[Expat-discuss] FW: OverExecuting characterDataHandler()

[Karl Waclawek]

James Renton wrote:
> I am using expat to quickly extract and parse an audio stream encoded in
> base64.  This way I can include all kinds of useful meta-data as
> attributes of each chunk.
>   

You mean the stream is XML, and the chunks are base64? I would be 
surprised of Expat parsed base64 directly.
> The source of my audio produces 8000 – 32000 bytes per second.  I just
> produced a working prototype and noticed that calls to my
> characterDataHandler() (set by using XML_SetCharacterDataHandler()) are
> occurring as follows:
> 1.	All bytes up to but excluding the first/next carriage return (in my
> case 80)
> 2.	Carriage return (single byte)
> 3.	Repeat (1)
>
> I am pushing this data off-process and streaming it to various endpoints
> using sockets (not decoding).  This is producing a VERY LARGE quantity of
> function calls to my handler which is not optimum.  It would be much
> better if at least 1-15 kilobytes of data were passed to the
> characterDataHandler  at a time rather.  My questions:
>    1. How exactly does the character data parsing in expat work?  Some
> insight in this regard might answer all of my questions.
>    2. Is the behavior of the libraries calls to the handler configurable
> in any way (possibly using #define’s, custom builds etc…)?
>    3. Is there any other way to decrease the number of function calls in
> this case?
>
>   
The call-back logic would be hard to change, but you can always 
accumulate character data in a buffer and make your other calls based on 
having accumulated enough data to warrant a network call.

Karl