Data.JSString incompatible with Data.Text for certain code points

[ali-abrar]

Data.Text performs a replacement on code points ["\55296" .. "\57343"] during certain operations (see this issue). Data.JSString, which uses JavaScript's builtin fromCodePoint (available as of ECMAScript 6) does not do this replacement.

Here's the behavior of Data.Text:

Prelude T> T.pack ['\xD800']
"\65533"
Prelude T> print '\xD800'
'\55296'
Prelude T> T.unpack (T.pack ['\xD800'])
"\65533"

As you can see, it doesn't roundtrip. By comparison, fromCodePoint and codePointAt in JavaScript do roundtrip on this value.

[ali-abrar]

This is the replacement Data.Text does: https://hackage.haskell.org/package/text-1.2.4.0/docs/Data-Text-Internal.html#v:safe

[mightybyte]

Can confirm that this is an issue that is impacting a production codebase. Would be great if we could get a fix.