Re: the pager

[Jeff King <peff@peff.net>]

On Mon, Sep 02, 2013 at 10:37:45PM -0400, Dale R. Worley wrote:

> > I've noticed that Git by default puts long output through "less" as a
> > pager.  I don't like that, but this is not the time to change
> > established behavior.  But while tracking that down, I noticed that
> > the paging behavior is controlled by at least 5 things:
> > 
> > the -p/--paginate/--no-pager options
> > the GIT_PAGER environment variable
> > the PAGER environment variable
> > the core.pager Git configuration variable
> > the build-in default (which seems to usually be "less")

This list has some orthogonal concepts. The "-p" and "--no-pager"
variables decide _whether_ to run the pager. The GIT_PAGER and PAGER
environment variables, along with core.pager and the compile-time
default, decide _which_ pager to run.

The fact that "cat" or the empty string becomes "no pager" is purely an
optimization (we could fork and run "sh -c ''" or "cat", but it would be
a no-op). So even though you might have instructed git to run the pager,
it may be a noop if your pager is "cat", and we optimize it out.

The confusing one (and missing from your list) is pager.$program, which
originally was a "whether", but later learned to optionally be a
"which". And you also omit the built-in defaults for "whether" on each
command (e.g., "log" runs a pager, "push" does not).

> There is a somewhat independent question of when the pager is
> activated.  What I know so far is that some commands use the pager by
> default and some by default do not.  My expectation is that
> --no-pager can be used to suppress the pager for *any* command.  Is it
> also true that -p can force the pager for *any* command, or are there
> commands which will not page even with -p?

Yes, --no-pager and -p suppress or force, respectively, for any command.
They take precedence over config (pager.$command), which in turn takes
precedence over builtin defaults (per-command defaults, in this case).

Environment variables should generally be less than command-line
options, but greater than config. But there is no "definitely use a
pager" environment variable, so it doesn't apply here.

And I say generally because we should put git-specific environment
variables over git-specific config, but git-specific config over general
environment variables (so similarly we should respect user.email in the
config over $EMAIL in the environment, but under $GIT_COMMITTER_EMAIL).

> I assume that if -p is specified but the "which pager" selection is
> "cat" (or some other specification of no pager), then there is no
> paging operation.

There is a pager in that case, but it doesn't do anything. And then we
optimize it out because it doesn't do anything. :) That is somewhat
tongue-in-cheek, but I hope it shows the mental model that goes into the
decision.

-Peff