[00/11] refactor git switch completion
mbox series

Message ID 20200425022045.1089291-1-jacob.e.keller@intel.com
Headers show
Series
  • refactor git switch completion
Related show

Message

Jacob Keller April 25, 2020, 2:20 a.m. UTC
From: Jacob Keller <jacob.keller@gmail.com>

completion support for git switch is subpar for a number of cases. Most
notable is difference between these two completions:

  $git switch <TAB>
  Display all 784 possibilities? (y or n)
  <list of all references and DWIM remotes>

  $git switch --track<TAB>
  jk-refactor-git-switch-completion master`

If --track is provided, tab completion becomes almost useless, because we
would expect to complete remote references, but instead can only complete
local branches!

This series was motivated by a desire to fix the completion for the above
two cases, but I noticed several other issues on the way, including some
issues understanding what the current logic did.

This series aims to improve the completion support, and comes with many
additional test cases that should help highlight the buggy behavior and
hopefully prevent future regressions.

The first few commits just add new test cases, most of which currently fail.

Following this is a commit to change __git_complete_refs so that it uses
"--dwim" instead of "--track", since this made reading _git_checkout() and
_git_switch() difficult to read. "--track" was both used as the "enable DWIM
remote branch names" and also the option name for --track.

Following this are some patches to extract displaying DWIM remote branch
names from __git_refs() and refactoring __git_complete_refs to take a mode
argument that switches between calling __git_heads, __git_refs, and a new
__git_remotes.

By doing this, it becomes easier to do things like complete DWIM remote
branches in addition to just regular branches, rather than all references.

With this series applied, completion for git switch behaves more like the
following examples:

  $git switch <TAB>
  HEAD                                master         todo
  jk-refactor-git-switch-completion   next
  maint                               pu

  $git switch --track <TAB>
  origin/HEAD     origin/maint    origin/master   origin/next     origin/pu
  origin/todo

Jacob Keller (11):
  completion: add some simple test cases for git switch completion
  completion: add test showing subpar git switch completion
  completion: add test highlighting subpar git switch --track completion
  completion: add tests showing lack of support for  git switch -c/-C
  completion: remove completion for git switch --orphan
  completion: rename --track option of __git_complete_refs
  completion: extract function __git_dwim_remote_heads
  completion: perform DWIM logic directly in __git_complete_refs
  completion: fix completion for git switch with no options
  completion: recognize -c/-C when completing for git switch
  completion: complete remote branches for git switch --track

 contrib/completion/git-completion.bash | 146 +++++++++++++++++++------
 t/t9902-completion.sh                  | 103 +++++++++++++++++
 2 files changed, 215 insertions(+), 34 deletions(-)

Comments

Jacob Keller April 25, 2020, 10:14 p.m. UTC | #1
On Fri, Apr 24, 2020 at 7:20 PM Jacob Keller <jacob.e.keller@intel.com> wrote:
> Following this are some patches to extract displaying DWIM remote branch
> names from __git_refs() and refactoring __git_complete_refs to take a mode
> argument that switches between calling __git_heads, __git_refs, and a new
> __git_remotes.

This should read __git_remote_heads, since I changed the name to be
more descriptive to avoid thinking it prints the remote names on their
own.

Thanks,
Jake
Junio C Hamano April 30, 2020, 10:56 p.m. UTC | #2
Jacob Keller <jacob.e.keller@intel.com> writes:

> From: Jacob Keller <jacob.keller@gmail.com>
>
> completion support for git switch is subpar for a number of cases. Most
> notable is difference between these two completions:
>
>   $git switch <TAB>
>   Display all 784 possibilities? (y or n)
>   <list of all references and DWIM remotes>
>
>   $git switch --track<TAB>
>   jk-refactor-git-switch-completion master`
> ...

We've discussed that it may be a good idea to make sure that "switch
-c", "checkout -b" and "switch/checkout --orphan" complete the new
branch name the same way, but haven't done anything else.  I'd very
much appreciate to see the patches reviewed by those involved more
in the completion script, before we decide to merge the topic to
'next'.

Thanks.
Jacob Keller May 1, 2020, 9:53 p.m. UTC | #3
On Thu, Apr 30, 2020 at 3:57 PM Junio C Hamano <gitster@pobox.com> wrote:
>
> Jacob Keller <jacob.e.keller@intel.com> writes:
>
> > From: Jacob Keller <jacob.keller@gmail.com>
> >
> > completion support for git switch is subpar for a number of cases. Most
> > notable is difference between these two completions:
> >
> >   $git switch <TAB>
> >   Display all 784 possibilities? (y or n)
> >   <list of all references and DWIM remotes>
> >
> >   $git switch --track<TAB>
> >   jk-refactor-git-switch-completion master`
> > ...
>
> We've discussed that it may be a good idea to make sure that "switch
> -c", "checkout -b" and "switch/checkout --orphan" complete the new
> branch name the same way, but haven't done anything else.  I'd very
> much appreciate to see the patches reviewed by those involved more
> in the completion script, before we decide to merge the topic to
> 'next'.


It is my intention to extend this series with a notion of reading the
previous word on the command line, and using that to determine if we
are completing an argument for an option.

Thanks,
Jake