[
Lists Home |
Date Index |
Thread Index
]
I'm really sympathetic to the calls for counted rather than
null-terminated strings, if only for the genxText() call.
So I was thinking about this back before we nuked the codePoint *
versions, and I realized that a "length" argument could be confusing
because in the codePoint* version it would naturally be the number of
characters, while in the utf8Byte * version it would naturally be the
number of bytes. Blecch. So I thought it would be more natural to
have something like
genxText(genxWriter w, utf8Byte * start, utf8Byte * end)
i.e. a pointer to the end of the string, which would have the same
semantics in both versions of the call. Well, we're losing the
codePoint * stuff (good riddance) but I'd kind of like to stay with the
stop-here argument rather than the byte (or character) count argument.
Of course if you want to null-terminate, you can, just do
genxText(w, buf, NULL)
Two questions:
- if you have a zero byte in the string before you get to the end mark,
should it just stop, or throw an error? The first is more consistent
with C culture (cf strncpy) but the latter a bit more stringent.
Moderately leaning to just stopping.
- if the stop marker is stupidly in the middle of a UTF-8 character,
genx should detect this and declare an error. The existence of this
situation is the only good argument for a count rather than a stopper.
But not quite good enough. -Tim
|
