Closed GoogleCodeExporter closed 8 years ago
I've tried to apply the optimizations outlined in that article, but they
actually turn out to be slower than the
current implementation. I've attached a patch to illustrate the changes I was
using.
For 10k rows, the patched version is 20-50ms slower on my machine.
If I remove the error checking, then this version is indeed faster, but then
the parser is no longer correct.
Since the overall goal of this library is to be correct, I do not feel
comfortable sacrificing correctness for speed
in this case.
I also tried using regular expressions to break some things apart, but that
ends up being MUCH slower,
especially on larger strings. The regular expression engine doesn't seem to
perform well in my testing of
large strings.
Original comment by darron.schall
on 4 Aug 2009 at 12:57
Attachments:
Original comment by darron.schall
on 4 Aug 2009 at 12:57
In testing again, it looks like the patched version actually is faster by 100ms
or so on my 10k records test. My
mistake. I also made a slight mistake in the patch that was causing the unit
tests to fail, but have corrected that
before check-in.
These changes have been made in r95. All of the unit tests continue to pass.
However, if anyone runs into
errors, please file additional issues with reproducible test cases so I can
expand the test cases.
Original comment by darron.schall
on 4 Aug 2009 at 1:09
Original issue reported on code.google.com by
dasa.pad...@gmail.com
on 25 Feb 2009 at 1:19