Michael Hanke

`subdatasets` slow for no apparent reason

1

Scenario: dataset with two non-installed subdatasets and 40k top-level subdirectories: ``` % datalad -f disabled subdatasets datalad -f disabled subdatasets 91.01s user 11.94s system 100% cpu 1:42.54 total % time...

`GitRepo.get_content_info()` is called by `GitRepo.get_submodules_()` and also calls `GitRepo.get_submodules_()` in order to ensure compatibility with recent Git versions. This can lead to ``` RecursionError: maximum recursion depth exceeded ``` "Introduced"...

As speculated in https://github.com/datalad/datalad-next/pull/92#issuecomment-1213303309, it seems that a call to `ui.message()` is causing a significant load on the logging system. For commands that issue a lot of messages (e.g. rendered...

https://github.com/datalad/datalad-next/pull/92 brings a new implementation that is based on `pathlib`. We should use it for the `Dataset` method also.

Introduce typechange detection for status reporting and saving

3

The proposed changes improve the accuracy of `status/diff` reporting, and consequently the reliability with which `save` can act on datasets modified in more complex ways (files replaced by subdatasets, etc.)...

`RI` unconditionally whines about obscure path

1

The tests use obscure path names for good reasons. Given that, I do not understand this behavior: ```py >>> from datalad.support.network import RI >>> RI("/tmp/datalad-fs-DAkJS/datalad_temp_tree_test_create_subdataset_hierarchy_from_topkfwtympw/origin/ds-;&%b5{}'ΔЙקم๗あ.datc") [WARNING] ParseResults contains params "&%b5{}'ΔЙקم๗あ.datc",...

Align `status/diff` reporting behavior with `git-diff-files`

1

Git does not report the previously recorded gitsha for content that is modified in the worktree. This change removes the `gitshasum` report from such a result record, in order to...

Report `prev_type` rather than `type` for `diff/status` of deletions

1

This PR sits on top of https://github.com/datalad/datalad/pull/6797 to enable a simple and clear fix. Only the last commit is relevant here. This change aligns the behavior with that of `git...

status/diff report (wrong) type for deletions

2

In the same spirit as #6875: ``` % git diff-files -- README.md :100644 000000 0696c43208dbcc378ed4c0531899eadca6165ca3 0000000000000000000000000000000000000000 D README.md ``` but (even with #6876) ``` % dl -f json_pp status --...

`datalad_next/patches/diff.py` is modified in worktree. Confirmed by `diff-files`. Note how the `gitsha` for the current state is `00000`. ``` % git diff-files -- datalad_next/patches/diff.py :100644 100644 6ac299544a6d7e60afb6bc36012d8578689cea21 0000000000000000000000000000000000000000 M datalad_next/patches/diff.py...