sksp2024-mcu/libucw/ucw/doc/generic.txt

Generic data structures and algorithms
======================================

The C preprocessor is a very powerful tool. One handy way to use it
can be generating generic data structures and algorithms. Here you can
find some conventions that are used in all such generic structures in
libUCW, and also hints for use of these structures.

- <<idea,General idea>>
- <<use,How to use them>>
- <<implement,How it is implemented>>
- Modules with generics
  * <<growbuf:gbuf,`gbuf.h`>>
  * <<sort:,Sorting>>
  * <<binheap:,`binheap.h`>>
  * <<hashtable:,`hashtable.h`>>

// TODO The module list

[[idea]]
General idea
------------

The idea is simple. If you have some code, you can customize it a
little by preprocessor macros. You can change constants, data types it
operates on, whole expressions, or you can compile parts of the code
conditionally. You can generate new function names using macros.

So if you provide few macros for data types, function names and
parameters and include some code using them, it gets modified by it
and a code for a specific data type is created. Then you can provide
new macros and include it again, to get another version of the code,
with different function names and types.

[[use]]
How to use them
---------------

The use is best explained with an example, so we will suppose there
is a header file `array.h`, which contains a generic array data type
and an indexing function, which returns a pointer to n'th element.

To get an array of integers, we need to provide macro for used data
type and macro that will provide prefixes for identifier names. Then
we include the file. Then we could get another array with unsigned
integers, so we will do the same:

  #define ARRAY_TYPE int
  #define ARRAY_PREFIX(name) intarray_##name
  #include <array.h>

  #define ARRAY_TYPE uint
  #define ARRAY_PREFIX(name) uintarray_##name
  #include <array.h>

This will generate the data types (presumably `intarray_t` and
`uintarray_t`) and the index functions (`intarray_index` and
`uintarray_index`). We can use them like anything else.

Maybe the `ARRAY_PREFIX` deserves some attention. When the header file
wants to generate an identifier, it uses this macro with
some name. Then the macro takes the name, adds a prefix to it and
returns the new identifier, so `ARRAY_PREFIX(t)` will generate
`intarray_t` in the first case and `uintarray_t` in the second. This
allows having more than one instance of the same data structure or
algorithm, because it generates different identifiers for them.

A similar macro is needed for every generic header in libUCW.

[[implement]]
How it is implemented
---------------------

For those who want to write their own or are just interested, how it
works, here is the `array.h` header and some description to it.

  #define ARRAY_A_TYPE ARRAY_PREFIX(t)
  typedef ARRAY_TYPE *ARRAY_A_TYPE

  static ARRAY_TYPE *ARRAY_PREFIX(index)(ARRAY_A_TYPE array, uint index)
  {
    return array + index;
  }

  #undef ARRAY_A_TYPE
  #undef ARRAY_TYPE
  #undef ARRAY_PREFIX

There are few things that are worth noticing. The first two lines
define the data type. The macro (`ARRAY_A_TYPE`) is only for
convenience inside the header, since such type names can be used quite
often inside the header (if it is large).

Then there is the function with its name generated (do not get scared
by the double parenthesis, ones will be eaten by the macro, the second
ones are real function parameters). The function is static, since more
than one `.c` file might want to use the same header with the same
prefix -- each one generates it's own instance.

And the end just undefines all the macros, so user may define them
again and get another instance of the data structure.

Also note it is not protected against multiple inclusion in the usual
way (eg. `#ifndef ARRAY_H` ...), since multiple inclusion is desired
-- it generates multiple versions of the data structure.