Consider miniutf over boost::locale

[springmeyer]

miniutf looks like it got its start for the same reason we fell back to using boost::locale on linux (#270): because libstdc++ is still missing full codecvt_utf8 / codecvt_utf16 support.

TODO:

• - benchmark to make sure miniutf is as fast as boost::locale/icu/std:: - mapnik/mapnik@0859a74

[kkaefer]

Current case conversion support from #348 is lacking:

[kkaefer]

https://bitbucket.org/alekseyt/nunicode might be a better alternative than miniutf since it implements the Unicode Special Casing.

[springmeyer]

Cool, had not seen nunicode before. Landing here again because I noticed that V8 has what look like depedency-free UpperCase and LowerCase routines that might be worth comparing/(and extracting?): https://github.com/v8/v8/blob/master/src/unicode.h#L221-L236

[kkaefer]

Sounds like a pretty good idea to do that. Likely, this has implements the same uppercase/lowercase behavior as JavaScript does, so we'd have easy parity between WebGL and native on that front.

[springmeyer]

appears blink/chromium has diff uppercase impl (not sure how it relates to v8): https://github.com/yoavweiss/Blink/blob/0e22e374087693ab579b336e0a639e13f637fed1/Source/wtf/text/StringImpl.cpp#L733-L771

[artemp]

using nunicode for language aware to upper/lower conversions #530

[springmeyer]

@artemp thanks for pushing on nunicode route.

• Are there any regression tests validating the improved case conversion?
• What do you think about my ideas above about using v8/chromium methods?

[artemp]

@springmeyer - re: blink upper/lower

• It uses ICU and it has hardcoded locale identifiers in that method - doesn't look good to me at all.

[artemp]

Content based to_upper/lower text conversions : https://github.com/mapbox/mapbox-gl-native/tree/nunicode - but we need to wait for nunicode 1.5.x release to merge into master (ETA ~ 1 dec 2014)

[kkaefer]

This is merged and we're using nunicode.