Throttle rate of output messages

[fischman]

Building a large project with ninja output going to an emacs compilation buffer is painful b/c emacs doesn't do so well with 10k-line-buffers.

IWBN to have a rate-limit on progress messages, such as "don't emit a progress message more often than once per seconds". Could be done with an env var or a cmd-line option to ninja (I weakly prefer env var).

WDYT, worth writing a patch for?

[nico]

This might make sense for the one-line-overwriting mode, but terminals seem to be fast enough to handle the current behavior.

emacs probably gets the output one sees when running ninja | cat – the more unixy way of handling this would be to write a standalone program that reads lines from stdin and only writes one every N ms to stdout, dropping the rest, and have emacs consume ninja output through that. This sounds general enough that it might already exist.

[fischman]

Nico: I thought about an external throttler, but that seems to fail because
it'd be unable to keep line that are associated together, i.e. multi line
errors, or output belonging to an action.
On Oct 10, 2013 6:36 PM, "Nico Weber" notifications@github.com wrote:

This might make sense for the one-line-overwriting mode, but terminals
seem to be fast enough to handle the current behavior.

emacs probably gets the output one sees when running ninja | cat – the
more unixy way of handling this would be to write a standalone program that
reads lines from stdin and only writes one every N ms to stdout, dropping
the rest, and have emacs consume ninja output through that. This sounds
general enough that it might already exist.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/663#issuecomment-26106826
.

[nico]

Clearly, ninja should have xml output, then the filter would be easier to write :-D

Does this help: http://stackoverflow.com/questions/11239201/can-i-limit-the-length-of-the-compilation-buffer-in-emacs ?

[fischman]

Sadly comint-truncate-buffer actually makes things far worse.

[fischman]

Nico: since none of the workarounds work WDYT of the proposed patch above?

[nico]

There's a long list of things people want to configure ninja's output. I'd like to collect a list of all of that and put that in a tracking bug, and ideally come up with some cohesive way to address most of them.

[fischman]

Filed #746 for the tracking bug.

[fischman]

I've been out of the ninja game for years so unsubscribing from this. Feel free to close if no longer useful, or @ me if you want me to see a followup.

[bradneuman]

I recently started using cmake with ninja in emacs and encountered this same problem. Built-in support for this would be much nicer, but I took the approach of creating a quick-and-dirty "throttling" script in python here: https://gist.github.com/bradneuman/3f2bcb3b3a31f00129f69fead4d94c5f

YMMV and you may have to change the (very simple) regex I used there, but for now it did the trick for me. It just throttles to one-per-second for lines matching a certain regex and prints everything else

[jonesmz]

Here's one way that this could be implemented: #1631

(Personally, I think librari-fying ninja is a better approach, so that the lib can be loaded by an application that offers customized output as desired)