The PagingMemoryProxy implementation has a couple of performance issues
The current implementation will parse the records on every load attempt
(every new page), and resort the full result set (if sorting is defined).
If sorting is not defined, then sorting is skipped (which is considerably
faster for a few hundred records or more), otherwise it severely degrades
with the size of the results.
Would it be possible to
a) cache parsed records after the initial load? Since the contents are
fixed, maybe even give it a constructor that takes records?
b) remember the last sort (column + direction) and avoid resorting if
unchanged?
This will significantly speed up paging, and ensure that paging is a
constant time operation for a given resultset regardless of whether/which
column is sorted.
Original issue reported on code.google.com by froc...@gmail.com on 30 Nov 2007 at 10:18
Original issue reported on code.google.com by
froc...@gmail.com
on 30 Nov 2007 at 10:18