Re: Force usage of pager for diff, show, etc when piping to non-TTY

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Jeff, thank you for your expert and thorough response!
You make a good point that users usually pipe the output of a prettier
tool to a pager themselves. In the case of delta, delta can be
configured to selectively pipe to a pager itself, so delta's docs
doesn't tell users to pipe: `core.pager = delta` (source:
https://dandavison.github.io/delta/configuration.html)

However, for diff-so-fancy, it might not work cleanly because users
are told to explicitly pipe it to a pager. `core.pager "diff-so-fancy
| less --tabs=4 -RFX"` (source:
https://github.com/so-fancy/diff-so-fancy#with-git).

I haven't looked at other tools but it's clear this approach is not
guaranteed to work for all configurations so I should look into
something else.

Jeff, would you be so kind as to elaborate more on the
interactive.diffFilter approach? My understanding is that
interactive.diffFilter is only used for git add -p or git reset -p.
However, the limitation for my use case is I need to use the pager
for git log and git show so that won't work. So then, you are
suggesting that I ask my users to opt in by setting an arbitrary git
config like fzf.pager and then read out the pager from that git var?

On Tue, Aug 15, 2023 at 7:57 PM Jeff King <peff@xxxxxxxx> wrote:
>
> On Tue, Aug 15, 2023 at 05:09:13PM -0700, Patrick wrote:
>
> > I noticed there is no option I can pass to git to use the pager set in
> > my gitconfig even when piping to non-TTY. Is there a workaround? If
> > not, may I request this as a new feature?
>
> I don't think there is a workaround. We have "git --paginate", but it
> really means "paginate this command using the default rules, even if it
> is not a command that is usually paginated".
>
> Looking at the code in setup_pager(), I think the check for "is stdout a
> tty" is fed directly into the decision of whether to use a pager.
> There's no way to override it within Git. You'd have to trick Git by
> opening a pty in the calling program and pumping the data through it
> (which is what our test suite does; see t/test-terminal.perl).
>
> So I think it would need a new option. But...
>
> > Use case: integrate tools like Delta or diff-so-fancy when building
> > wrappers around git commands. See
> > https://github.com/dandavison/delta/discussions/840 and
> > https://github.com/PatrickF1/fzf.fish/discussions/202 for examples.
>
> I'm not quite sure that's what you want. When a user configures a custom
> pager using a prettifier tool like that, they usually further pipe the
> output to a pager like "less". E.g., I have:
>
>   [pager]
>   log = diff-highlight | less
>
> in my config. If you are trying to save output that looks like what the
> user would see on their tty, you want the first half of that pipeline
> (the diff-highlight here), but not the second (less). Of course it
> mostly works because less is smart enough to behave like a noop "cat"
> when stdout isn't a tty. So it might be OK in practice.
>
> I think your script might be better off doing the piping itself. In
> theory you could ask Git what the configured pager is and the run it
> yourself, but:
>
>   1. You can use "git var GIT_PAGER" to get the default pager, but not
>      command-specific ones. So you'd have to check "git config
>      pager.log", etc, yourself, which means reimplementing some of Git's
>      logic.
>
>   2. You'd get a string like "diff-highlight | less", and then you'd
>      have to decide whether to parse off the "| less" part yourself,
>      which is obviously error-prone since the string can be an
>      arbitrarily complex shell expression.
>
> When we were faced with a similar situation within Git, we ended up
> adding a new config option: interactive.diffFilter. This is used by "add
> -p", etc, to filter diffs that are shown to the user. It does mean the
> user has to repeat themselves (I have to set it to "diff-highlight"
> separately from my settings for pager.log, pager.diff, etc). But it's
> unambiguous, and it gives the user the flexibility of configuring
> various outputs differently if they choose.
>
> So depending on what your script does, you could use a similar config
> option. Or even just use interactive.diffFilter if your use case is
> conceptually the same.
>
> -Peff




[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux