an op.

[Xuanwo]

Is there anything i am missing regarding above question regarding ordering? Evaluating remove_all if it is suited to purge remnant directories in an op.

files will be guaranteed to be return before the dir itself.

[damooo]

seems fs, hdfs, webdav, etc doesn't support scan.

Also it probably can fail unnecessarily, if a backend supports scan, but has true directories.
Then deleting directory before a child, if it's entry came first in scan, can cause error.