You can clone with
HTTPS or Subversion.
Update benchmark results.
With some undesirable backtracking eliminated in attoparsec 0.9.0.0, we
see a nice little bump in parsing performance: between 5% and 12%.
This has also had the nice effect of reducing memory use by about 15%.
Tighten up the top-level json parser to be safer
An obvious simplification that potentially lets us avoid backtracking
Added tag 0.3.2.6 for changeset 9311d533b5d0
Drop a deleted file
Bump dependency on hashable
Bump version, and performance-dependent version of attoparsec needed
Update 32-bit GHC 6.12.3 performance
Add benchmark running script
Make Value an instance of Hashable.
Enable RTS options for GHC >= 7
Added tag 0.3.2.5 for changeset 21e7a84b5762
Port a bunch of code over to blaze-textual for reuse.
Added tag 0.3.2.4 for changeset c50d0f85ca2a
Build qc with threaded runtime
Fix up compilation warnings.
Make the roundtrip test of Double approximate.
Ignore the quickcheck test
Bump dependency on attoparsec.
Ensure that we parse null as NaN when expecting a floating point number.
RFC 4627 section 2.4 forbids encoding of NaN and +/- Infinity.
Remove some incorrect hashes from .hgtags.
Added tag 0.3.2.3 for changeset 14f3d533dbbb
Parse and serialize .NET System.DateTime values correctly.
The number in parentheses in the default JSON serialization of a a .NET
DateTime is milliseconds since the Unix epoch, whereas I had been treating
it as seconds.
Fix a compilation error with GHC HEAD.
Added tag 0.3.2.2 for changeset c3a1a4557a29
Switch from monads-fd to mtl, and bump version
Ignore cabal-dev sandboxes
Drop use of bytestring-show
Make Result an instance of NFData