[Opencxx-users] [opencxx] changes/occ2.5.12

[<ro...@lo...>]

Hi List,

over the last month I've worked with OpenC++. Now there is the end of my project
near and I would like to share the changes I've made on occ with you in case
someone else might find useful what I did. Please note that this patch
is for version 2.5.12, I haven't found time to tweak the source to
work with the daily snapshots and 2.5.12 fit my needs most of the time.

In all the changes I thought about backward-compatibility so even I you wrote
code which depends heavily on the occ internal object implementations this
shouldn't break anything:

- for all pure virtual functions you can remove the pure specifier with
  Member::RemovePureSpecifier(). Had to do this, since the member's declarator
  is private and I didn't want to mess around with the access specifiers.

- for all array declarations you can get a PtreeList of the array range
  expressions by calling Member::GetArrayRange (e.g. int a[3][b][4] would
  get [[3],[b],[4]]). Had to do this since the declarator is private.

- as mentioned in another mail about 1 month ago: member variables of
  type array of builtin-type weren't treated as member variables in respect
  to the various TranslateMember(Write/...) routines. I fixed that with
  that a rudimentary patch, which handles these cases as special cases, so
  you would have to define new methods (TranslateArrayMemberWrite,
  TranslatePostfixOnArrayMember, etc.) to get the corresponding hooks in
  the translation. The added member methods are:

    virtual Ptree* TranslateMemberSubscript
    virtual Ptree* TranslateArrayMemberWrite
    virtual Ptree* TranslateUnaryOnArrayMember
    virtual Ptree* TranslatePostfixOnArrayMember

- I had to set MaxNameLen in encoding.h to 1024, due to some complicated
  namespace nestings, which exceeded the 512 character limit. I personally
  would prefer to make this limit unnecessary at all but what I understand
  from the source it seems that this might mean to change code at various,
  essential places so I am not sure if I will do it.

- I added a new suffix as an indicator for C++ source (.eti, it's really an
  internal thing but I don't have the time to clean up my patch at the moment).

- I added a new command line switch (-T) which let's occ output the
  translated parse tree on stdout in addition to writing it to an .ii file.

- last, but not least: I added a method similar to CopyList to ptree-core.
  CloneList has some advantages over CopyList which I found to be very
  useful:
    1.) in contrast to CopyList CloneList creates a copy of *each* node
        in the parse tree. CopyList on the contrary just makes a copy
	of the topmost list of nodes and let's the corresponding nodes
	Car fields stay the same (at pointer equality) so you won't really
	get a totally independent list of Ptree.

	CloneList on the other hand resembles the exact structure of the
	original tree but with copies of each node (either non-leaf or
	leaf).

     2.) CloneList uses a kind of a virtual constructor which I added to
         every subclass of (and including) Ptree. It therefore is possible
	 to do something like this:

	   ClassWalker walker(env);
	   Ptree* body = Ptree::CloneList(member.FunctionBody());
	   body = walker.Translate(body);

	 where Translate will call the corresponding TranslationHooks
	 of your method class *without* relexing and parsing the created
	 parse tree, since a Ptree-subclass will stay of the same type as the
	 original nodes one. (thanks to Grzegorz Jakacki for pointing me in the
	 right directions).

I thought this might be interesting to others so I will make my patch
available here:
http://lo-res.org/~rost/patch-openc++2.5.12-changes.rost.19092002.gz

Any comments are welcome,
cu,
Robert

---
GPG: gpg --keyserver wwwkeys.pgp.net --recv-keys CCFDEB89