[ jEdit-devel ] [ jedit-Patches-1999448 ] More efficient and general fix for black hole bug

[SourceForge.net <no...@so...>]

Patches item #1999448, was opened at 2008-06-21 07:08
Message generated for change (Comment added) made by ezust
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=300588&aid=1999448&group_id=588

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Kazutoshi Satoda (k_satoda)
Assigned to: Matthieu Casanova (kpouer)
Summary: More efficient and general fix for black hole bug

Initial Comment:
I found joining lines is very slow only when folding
mode is set to other than 'none'.

Reproduction recipe:
  (WARNING: This will cause vary long hang up.)
  - Open doc/CHANGES.txt in jEdit source tree.
  - Do [Expand All Folds] (C+e x).
    (without this, another bug happens.)
  - Do [Select All] (C+a).
  - Try [Join Lines] (C+j).

This patch will speed up this situation by minimizing
calls to buffer.getFoldLevel().

Also, this patch will expand the care about black hole
bug to support ...
  - multi line edit including folded line
  - more folding modes other than 'explicit'
  - modification at the middle of folded range.

I want approval from Matthieu who made the current
fixes for the bugs.

----------------------------------------------------------------------

>Comment By: Alan Ezust (ezust)
Date: 2008-08-11 13:54

Message:
Logged In: YES 
user_id=935841
Originator: NO

if optimized join_line.patch was applied to trunk, could the person who
applied it please remove the attachment from this tracker item?

It fails to apply now, in any case.


----------------------------------------------------------------------

Comment By: Kazutoshi Satoda (k_satoda)
Date: 2008-08-02 12:29

Message:
Logged In: YES 
user_id=1483238
Originator: YES

That's a greate idea. Applied in r13153. The speed problem has gone.
Thanks.

However, this patch still has many fixes about black hole bugs. I still
think this should be applied. What is your evaluation?


----------------------------------------------------------------------

Comment By: Matthieu Casanova (kpouer)
Date: 2008-07-31 14:59

Message:
Logged In: YES 
user_id=285591
Originator: NO

Hi, I analyzed what was slow in the joinline method and found this :
When removing a line from the buffer, the
LineManager.setFirstInvalidFoldLevel() method is called for the following
lines.
And when asking if the line is the foldStart, the foldLevels must be
calculated if they are invalid.

When joining lines, we will start from first line of the selection and
join with the next line until the selection is 1 line length. And each time
the line is invalidated and will require to recalculate fold levels..
So I inverted the method to join lines from last to first, doing this, we
don't need to recalculate fold levels everytime.
It seems far faster now, I attached a patch, please try it and tell me
what you think

File Added: optimized_join_line.patch

----------------------------------------------------------------------

Comment By: Kazutoshi Satoda (k_satoda)
Date: 2008-07-05 06:08

Message:
Logged In: YES 
user_id=1483238
Originator: YES

Thank you for trying the patch.

Yes, it is a problem. But I think it is a small problem which I can
recover by folding it again after editing the line, and the black hole
bugs is big problem which I can't recover it without reloading. I think
it is fair to trade them. How do you think?

Expanding with actual change of fold level should implemented for
all fold handlers, not for only 'explicit'. ... by adding a method
boolean mayChangeFoldLevel(<information of modification>) to the
FoldHandler which returns true by default ... ?

----------------------------------------------------------------------

Comment By: Matthieu Casanova (kpouer)
Date: 2008-06-25 01:08

Message:
Logged In: YES 
user_id=285591
Originator: NO

Hi, I just tested the patch but there is a problem :

if I use explicit fold, with this buffer

{{{ hello

something

}}}

all folds are folded. I remove one "l" from hello. I think it should not
expand folds since the fold level is not changed

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=300588&aid=1999448&group_id=588