bug#59502: 29.0.50; [PATCH] Dedicated buffers per project

[Dmitry Gutov]

On 06/12/2022 19:21, Juri Linkov wrote:
> > > Idly wondering if it'd make sense to also pass the major-mode to
> > > project-buffer-name-function, since string matching might be unreliable
> > > in some situations?  Can't think of any concrete problem off the top of
> > > my head, I just have a vague expectation that mode symbols might be more
> > > stable than buffer names.  It's unsubstantiated though, so feel free to
> > > dismiss.
> >
> > I agree, and actually my first implementation used the major-mode symbol
> > as the argument to function project-buffer-name-function.  I changed to
> > a string to be able to handle the following cases:
>
> A string is too confusing.  After seeing
>
> +        (shell-command-buffer-name (funcall project-buffer-name-function
> +                                            "Shell Command Output")))

I think that's not too bad.

> the first thought was: why the buffer name is without asterisks
> as in the standard name "*Shell Command Output*".

> > | Command                     | major-mode       |
> > |-----------------------------+------------------|
> > | project-shell-command       | fundamental-mode |
> > | project-async-shell-command | shell-mode       |
>
> More consistent would be to accept a symbol as the argument of
> project-buffer-name-function.  Then every project-supported command
> could have the corresponding symbol that is part of the command name:
>
> | Command                     | name        |
> |-----------------------------+-------------|
> | project-shell-command       | shell       |
> | project-async-shell-command | async-shell |
>
> Also for calling the user customized function it would be useful to
> provide an argument with the standard buffer name.  So users won't need
> to copy the mapping from project-buffer-name-default.  Then the default
> implementation will be just:
>
> (defun project-buffer-name-default (project command-symbol buffer-name)
>    (or buffer-name (format "*%s-%s*" (project-name project) command-symbol)))
>
> And here is an example of how I'd like to customize it:
>
> (setopt project-buffer-name-function
>    (lambda (project _command-symbol buffer-name)
>      (format "%s<%s>" buffer-name (project-name project))))

I think *prjname/Shell Command Output* sounds nice enough.

If it doesn't, we probably wouldn't choose *Shell Command Output* as the 
name for non-project buffers of this type, would we?