How can I save a DiagramData to a dia file in the python plug-in?

[Daniel Katz <dpkatz bc gmail com>]

Hi -

If I have a DiagramData object (but not a Diagram object) in the 
python plugin, is there a simple way to export that DiagramData 
out to a .dia file?


The context is this: I have an export filter that I've written 
using the python plug-in.  In the "normal operation" mode, 
everything is working fine and I get an export to the format in 
question.  What I was hoping to do is to add some error reporting 
such that if the input Dia diagram had a "semantic error" (in 
terms of the semantics of my diagrams), then I could put out a new 
Dia diagram which was identical to the original but had the 
offending object marked (e.g. by coloring it red).  Since this is 
all part of an export filter, I have the DiagramData through the 
interface of the begin_render() method, but I don't have a Diagram 
object.


I did try a few things:

My first thought was to get the Diagram object from 
dia.diagrams(), but that turned out to be empty when I ran the 
filter from the command line as 'dia -t foo bar.dia'.

My second thought was to create a new Diagram object via 
dia.new(), but I couldn't find a way to use my existing 
DiagramData object as the superclass instance - it doesn't seem 
like dia.new() takes a DiagramData as an argument, and I couldn't 
figure out a way to set my newDiagram.data to my existing 
DiagramData from begin_render() after the fact.

My third thought was to try to call the existing export filter for 
dia files directly, but I couldn't figure out how to get a filter 
instance in python.  (I saw that in bindings/test.py there were 
references to dia.filter_guess_export_filter()', but when I tried 
to invoke that function through the Python Console dialog I got a 
missing attribute error - not part of the python API, I guess.)

My fourth thought was to bypass the passed in DiagramData object 
and instead use dia.load() to load a copy of the original input 
diagram directly.  This works (since begin_render() also gives me 
the output filename and in my particular use case I can infer the 
full path of the input file from the output filename and from 
there make a copy, dia.load() it, modify object colors, save it, 
and so on), but it feels sort of roundabout and the input-file 
finding is fragile.



So, any advice on a correct and robust way to do this?

Thanks.

Dan


P.S.  This was all attempted with Dia 0.97.1 and also with the 
latest from the git repository (with python and swig included as 
described at 
http://dia-installer.de/howto/compile_msvc/index.html) on Ubuntu 
Linux 11.10.