On Thursday, January 30, 2003, at 09:47 PM, Ryan Hoegg wrote:
There's a new comment added to that bug, to the effect that Base64.decode is broken. I may have a fix for that, and a considerable encode speedup coming later today --- I'll post them to buzilla. I also think that a trailing newline should be added to encoded data (the perl Base64 module does this). I'll include this in the patch as well.Daniel Rall wrote:Not yet. I think the issue is basically closed except for some uncertainty on performance with large inputs.p.s. Did the base-64 fixes make it into both 1.2 branch _and_ HEAD?
I was going to get around to making a test for that this weekend. I found a tasty Project Gutenberg text for it :)My tests with large inputs showed consistently better performance for the patched code (by up to 50%), which should be pretty robust --- it might not be worth cluttering up the tests with another test.
I've got some code I can send you (that reads the large text from a file), that I used for my speed tests, in case you still want to add it.
Also, I think our changes should go upstream to Codec, and we should change the package name on Base64.java to org.apache.commons.codec. You're a committer on Codec, so you could make it happen.
*yes*, the Base64 code definitely wants to be consolidated.
