threads / discuss / 43893

Re: git rm --cached should be git rm --cache or git rm --stage

Subject: Re: git rm --cached should be git rm --cache or git rm --stage

## tl;dr

6 messages between Aug 20, 2016 and Aug 20, 2016.

replies: 5people: 3as markdown or json

Zenaan Harkness· Aug 20, 2016, 07:22 UTC · lore

git rm --cached should be git rm --cache or git rm --stage

  Please CC me :)
From man git-rm:
--cached
  Use this option to unstage and remove paths only from the index.
  Working tree files, whether modified or not, will be left alone.
This wording is unclear and dangerous, and ought be cleaned up somehow.
Probably also the option name should change.

See here for examples of community self-help compensating for git-rm man page's particularly bad wording: http://stackoverflow.com/questions/6919121/why-are-there-2-ways-to-unstage-a-file-in-git

I suggest:
 - change "--cached" to "--cache" at the least (allow both)
 - probably deprecate --cached and add "--stage" to replace it
 - perhaps change the wording in the man page above to something similar
   as on the stackoverflow page above,
   e.g.:
   "Removes a file from the index. In the case where the file is already
   in the repo, git rm --cached will remove the file from the index,
   leaving it in the working directory and a commit will now remove it
   from the repo as well. Basically, after the commit, you would have
   unversioned the file and kept a local copy."
   or perhaps something like:
   "does not unstage a file, it actually stages the removal of the
   file(s) from the repo (assuming it was already committed before) but
   leaves the file in your working tree (leaving you with an untracked
   file)"

The git "stage" is a primary concept, and a primary noun (one reason many of us have come to appreciate git), and git's cmd line options and help docs ought reflect this.

Thanks, Zenaan

Zenaan Harkness· Aug 20, 2016, 07:25 UTC · re: Zenaan Harkness · lore
   Please CC me :)
Show 14 quoted lines
> From man git-rm:
> 
> --cached
>   Use this option to unstage and remove paths only from the index.
>   Working tree files, whether modified or not, will be left alone.
> 
> 
> This wording is unclear and dangerous, and ought be cleaned up somehow.
> 
> Probably also the option name should change.
> 
> See here for examples of community self-help compensating for git-rm man
> page's particularly bad wording:
> http://stackoverflow.com/questions/6919121/why-are-there-2-ways-to-unstage-a-file-in-git

Also, perhaps a git porcelain command "git unstage" to unstage staged changes. Something small like this might make a huge difference for git newbies.

Junio C Hamano· Aug 20, 2016, 17:48 UTC · re: Zenaan Harkness · lore
Zenaan Harkness <zen@freedbms.net> writes:
Show 7 quoted lines
> From man git-rm:
>
> --cached
>   Use this option to unstage and remove paths only from the index.
>   Working tree files, whether modified or not, will be left alone.
>
> This wording is unclear and dangerous, and ought be cleaned up somehow.

I think "unstage and" can be removed to make it more clear [*1*], but otherwise I do not see much wrong with the description to make it dangerous.

Can you elaborate?  Which part of the description is unclear?

"git help cli" gives a brief description as to why it should be called "--cached".

[Footnote]

*1* The reason why "unstage" makes it ambiguous is because people tend to use the word "stage" loosely. Even though Git keeps track of the entire state of a file, it is easy to confuse oneself to mistakenly think that "git add" a modified file as "staging the difference between the last committed state and the contents being added", and from that mistaken world view, you would imagine "unstage" may bring the last committed state back in the index, which is not what "git rm --cached" does (it does "remove" the path from the index).

Philip Oakley· Aug 20, 2016, 19:14 UTC · re: Zenaan Harkness · lore
From: "Zenaan Harkness" <zen@freedbms.net>
Show 21 quoted lines
>
>  Please CC me :)
>
>
> From man git-rm:
>
> --cached
>  Use this option to unstage and remove paths only from the index.
>  Working tree files, whether modified or not, will be left alone.
>
>
> This wording is unclear and dangerous, and ought be cleaned up somehow.
>
> Probably also the option name should change.
>
> See here for examples of community self-help compensating for git-rm man
> page's particularly bad wording:
> http://stackoverflow.com/questions/6919121/why-are-there-2-ways-to-unstage-a-file-in-git
>
> I suggest:
> - change "--cached" to "--cache" at the least (allow both)

Such small nuance changes rarely have the effect hoped for when there are such gross misunderstandings by the user.

Show 11 quoted lines
>
> - probably deprecate --cached and add "--stage" to replace it
>
> - perhaps change the wording in the man page above to something similar
>   as on the stackoverflow page above,
>   e.g.:
>   "Removes a file from the index. In the case where the file is already
>   in the repo, git rm --cached will remove the file from the index,
>   leaving it in the working directory and a commit will now remove it
>   from the repo as well. Basically, after the commit, you would have
>   unversioned the file and kept a local copy."

The idea of 'removing from the repo' can have a connotation that it means total removal from history(the repository hold the history!), and not just future commits.

Show 7 quoted lines
>
>   or perhaps something like:
>   "does not unstage a file, it actually stages the removal of the
>   file(s) from the repo (assuming it was already committed before) but
>   leaves the file in your working tree (leaving you with an untracked
>   file)"
>

The easiest way is to simply swap around the two sentences so that the positive action is listed first - this better matches people's typical cognition. Human Error (by Reason)[1] tells us to Never state warnings and caveats after the instruction, and preferably be positive.

"--cached: Working tree files, whether modified or not, will be retained unchanged. The option will remove paths from the index (only) to unstage them from future commits."

Show 8 quoted lines
>
> The git "stage" is a primary concept, and a primary noun (one reason
> many of us have come to appreciate git), and git's cmd line options and
> help docs ought reflect this.
>
> Thanks,
> Zenaan
> --

Philip [1] https://www.amazon.com/Human-Error-James-Reason/dp/0521314194

Zenaan Harkness· Aug 20, 2016, 22:25 UTC · re: Philip Oakley · lore
On Sat, Aug 20, 2016 at 08:14:25PM +0100, Philip Oakley wrote:
> From: "Zenaan Harkness" <zen@freedbms.net>
> >
> > Please CC me :)
Show 16 quoted lines
> >  or perhaps something like:
> >  "does not unstage a file, it actually stages the removal of the
> >  file(s) from the repo (assuming it was already committed before) but
> >  leaves the file in your working tree (leaving you with an untracked
> >  file)"
> >
> 
> The easiest way is to simply swap around the two sentences so that the
> positive action is listed first - this better matches people's typical
> cognition. Human Error (by Reason)[1] tells us to Never state warnings and
> caveats after the instruction, and preferably be positive.
> 
> "--cached:
> Working tree files, whether modified or not, will be retained unchanged.
> The option will remove paths from the index (only) to unstage them from
> future commits."
That's much better. +1. Thanks.
Show 11 quoted lines
> >
> >The git "stage" is a primary concept, and a primary noun (one reason
> >many of us have come to appreciate git), and git's cmd line options and
> >help docs ought reflect this.
> >
> >Thanks,
> >Zenaan
> >--
> 
> Philip
> [1] https://www.amazon.com/Human-Error-James-Reason/dp/0521314194
Philip Oakley· Aug 20, 2016, 23:06 UTC · re: Zenaan Harkness · lore
From: "Zenaan Harkness" <zen@freedbms.net>
Show 25 quoted lines
> On Sat, Aug 20, 2016 at 08:14:25PM +0100, Philip Oakley wrote:
>> From: "Zenaan Harkness" <zen@freedbms.net>
>> >
>> > Please CC me :)
>
>> >  or perhaps something like:
>> >  "does not unstage a file, it actually stages the removal of the
>> >  file(s) from the repo (assuming it was already committed before) but
>> >  leaves the file in your working tree (leaving you with an untracked
>> >  file)"
>> >
>>
>> The easiest way is to simply swap around the two sentences so that the
>> positive action is listed first - this better matches people's typical
>> cognition. Human Error (by Reason)[1] tells us to Never state warnings 
>> and
>> caveats after the instruction, and preferably be positive.
>>
>> "--cached:
>> Working tree files, whether modified or not, will be retained unchanged.
>> The option will remove paths from the index (only) to unstage them from
>> future commits."
>
> That's much better. +1. Thanks.
>
One better maybe to add (at the end): "The paths become untracked."

That way we hit all the key codewords that the user may recognise, or need to see, to understand.

Submittimg a patch is faily easy, especially if you have git send-email set up. "patches to git@vger.kernel.org (read Documentation/SubmittingPatches for instructions on patch submission)." (from the README.md)

https://github.com/git/git/blob/master/Documentation/SubmittingPatches
Show 17 quoted lines
>
>> >
>> >The git "stage" is a primary concept, and a primary noun (one reason
>> >many of us have come to appreciate git), and git's cmd line options and
>> >help docs ought reflect this.
>> >
>> >Thanks,
>> >Zenaan
>> >--
>>
>> Philip
>> [1] https://www.amazon.com/Human-Error-James-Reason/dp/0521314194
> --
> To unsubscribe from this list: send the line "unsubscribe git" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

← back to recent threads