On Mon, Jun 23, 2008 at 09:50:16AM -0700, NotFound wrote: > The chr opcode is documented in docs/ops/string.pod as: > > chr(out STR, in INT) > The character specified by codepoint integer $2 in the > current character set is returned in string $1. > > But the implementation just calls string_chr, which return a string > ascii, iso_8859_1 or utf8 depending on the codepoint value. > > If we change that, we break some code, like > examples/shotout/mandelbrot.pir, that seems to expect that the result > will always be a string with just a byte with the numeric value of his > argument for every 0-255 value. > > Maybe we need another opcode, or a way to tell the desired character > set and encoding.
I don't think there's a notion of "current character set", so we should remove that from the documentation. It should simply say "The character specified by codepoint integer $2 is returned in string $1." I think the existing behavior (returning ascii, iso8859-1, or unicode depending on the codepoint) is reasonable. Mandelbrot.pir only sends values 0-255 to C<chr>, so it always gets back a single-character string in fixed8 encoding. I think the C<chr> opcode as it currently exists is fine for most purposes. If we need a special-purpose operation to "build a fixed8 string" or "generate a codepoint in a given charset" then perhaps a method or function to do this would be more appropriate than an opcode. Pm
