Am 05.10.2011 um 10:24 schrieb Jean-Marc Lasgouttes:

> Le 04/10/2011 22:22, Stephan Witt a écrit :
>> Am 30.09.2011 um 13:13 schrieb LyX Ticket Tracker:
>> 
>>> #7789: Word count will count words in deleted notes (with change tracking)
>>> ---------------------+------------------------------------------------------
>>> Reporter:  bhelm    |       Owner:  rgheck
>>>     Type:  defect   |      Status:  new
>>> Priority:  normal   |   Milestone:
>>> Component:  changes  |     Version:  2.0.Xsvn
>>> Severity:  normal   |    Keywords:  patch
>>> ---------------------+------------------------------------------------------
>>> Changes (by stwitt):
>>> 
>>> * cc: stwitt (added)
>>>  * keywords:  =>  patch
>>> --
>>> Ticket URL:<http://www.lyx.org/trac/ticket/7789#comment:1>
>> 
>> Does anyone has any objection? I'd like to apply the patch.
> 
> The patch looks reasonable, provided that it works. Did you test it by hand 
> for the different cases?

Yes. I did test it with different cases.
Bennett had the patch applied already and tested it too without negative side 
effects.

> I would get rid of the private wordCount() and charCount() methods that are 
> too trivial and have only one user.

I wanted to make the statistics status variables private and therefor I had to 
make the accessor functions.
Note: there are inline code.

> Actually, I do not know why this stuff goes to Buffer, and even less 
> Buffer::Private.

We had a discussion already about the merge of the different count functions.
I think these values belong to the Buffer. And to keep the Buffer constness
I put it into Buffer::Private. 

> Is performance correct on very large files?

The performance is improved - although I didn't profile it.
The former implementation did iterate the document 3 times.
Now there is only one iteration. I cannot imagine that's not faster.

> Otherwise, such functionality could be implemented (like updateLabels) by a 
> Inset::updateStatistics. BUt this is a different story.

I thought about that for a short time and decided to postpone this. 
I don't believe the statistics have to be correct all the time.
It's ok to compute them when needed.

Stephan

Reply via email to