Community
Participate
Working Groups
Our p2 metadata contains lots and lots of strings: ids of IUs, required/provided capabilities, property keys/values, touchpoint types, etc, etc. Simply uniquifying strings while parsing our data files would save a lot of memory.
I have released a simple canonicalization of strings in XMLParser. I did a quick benchmark of the admin UI application with a single metadata repository, and a profile registry with two profiles, each containing the Eclipse SDK. Before the change, there were 8,909,576 bytes worth of strings (transitive size), and after the change there were 5,081,024 bytes of strings. There is lots more room for optimization, but this was an easy first start (only a couple of lines of code).