From: Stefan Monnier <monnier@iro.umontreal.ca>
To: Joakim Jalap <joakim.jalap@fastmail.com>
Cc: emacs-devel@gnu.org
Subject: Re: Overlays as an AA-tree
Date: Thu, 22 Sep 2016 08:17:40 -0400 [thread overview]
Message-ID: <jwvponwdtxk.fsf-monnier+Inbox@gnu.org> (raw)
In-Reply-To: <87a8f0p69w.fsf@fastmail.com> (Joakim Jalap's message of "Thu, 22 Sep 2016 12:35:29 +0200")
>> which makes me suspect your design might not be quite right.
> Well, I didn't think *too* hard about it. I didn't really think about
> what to do when the intervals of two overlays are the same. Does it even
> matter?
Not much, no. Do note that in compare_overlays, we do need a total ordering
(this is used for sort_overlays). Your sorting and that of
compare_overlays can't be identical (compare_overlays has to obey the
`priority` property, whereas yours shouldn't pay attention to it), but
you can still use the same idea: i.e. when start and end are equal, just
use the overlays's memory addresses to decide which comes first.
>> This said, reading overlay_tree_insert_internal, I have the impression
>> that you're implementing what wikipedia describes as an "Augmented tree"
>> where the basic tree is your AA-tree, indexed by the left position of
>> the overlays, with the `max` field being the "augmented" data, which
>> sounds just fine.
>> Is that right?
> That is exactly right.
Great.
>> The only places you absolutely need to replace are all the places that
>> need to modify the tree. There shouldn't be that many of them
>> (basically, make-overlay, move-overlay, delete-overlay, plus text
>> insertion/deletion).
>> Then the rest can be modified little by little.
> I think I ran into some other subtle things. For example an overlay can
> be in "no buffer" if both it's markers were detached from their buffer,
> I think.
The user doesn't have direct access to the markers, so IIUC this
situation happens through delete-overlay or move-overlay.
> So in the case of the overlay tree, I need to detect this and
> remove the overlay from the buffers tree.
Similarly, move-overlay can necessitate moving the overlay from one tree
to another.
>> Some tricky parts:
>> - because of the insertion_type, two overlays may start with end1<end2
>> and finish with end1>end2 without any call to move-overlay.
> Hm, ok. I will have to look into this further.
I think a more important aspect is that insertion/deletion of text has
to update all char-positions of overlays after the point of
insertion/deletion. I.e. it's an O(n) operation. In the intervals tree
used for text-properties we avoid this by keeping track of distances
rather than positions. and then positions get (re)computed when needed.
>> - the overlay tree needs to be "weak" (i.e. we'll need to add special
>> code in the GC).
> I don't really understand what this means. Could you please elaborate?
> I was hoping it would work with the current marking and sweeping.
Hmm... forget it, I was confused.
>> I wouldn't worry about merging (better yet: merge from master right away
>> and then keep doing that on a regular basis, which should be easy since
>> I don't think we've touched (nor will touch) this code very much).
> I tried to do that yesterday. There were some conflicts in insdel.c
> (which I think I fixed) but now I can't commit, because 'git commit'
> complains about trailing whitespace in some regex test files. How do I
> get around this?
Just remove the trailing whitespace.
Stefan
next prev parent reply other threads:[~2016-09-22 12:17 UTC|newest]
Thread overview: 110+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-09-20 10:32 Overlays as an AA-tree Joakim Jalap
2016-09-20 12:14 ` Clément Pit--Claudel
2016-09-20 12:43 ` Lars Ingebrigtsen
2016-09-20 16:19 ` Joakim Jalap
2016-09-20 23:19 ` Richard Stallman
2016-09-20 14:40 ` Eli Zaretskii
2016-09-20 17:13 ` Joakim Jalap
2016-09-21 16:14 ` Eli Zaretskii
2016-09-22 10:35 ` Joakim Jalap
2016-09-22 12:24 ` Stefan Monnier
2016-09-21 14:52 ` Stefan Monnier
2016-09-21 15:58 ` Eli Zaretskii
2016-09-21 16:24 ` Stefan Monnier
2016-09-21 16:43 ` Eli Zaretskii
2016-09-21 18:41 ` Stefan Monnier
2016-09-21 19:28 ` Eli Zaretskii
2016-09-22 10:35 ` Joakim Jalap
2016-09-22 12:17 ` Stefan Monnier [this message]
2016-09-22 20:11 ` Joakim Jalap
2016-09-23 1:29 ` Stefan Monnier
2016-09-27 6:26 ` Joakim Jalap
2016-09-27 11:50 ` Stefan Monnier
2016-09-27 14:38 ` Eli Zaretskii
2016-09-27 16:07 ` Joakim Jalap
2016-11-21 17:32 ` Clément Pit--Claudel
2016-11-22 8:09 ` Joakim Jalap
2016-11-22 13:44 ` Stefan Monnier
2016-11-23 6:58 ` Joakim Jalap
2016-11-22 13:44 ` Clément Pit--Claudel
2016-11-22 13:55 ` Evgeny Roubinchtein
2016-11-23 7:16 ` Joakim Jalap
2016-11-23 15:42 ` Eli Zaretskii
2016-11-23 16:06 ` Stefan Monnier
2016-11-24 18:33 ` Evgeny Roubinchtein
2016-11-23 7:13 ` Joakim Jalap
2017-02-03 8:49 ` Andreas Politz
2017-02-03 9:13 ` Eli Zaretskii
2017-02-03 10:24 ` Andreas Politz
2017-02-03 11:33 ` Joakim Jalap
2017-02-03 12:44 ` Andreas Politz
2017-02-03 14:11 ` Joakim Jalap
2017-02-03 15:02 ` Andreas Politz
2017-02-03 15:23 ` Joakim Jalap
2017-02-03 15:54 ` Andreas Politz
2017-02-04 21:22 ` Stefan Monnier
2017-02-04 23:10 ` Andreas Politz
2017-02-06 9:56 ` Joakim Jalap
2017-02-06 11:33 ` Andreas Politz
2017-02-06 15:33 ` Joakim Jalap
2017-02-06 15:46 ` Stefan Monnier
2017-02-06 16:52 ` Joakim Jalap
2017-02-06 17:34 ` Stefan Monnier
2017-02-06 17:32 ` Andreas Politz
2017-02-07 12:35 ` Andreas Politz
2017-02-07 14:46 ` Joakim Jalap
2017-02-07 22:00 ` Andreas Politz
2017-02-08 0:36 ` Andreas Politz
2017-02-08 6:53 ` Joakim Jalap
2017-02-08 7:34 ` Andreas Politz
2017-02-08 8:18 ` Joakim Jalap
2017-02-08 8:44 ` Andreas Politz
2017-02-08 16:34 ` Andreas Politz
2017-02-08 17:38 ` Eli Zaretskii
2017-02-08 13:04 ` Stefan Monnier
2017-02-08 22:27 ` Andreas Politz
2017-02-08 22:34 ` Stefan Monnier
2017-02-09 9:34 ` Andreas Politz
2017-02-09 10:05 ` Joakim Jalap
2017-02-09 10:19 ` Andreas Politz
2017-02-13 3:44 ` Andreas Politz
2017-02-13 6:11 ` Eli Zaretskii
2017-02-13 9:04 ` Andreas Politz
2017-02-13 14:36 ` Eli Zaretskii
2017-02-14 10:07 ` Andreas Politz
2017-02-17 4:58 ` Andreas Politz
2017-02-17 7:12 ` Eli Zaretskii
2017-02-19 22:39 ` Andreas Politz
2017-02-19 23:10 ` Stefan Monnier
2017-02-19 23:44 ` Andreas Politz
2017-02-20 15:34 ` Eli Zaretskii
2017-02-21 6:56 ` Andreas Politz
2017-02-21 15:11 ` Stefan Monnier
2017-02-21 18:26 ` Andreas Politz
2017-02-21 19:18 ` Stefan Monnier
2017-02-21 23:36 ` Andreas Politz
2017-02-24 8:43 ` Andreas Politz
2017-04-08 13:28 ` Clément Pit-Claudel
2017-05-03 19:20 ` Andreas Politz
2017-05-03 19:40 ` Stefan Monnier
2017-05-05 20:39 ` Andreas Politz
2017-05-04 0:54 ` Clément Pit-Claudel
2017-05-05 20:10 ` Andreas Politz
2017-05-05 22:22 ` Clément Pit-Claudel
2017-05-06 8:05 ` Andreas Politz
2017-05-04 14:21 ` Eli Zaretskii
2017-05-05 20:08 ` Andreas Politz
2017-05-05 21:41 ` Eli Zaretskii
2017-09-24 13:09 ` Clément Pit-Claudel
2017-10-04 8:17 ` Andreas Politz
2017-10-04 9:22 ` Eli Zaretskii
2017-10-04 20:36 ` Andreas Politz
2017-02-14 0:45 ` Richard Stallman
2017-02-14 8:32 ` Andreas Politz
2017-02-06 13:51 ` Stefan Monnier
2017-02-06 14:26 ` Andreas Politz
2017-02-06 15:06 ` Stefan Monnier
2017-02-06 15:40 ` Joakim Jalap
2017-02-06 16:24 ` Clément Pit-Claudel
2017-02-03 13:55 ` Stefan Monnier
2017-02-03 15:14 ` Andreas Politz
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=jwvponwdtxk.fsf-monnier+Inbox@gnu.org \
--to=monnier@iro.umontreal.ca \
--cc=emacs-devel@gnu.org \
--cc=joakim.jalap@fastmail.com \
/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).