A rant about cross-platform programming with wchar_t | Hacker News Reader

A rant about cross-platform programming with wchar_t

losingfight.com

6 pointsBudVVeezer5 comments

ANSI C specified wchar_t in 1989, two years before Unicode 1.0. They couldn't even be sure Unicode was going to win.

Besides, the whole point of wchar_t is to not be variable width. UTF-16 in wchar_t is an abomination that dates back to the industry building APIs that take UCS-2 (which the author really ought to cover) before they realized UCS-2 was too narrow to do its job. So now we have a lot of code that appears to support Unicode but may not handle it correctly, depending on whether QA knew they should try surrogate pairs. Almost nobody realizes UTF-16 needs to be searched and spliced as carefully as UTF-8. Each is just a compression scheme for the million or so actual codepoints, and there aren't many reasons to favor one over the other (in memory, at least).

What's the actual problem here, the team made assumptions that ANSI warned against making? Apple failed to accept UCS-4 for their API?

The fact still remains that wchar_t is still a royal pain to use across platforms. Sure, it defines the size of a single character consistently within the platform. But it doesn't define what encoding that is. USC-2? USC-4? And the fact is: different compilers define it differently, which sucks in practice.

This is silly... not only is the article inaccurate but the problem described is trivial to solve. As long as you know what encoding wchar_t uses and what encoding your data is stored in this is not a big problem, use one format internally and convert your data on the way in as appropriate. Trust me, I solved it with no prior knowledge, no formal education in less than a day, as a distraction during my day job... I did not have to re-write the entire library from scratch.

Inaccurate in what way?

It may be trivial to solve, but it's certainly not an issue anyone should have to solve, if the C++ standards committee would have dictated size and encoding information along with the datatype.

the MS compiler uses wchar_t as UCS2, not UTF-16 for example

all of the C++ types have platform dependence - expecting a specific number of bits or encoding for wchar_t would make an exception of it, and one that would be impossible to implement in practice.

Reply on news.ycombinator.com