[Mulgara-general] ARP vs RIO parser (was Re: Mulgara size limits)

[Alex Hall]

Edwin Shin wrote:
> On 07/02/2008 01:18 AM, Paul Gearon is rumored to have said:
>> Alex's comment about RIO reminds me.... wasn't this parser supposed to
>> replace the ARP-based one?
> 
> This reminded me that there's a "HACK!!!!!" noted in the build targets
> to "get the ARP parser working". When bundling Mulgara with Fedora, I
> noticed some classloader-related problems that had to do with some of
> the classes that are being included as part of the this hack. I remember
> I had to fuss with the load order to work around it.

Ahh yes, I am familiar with this hack.  It predates my involvement with 
the project so I'm not sure why it was necessary.  It's been on my 
wish-list of things to fix for a while, albeit not very high on that 
list.  The hack also predates Java 1.5, when Xerces was added to the 
JDK, so I'm hoping that it isn't even necessary any more; however, I'm 
not familiar enough with the various API's for XML in Java to know for 
sure.  Regardless, if we move ahead with the build system refactoring 
that Paul has proposed, this would be a good candidate for investigation.

> It sounded as though this bit of the build file had become part of a
> forgotten history. Is there indeed a plan to switch the parser, and if
> so, for 2.0?

The entire ARP vs. RIO parser issue seems to have become a part of 
forgotten history.  I seem to remember hearing that the RIO parser was 
supposed to have better performance, but I would prefer to hear some 
verification of this before proceeding with the switch.  At any rate, 
the "switch" will just be a change in the config file; Mulgara ships 
with both parsers included and will probably continue to do so unless 
somebody comes up with a good reason to remove the ARP parser altogether.

Regards,
Alex