[Geany-devel] Symbol tree discussion

[Nick Treleaven]

On Sat, 4 Oct 2008 11:00:13 -0700
"Conrad Steenberg" <conrad.steenberg at caltech.edu> wrote:

> >1. Replace the GeanySymbol list with a TMTag list. This avoids
> >having to parse the parent name, avoiding possible bugs.
> 
> I'm not exactly sure what you mean - I thought that parsing the symbol
> names was the only way to determine symbol tree structure?

I should have been clearer. The patch 'parses' the parent
name in the GeanySymbol::str field e.g. "Parent" from "Parent::Child".
Line 458 of symbols.c shows:

symbol->str = g_strconcat(tag->atts.entry.scope, cosep, utf8_name,
	NULL);

So it would be cleaner and safer to remove the old GeanySymbol code and
let the tree creation code read tag->atts.entry.scope instead.

...
> >2. Sort the tags by appearance before adding to the tree model. This
> >ensures that parent items are added before children, and we can
> >remove all the complex 'skipped' code. The model can then be sorted
> >after all entries have been added (for the 'sort by name' case).
> 
> Yes, presorting sounds like a good idea to avoid modifying the list of
> symbols. FWIW the current logic is to move a symbol to the end of the
> list if its parent cannot be found.
> 
> Note that the 'skipped' list is actually not used in the current
> version of the patch - it was there to catch errors caused by a bug in
> the the C++ tag generator (nested namespaces, see bug "[ 2141317 ] C++
> nested namespace tags created incorrectly")

OK. Possibly that bug is already fixed in ctags, haven't checked yet.

> >3. Instead of looking for children for certain tags (e.g. class,
> >etc), just see if the TMTag struct has a parent field. This should
> >be much more flexible for filetypes that overload e.g. the
> >tm_tag_class_t fields to mean something else.

BTW I meant the 'scope' field, there's no parent field.

> >I'm happy to work on these changes over time. If you want to discuss
> >them, the geany-devel list might be better than here.
> 
> Of course you're welcome to do that - I'm simply happy to have the
> feature since it really helps navigating complex code.
> 
> Thanks again for your and all the other geany developers' work :-)

Thanks for writing the patch ;-) It's great for the filetypes I use
such as C/C++. I'll work on these changes in the branch [1] and merge
them to the trunk after 0.15 is released.

Regards,
Nick

[1]:
https://geany.svn.sourceforge.net/svnroot/geany/branches/symbol-tree