[Geany] remembering folding status

[daspostloch at googlemail.com]

i got a statement from neil on the state of folding remembrance
in scintilla and wanted to share:

> There is code in SciTE in the FoldState class in
> scite/src/SciTEBase.h for storing folding of buffers that aren't
> shown. There is also code in scite/src/SciTEBuffers.cxx for saving and
> loading this from file: search for "session.folds". SciTE only allows
> a single view on a file.
>
 >  Neil

wont be looking into this for the next 3 weeks, but thanks to
everyone for laying out the roadmap :)
- matt

On 9/21/09, Thomas Martitiz <s0523936 at htw-berlin.de> wrote:
> Am 18.09.2009 14:17, schrieb Lex Trotman:
>> Hi,
>>
>> Another 1.5c.  It was more fun to think about this than what I'm
>> supposed to be doing, especially as g++ just gave me a 75 line error
>> message (and all it wanted to say was I had an extra const that I
>> shouldn't have)
>>
>>  From past experience in another system, I think the problems that Neil
>> is worried about are that folds are on a per line basis (confirmed by
>> looking at the interface), but of course any saved fold data is
>> garbage if the lines in the file have changed.
>>
>> As Neil points out thats hard enough when only worrying about
>> in-memory views, its really difficult to ensure when relating saved
>> fold lines to an external file.  You must ensure that the file hasn't
>> changed in a way that invalidates the fold data.
>>
>> The fold data therefore needs to save a suitable signature of the file
>> to decide on its validity, I guess mtime is the minimum
> This breaks on just resaving/touching the file.
>> or some MD5 or
>> SHA signature on the contents.
>>
> That means external dependencies.
>
>
> Really, the easiest and simplest way I see is to store the line number
> if the folding mark and the content line. Then iterate through all line
> numbers, and strcmp() the content with the dumped one. If it doesn't
> match -> don't fold. That sounds pretty save to me-
>
>> And there needs to be a way of finding the fold data for a particular
>> file, maybe same pathname with an added extension  would do, (eg
>> /home/fred/fred.cpp.fd~ ,the ~ so it would be ignored by VCS).  That
>> is if you don't mind such files littering your src directories.  If
>> its not in the same directory it gets a little bit harder still.
>>
>>
>
> I think a single file would be most easy to handle.
>
>
>
> _______________________________________________
> Geany mailing list
> Geany at uvena.de
> http://lists.uvena.de/cgi-bin/mailman/listinfo/geany
>