changes:
1.0.0.1
- reChunkIn: Fix insufficient laziness
The redundant case for [y] was always forcing one more input chunk to
be evaluated than necessary to generate an output chunk.
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
- typo of "decoding"
- Simplify reChunkWith
- Add lazy benchmarks
- Improve encoding performance by 5%
- Improve decodeLenient performance by 50%
1.0.0.0
- Refactor the testsuite a little
- Add .Lazy variants of the two Base64 modules
0.1.2.0
- Ensure that joinWith really does what it claims.
0.1.1.3
- Bump version constraint on bytestring
- trustworthy flags; fix BM
0.1.1.1
- Fix performance regression in encoding.
The encoding table wasn't being let-floated when URL encoding support was
added, and so it was being regenerated on every call, with catastrophic
performance consequences (600x slowdown for small inputs on my machine).