Export SDL_UCS4ToUTF8()? about sdl HOT 4 CLOSED

Pros:

• much less heavyweight than SDL_iconv
• people rolling their own will either get it wrong, or get it really wrong by casting to char.

Cons:

...?

from sdl.

Dumb questions:

• What should this do with too-large values? A value > 0x10FFFF not only isn't a valid Unicode codepoint, it also won't fit in a UTF-8 4-byte sequence. The current code just generates a 4-byte sequence with apparently-random bits.
• Should it fail if they try to encode a UTF-16 surrogate codepoint? I'd say yes, but right now it does not.

I'm thinking in both cases this should either encode an SDL_INVALID_UNICODE_CODEPOINT instead (which is the Unicode REPLACEMENT CHARACTER: that weird question mark diamond shape thing we've all seen at some point), or report an error in some way, or both.

from sdl.

Dumb questions:

• What should this do with too-large values? A value > 0x10FFFF not only isn't a valid Unicode codepoint, it also won't fit in a UTF-8 4-byte sequence. The current code just generates a 4-byte sequence with apparently-random bits.
• Should it fail if they try to encode a UTF-16 surrogate codepoint? I'd say yes, but right now it does not.

I'm thinking in both cases this should either encode an SDL_INVALID_UNICODE_CODEPOINT instead (which is the Unicode REPLACEMENT CHARACTER: that weird question mark diamond shape thing we've all seen at some point), or report an error in some way, or both.

Agreed.

from sdl.

I'm thinking in both cases this should either encode an SDL_INVALID_UNICODE_CODEPOINT instead (which is the Unicode REPLACEMENT CHARACTER: that weird question mark diamond shape thing we've all seen at some point), or report an error in some way, or both.

Went with the first one; the pattern char *end = SDL_UCS4ToUTF8(ch, an_array_not_a_pointer); *end = 0; was the primary way we were using this, so reporting an error would be awkward and probably unnecessary.

from sdl.