Re: [Mingw-cross-env-list] TinyXML integration

[Alexandre Bacquart]

On 07/24/2010 09:55 PM, Volker Grabsch wrote:
> Alexandre Bacquart<address@hidden>  schrieb:
>    
> > After dealing with Xerces successfully, I quickly noticed a sad fact:
> > it's too damn big for what I need about XML. The static library adds
> > ~35Mo to the final executable.
> >      
> Although I generally agree, I don't think it's really 35 MB.
> Did you strip the debugging symbols before measuring?
>    

You're right. I was talking about the static library size. The added 
cost on the executable was 'only' ~20Mo.
I did not strip anything too.

> > Well, undoubtly Xerces has good reasons to be that big, but I don't need
> > a rocket launcher to kill my flies so I decided to use something
> > lighter. My first focus was on libxml2. But if libxml2 is a lot lighter
> > than Xerces (~4Mo and already integrated in MCE), I found there was
> > something even lighter than that: tinyxml (~120Ko, sure tiny in
> > comparison)
> >      
> General hint: Although Xerces is bigger, it is also more complete
> and much more bugfree than libxml2. I don't know about TinyXML,
> but be prepared to deal with an incomplete XML implementation.
>    

Well, I've dealt with it in no time and it replaced successfully my sole 
need of importing some big .dae files (Collada format) generated with 
Blender. Better than that, it simplified my importer code in many ways.

As I said, I guess there are good reasons for Xerces to be so big, but 
TBH, I don't need all the mechanisms provided by it. It's a far too big 
library for my needs here. I only need basic DOM-like parsing and had no 
problem porting my code to tinyXML.

As a final detail, I still had to use a little hack with Xerces which 
sometimes returned me text elements with randomly added whitespace 
characters (those ones "\t\n ") at the end even though texts were not 
like that in the XML file. Maybe I forgot to do something with Xerces 
inits, but with tinyXML, I do not need to do this ugly hack anymore:

            size_t wsLastNotOf = str.find_last_not_of("\n\t ") + 1;
            str.erase(wsLastNotOf);

So overall, I'm very happy with it.

--
Alex