Tab completions shouldn't move the cursor when there are no completions #4128
Conversation
The name `comp_empty` suggests that it's true whenever the vector `comp` is empty. However, it's actually the exact opposite because handle_completions() returns true if it inserted text into the command line.
This is something that you want to test with expect. See e.g.
That's something we definitely do not want to change. When a completion is accepted, the cursor should be moved to the end. What I don't like particularly is that this moves the cursor to the end, repaints and then moves it back. That's kinda visually jarring. Have you tried skipping the cursor movement entirely if there is no completion? |
|
This currently doesn't work as well as it should, so I'm closing this for now. |
Description
This fixes #4124
This fixes the first example,
mpv --|'https://...very-very-long-url...'. I'm not quite sure how to test it, however, because it changes thereader.cpploop, which doesn't seem to have any tests infish_tests.cpp.The second example,
foo /path-typo|:/usr/lib, is much more complicated and seems to be because the functioncompleter_t::complete_param_expand()incomplete.cppdoesn't know the cursor's position, and all surrounding code, including the maincomplete()function, assumes completions are done at the end of a token.I looked into this a little bit and made a WIP branch
complete_param_expand_enhancement_wipand it's possible to have the cursor not move when there are no completions and even expand paths before the:too. But with my current implementation I don't think it's possible to do this and move the cursor to the end when there is a completion e.g.fish --ver|si->fish --version|, which is probably a big breaking change.