genUnicodeString generates invalid unicode
martyall opened this issue · 4 comments
The unicode character generator for unicode characters is picking a random CodePoint in the BMP. The unicode string generator just generates an arbitrary array of such code points and turns it into a string. It turns out that this can generate invalid unicode via unpaired surrogates: https://unicode.org/faq/utf_bom.html#utf16-7
One solution here would be to restrict the code points to avoid such cases, another would be to figure out a more complicated but correct way to generate unicode which cannot be done CodePoint by CodePoint.
For context I discovered this while trying to write a quickcheck test for utf8 encoding/decoding, you can see the failing test here
On a related note, it appears that the unicode Char generator includes the code point 65536 (chooseInt is inlcusive?). Shouldn't this be 65535? (FYI, I've verified that this is not causing my error)
On a related note, it appears that the unicode Char generator includes the code point 65536 (chooseInt is inlcusive?). Shouldn't this be 65535? (FYI, I've verified that this is not causing my error)
It's wrong as a bound, but 65536 is just getting turned into 65535 via toEnumWithDefaults, which is effectively clamp.
Does something like this seem right?
At least it passes the utf-8 round trip
Seems reasonable to me 👍