ToEncodedUrlPiece

[fizruk]

This PR is based on #47 and adds efficient implementations for toEncodedUrlPiece together with some tests.

@phadej could you take a look?

[fizruk]

We could also implement more efficient versions of toEncodedUrlPiece for Maybe and Either... but I wonder who uses those anyway.

[phadej]

Seems fine. Have you benchmarked (or do you real world example) that this really matters?

[fizruk]

I did not perform a proper benchmarking, but it should matter, considering that URL-encoding demands per-char processing. A quick test in GHCi reveals at least some difference:

>>> :set +s
>>> import qualified Data.ByteString.Lazy as BSL
>>> import Data.ByteString.Builder (toLazyByteString)
>>> let check f = sum . map (BSL.length . toLazyByteString . f)
>>> check (toEncodedUrlPiece . toUrlPiece) [1..10000000]
68888897
(22.59 secs, 72,137,670,104 bytes)
>>> check toEncodedUrlPiece [1..10000000]
68888897
(15.44 secs, 55,719,891,472 bytes)

OTOH, now that I think of it,toEncodedUrlPiece only happens on clients, so it probably isn't much of a performance concern.

[phadej]

LGTM