GNU bug report logs - #63861
[PATCH] pp.el: New "pretty printing" code

Previous Next

Full log

Message #89 received at 63861 <at> debbugs.gnu.org (full text, mbox):

From: Stefan Monnier <monnier <at> iro.umontreal.ca>
To: Visuwesh <visuweshm <at> gmail.com>
Cc: Eli Zaretskii <eliz <at> gnu.org>, 63861 <at> debbugs.gnu.org
Subject: Re: bug#63861: [PATCH] pp.el: New "pretty printing" code
Date: Fri, 09 Jun 2023 10:59:18 -0400

>>> BTW, how does this compare to the newly added `pp-emacs-lisp-code`?
>> Very good question.  I had completely missed that (and its `pp-use-max-width`).
>> This points to a host of integration issues between my code and the
>> existing code.  I'll have to take a deeper look.
> From what I remember, pp simply switches to use pp-emacs-lisp-code when
> the relevant user option is set.

Yup, similar to my patch, except my patch hooks into `pp-buffer`,
i.e. a lower level which hence affects more uses.

> The poor performance of pp-emacs-lisp-code made me wish
> pp-use-max-width was only obeyed by user facing commands like
> pp-eval-expression & friends.

My tests yesterday suggest `pp-emacs-lisp-code` is simply too slow to be
used except when we know beforehand that the sexp to be printed
is small.  And I can't think of a single piece of code where that's
the case.  I suspect part of the code can be improved to bring the
computational complexity of the code closer to linear, but until
someone does that, I think `pp-use-max-width` is just unusable.
Instead we could/should provide ways for the user to interactively call
a command to reformat something using `pp-emacs-lisp-code`.  Or maybe
change the code so `pp-emacs-lisp-code` is used only when the total
printed size is below a certain threshold.

>> My new code is expected to be slower than the "normal" pretty-printer,
>> but barring performance bugs in `lisp-indent-line` (such as the one
>> fixed by the patch I just sent to Thierry) it should be approximately
>> a constant factor slower.
>> AFAICT the performance of `pp-emacs-lisp-code` can be more problematic.
> Hopefully, the constant factor is quite small.

In my tests, 10x seems to be the "worst case slowdown" of `pp-region`.
On some of the tests, it's actually faster, sometimes significantly so
(presumably due to some non-linear complexity in some parts of
`pp-buffer`).

> pp-emacs-lisp-code took a lot of time to print my modest bookmark
> alist (28 entries) and for the longest time I thought some other code
> in Emacs has gone awry.

AFAICT it suffers from a pretty bad complexity.

> I haven't tested your pretty printer but pp-emacs-lisp-code could give
> some really bizarre results for lisp data.

I haven't seen "really bizarre results" with `pp-region` yet, but
I wouldn't be surprised if that can happen: I've been playing with
various pretty-printing alternatives over the years and they all seem to
suffer from weird behaviors occasionally, except for those that insert
too many line breaks.


        Stefan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.