> Sure but now we're talking about a universal problem across languages, rather than a C-specific problem.
Of course, but that's my point - C already gives you the ability to fix the incorrect typing problem, using the existing foundational `str*` functions.
A team who is not using the compiler's ability to warn when mixing types are still going to mix types when there is a safe strbuf_t type.
The problem with the `str*` functions can be fixed today without modifying the language or it's stdlib.
Most C programmers don't do it (myself included). I think that, in one sense, you are correct in that removing the existing string representation (and functions for them) and replacing them with len+data representation for strings will fix some problems.
Trouble is, a lot of useful tokenising/parsing/etc string problems are not possible in a len+data representation (each strtok() type function, for example, needs to make a copy of what it returns) so programmers are just going to do their best to bypass them.
Having programmers trained to create new string types using existing C is just easier, because then you solve the whole 'mixing types' problem even when looking at replacements for things like `strtok`.
Or ... maybe I'm completely off-base and the reason that programmers don't create different types for string-stored data is because it is too much work in current C-as-we-know-it.
Of course, but that's my point - C already gives you the ability to fix the incorrect typing problem, using the existing foundational `str*` functions.
A team who is not using the compiler's ability to warn when mixing types are still going to mix types when there is a safe strbuf_t type.
The problem with the `str*` functions can be fixed today without modifying the language or it's stdlib.
Most C programmers don't do it (myself included). I think that, in one sense, you are correct in that removing the existing string representation (and functions for them) and replacing them with len+data representation for strings will fix some problems.
Trouble is, a lot of useful tokenising/parsing/etc string problems are not possible in a len+data representation (each strtok() type function, for example, needs to make a copy of what it returns) so programmers are just going to do their best to bypass them.
Having programmers trained to create new string types using existing C is just easier, because then you solve the whole 'mixing types' problem even when looking at replacements for things like `strtok`.
Or ... maybe I'm completely off-base and the reason that programmers don't create different types for string-stored data is because it is too much work in current C-as-we-know-it.