Description Valentin Baciu 2008-04-23 16:29:15 EDT

A recent profiling session yielded some interesting results. It looks as if the
WSDL parser's (WSDLParser) user data map, which gets populated when instructing
the WSDL resource to track location, can get pretty big when loading lots of
WSDL documents or just a few large WSDL documents in the same resource set.

So I looked at the way WSDLParser tracks location. The map of Node -> map of
string->Integer seems intuitively to be a bit expensive memory wise. All those
user data hash maps are created with the default size/load (16 entries for
example) even though they will normally only hold maximum 4 entries. Perhaps
this could be changed to something more economical memory-wise like Node ->
some object with 4 ints. We'd avoid creating lots of HashMap#Entry and Integer
objects and spare some memory.