Dave Berenbaum
Dave Berenbaum
Related ideas: * Show the status of these files for the workspace row. For example, we could add a letter next to the hash to indicate the status or color...
`checkout-noop` benchmarks are also slower:
Do you mean for the `checkout` behavior specifically? Is it expected that it will also impact `add-noop`? And by that much?
@skshetry Do we have any plan to address this yet?
> Before we used to be intelligent, and used `modifiedTime` of the directory to skip on `noop`. Is there any plan to get back to this? Optimizing diff and staging...
Interesting, thanks @skshetry. Not sure what's going on with the benchmarks in that case. I'm not seeing the reported regression when testing with an unzipped http://images.cocodataset.org/zips/test2014.zip. In fact, it appears...
Thanks. Using copy link type, I'm seeing a slight regression, but also it's way faster than it was for reflink. Is that expected? 2.6.3: ```console $ dvc config cache.type copy;...
> Even in copy linktype, dvc still tries to reflink first and fallbacks to copy. Any idea why the operation was so much faster?
True, that could be nice, but I was thinking of just testing in s3 like we do for regular remotes to make sure the basic workflow has reasonable benchmarks in...
Sorry, my mistake here. I guess we don't test s3 for regular remotes with dvc-bench now. So maybe we either need to test on at least one cloud remote or...