Both double readers (bytestring-lexing and Data.Text.Read.double) are
approximate: 81% of a 200k show/read sample comes back 1-7 ULP off,
subnormals parse as 0.0, and the largest finite double parses as
Infinity.
An exact byte-level conversion with a Clinger fast path fixes all
three, but this is the parsing hot path (#214 review), so it'll come as
its own PR with benchmarks.
Both double readers (bytestring-lexing and Data.Text.Read.double) are
approximate: 81% of a 200k show/read sample comes back 1-7 ULP off,
subnormals parse as 0.0, and the largest finite double parses as
Infinity.
An exact byte-level conversion with a Clinger fast path fixes all
three, but this is the parsing hot path (#214 review), so it'll come as
its own PR with benchmarks.