From: Tianxiang Xiong <tianxiang.xiong@gmail.com>
To: Stefan Monnier <monnier@iro.umontreal.ca>
Cc: Emacs developers <emacs-devel@gnu.org>
Subject: Re: Performance issue w/ `cl-loop`s `collect...into`
Date: Sun, 8 Apr 2018 16:29:56 -0700 [thread overview]
Message-ID: <CACMkxiw5Eag=y8P1FwEFrJNg-TpZmFU3QyJvXHPseS4rjdTgRg@mail.gmail.com> (raw)
In-Reply-To: <jwvtvsl8lqf.fsf-monnier+gmane.emacs.devel@gnu.org>
[-- Attachment #1: Type: text/plain, Size: 3566 bytes --]
One thing I don't understand is the common
(push `(progn (setq ...) t) cl--loop-body)
pattern found in the code. I'm not sure why the `(progn ... t)` is
necessary. If anyone could explain that I'd add it as a comment.
On Sun, Apr 8, 2018 at 2:13 PM, Stefan Monnier <monnier@iro.umontreal.ca>
wrote:
> > Avoid O(n^2) nconc-ing by keeping track of tail of collection.
>
> I took a quick look at your patch, and it looks pretty good.
> See comments below.
>
>
> Stefan
>
>
> > ((memq word '(collect collecting))
> > - (let ((what (pop cl--loop-args))
> > - (var (cl--loop-handle-accum nil 'nreverse)))
> > - (if (eq var cl--loop-accum-var)
> > - (push `(progn (push ,what ,var) t) cl--loop-body)
> > - (push `(progn
> > - (setq ,var (nconc ,var (list ,what)))
> > - t)
> > - cl--loop-body))))
> > + (let ((what (pop cl--loop-args)))
> > + (cl-multiple-value-bind (var var-tail)
>
> `cl-multiple-value-bind` is the "destructor" corresponding to the
> `cl-values` "constructor". Since your code doesn't use `cl-values` it
> should not use `cl-multiple-value-bind` either (you probably meant to
> use cl-destructuring-bind instead).
>
> > + (cl--loop-handle-accum nil 'nreverse)
> > + (if (eq var cl--loop-accum-var)
> > + (push `(progn (push ,what ,var) t) cl--loop-body)
> > + (push `(progn
> > + (if (null ,var-tail)
> > + (setq ,var (list ,what) ,var-tail (last ,var))
> > + (setq ,var-tail (setcdr ,var-tail (list ,what))))
> > + t)
> > + cl--loop-body)))))
>
> The cl-loop macro's code lacks comments. Could you take advantage of
> "being there" to try and add comments? E.g. in the above code I see
> that depending on (eq var cl--loop-accum-var) we end up accumulating in
> the from or in the back. Could you add a comments explaining why and
> mentioning where we correct this discrepancy?
>
> > + (let ((what (pop cl--loop-args)))
> > + (cl-destructuring-bind (var) (cl--loop-handle-accum nil 'nreverse)
> > + (push `(progn
> > + (setq ,var
> > + ,(if (eq var cl--loop-accum-var)
> > + `(nconc
> > + (,(if (memq word '(nconc nconcing))
> > + #'nreverse #'reverse)
> > + ,what)
> > + ,var)
> > + `(,(if (memq word '(nconc nconcing))
> > + #'nconc #'append)
> > + ,var ,what)))
> > + t)
> > + cl--loop-body))))
>
> In the `nconc` case (when (eq var cl--loop-accum-var) is nil) we could
> also use the `var-tail` to speed up the `nconc`.
>
> Also, to avoid the N² behavior for the `append` case, maybe we
> could/should make it use `copy-sequence`, i.e.
>
> `(nconc ,var-tail (copy-sequence ,what))
>
> > -(defun cl--loop-handle-accum (def &optional func) ; uses loop-*
> > - (if (eq (car cl--loop-args) 'into)
> > +(defun cl--loop-handle-accum (def &optional func type) ; uses loop-*
> > + (setq type (or type 'list))
>
> Please add a docstring explaining whatever you managed to understand of
> this code, and describing also what this new arg `type` does.
>
>
>
[-- Attachment #2: Type: text/html, Size: 4714 bytes --]
next prev parent reply other threads:[~2018-04-08 23:29 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-04-08 0:51 Performance issue w/ `cl-loop`s `collect...into` Tianxiang Xiong
2018-04-08 3:26 ` Clément Pit-Claudel
2018-04-08 5:56 ` Tianxiang Xiong
2018-04-08 6:12 ` Clément Pit-Claudel
2018-04-08 8:50 ` Tianxiang Xiong
2018-04-08 13:19 ` Clément Pit-Claudel
2018-04-08 16:07 ` Stefan Monnier
2018-04-08 19:58 ` Tianxiang Xiong
2018-04-08 21:13 ` Stefan Monnier
2018-04-08 23:29 ` Tianxiang Xiong [this message]
2018-04-09 1:10 ` Tianxiang Xiong
2018-04-09 1:59 ` Stefan Monnier
2018-04-09 2:16 ` Tianxiang Xiong
2018-04-09 2:20 ` Stefan Monnier
2018-04-09 3:34 ` Tianxiang Xiong
2018-04-09 3:38 ` Tianxiang Xiong
2018-04-09 12:07 ` Stefan Monnier
2018-04-09 12:22 ` Basil L. Contovounesios
2018-04-09 15:28 ` Tianxiang Xiong
2018-04-09 15:33 ` Tianxiang Xiong
2018-04-14 7:01 ` Tianxiang Xiong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://www.gnu.org/software/emacs/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CACMkxiw5Eag=y8P1FwEFrJNg-TpZmFU3QyJvXHPseS4rjdTgRg@mail.gmail.com' \
--to=tianxiang.xiong@gmail.com \
--cc=emacs-devel@gnu.org \
--cc=monnier@iro.umontreal.ca \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/emacs.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).