rmlint icon indicating copy to clipboard operation
rmlint copied to clipboard

Deprecation of --write-unfinished in develop branch

Open Flurrywinde opened this issue 3 years ago • 4 comments

In the develop branch, I noticed --write-unfinished is deprecated, so I was going to submit a PR to state this in the man page. However, I have a few questions:

  1. Does --xattr now include --hash-unmatched?
  2. rmlint -U still worked, but isn't -U a synonym for --write-unfinished?

Flurrywinde avatar May 11 '21 15:05 Flurrywinde

  1. Does --xattr now include --hash-unmatched?

Optionally. So:

  • rmlint --xattr will write checksums for twins only
  • rmlint --xattr --hash-unmatched will write [full file] checksums for size twins (files that have at least one other file of the same size)
  • rmlint --xattr --hash-uniques will write [full file] checksums for essentially all files. Although files filtered out by options such as --size or match-basename will be excluded. Not sure about zero-sized files, that would be kinda pointless, will check.
  1. rmlint -U still worked, but isn't -U a synonym for --write-unfinished?

I mapped -U to --hash-unmatched because I figured that was functionally very similar to --write-unfinished and unlikely to lead to nasty surprises. Only downside is slightly longer runtime than --write-unfinished.

SeeSpotRun avatar May 12 '21 06:05 SeeSpotRun

Awesome. Thanks for clarifying. While I'm editing the man page, are there any other changes I should put in?

Flurrywinde avatar May 12 '21 16:05 Flurrywinde

While I'm editing the man page...

Hey thanks so much for that, I really appreciate it.

...are there any other changes I should put in?

Probably. From a quick look through https://github.com/sahib/rmlint/pulls?q=is%3Apr+closed%3A%3E2021-01-01+ I'd say check on:

  • Options for rmlint --dedupe (run rmlint --dedupe -h for listing*)
  • Similarly for rmlint --is-reflink
  • New sub-option --rank-by [fF] to rank originals by full path

* note that `rmlint --dedupe --xattr is about to be deprecated / removed so don't waste any time documenting that.

Cheers, SSR

SeeSpotRun avatar May 12 '21 22:05 SeeSpotRun

Is there any chance that --write-unfinished is going to be reintroduced at some point?

On my dataset --hash-unmatched makes rmlint scan 20 TiB of data (basically all of it) instead of just 4 TiB with just --xattr.

intelfx avatar Sep 16 '21 01:09 intelfx