Thread: Static functions for data hiding.

Hi everyone, just curious. When creating C files that interact with other C files, normally this is managed through the .h file(s).

By excluding function declaration prototypes from the .h file(s) they can't be accessed/used outside of that file. (or as far as I am aware). So, my question is, why use static to mean private to this source file? isn't it overkill?

The only possible reason (might be) so that it is 'clear' when reading a source file which functions won't have a declaration prototype in the .h file.

Am I missing something important?

You also avoid name space pollution.

Consider, for example, a project where you have several .c files that have a helper function named process() they use internally. If they are not declared as static, it will create a link time conflict; the linker won't know (but will probably try to guess, with varying results) which process() is being called where. (The header files are not examined at link time, so it does not matter whether the functions are declared in the header file or not; only whether they're static/local or not.)

This is also why most C libraries name their public functions library_function().

Originally Posted by Nominal Animal

You also avoid name space pollution.

Consider, for example, a project where you have several .c files that have a helper function named process() they use internally. If they are not declared as static, it will create a link time conflict; the linker won't know (but will probably try to guess, with varying results) which process() is being called where. (The header files are not examined at link time, so it does not matter whether the functions are declared in the header file or not; only whether they're static/local or not.)

This is also why most C libraries name their public functions library_function().

So, let me make sure I have this right.

If I have two source files, both have a function called process().. even though they are helper functions and not included in any header file, there will be a naming clash?

If my understanding (thanks very much to your explanation) by using static functions I wouldn't have that problem?

Originally Posted by laserlight

Yes. The technical reason is that by default, function names have external linkage, hence they must be unique across the program. Declaring them static changes the linkage to internal linkage, hence they only need to be unique to each translation unit (i.e., headers + source file).


Yes, unless you happen to define two static functions named process in the same translation unit.

When you say translation unit, you just mean two static functions with the same name inside the same source file?

Never mind, you already answered that question!

Thank you laserLight.