An iterator is a description which allows to define a larger sets of tests. Instead of defining every combination of a parameter setting manually, generators provide iterators over permutation sets. For that HWUT generates code which allows users to apply ‘iterators’ to iterate over their parameter settings.

In C an iterator may be used as follows:

Generators are specified in text files (or in comments of actual source code). Each parameter that appears in a setting must be specified along with the set of possible values that it may take. For statically typed languages the ‘type’ must be specified along. The simplest case is an iteration over constant values. In this case the specification of a iterator basically a ‘CSV’ file content such as

int x;  int y; char* my_string;
    0;      1;         "hello";
    0;      0;         "hello";
   10;      1;         "hello";
   10;      0;         "hello";
    0;      1;       "bonjour";
    0;      0;       "bonjour";
   10;      1;       "bonjour";
   10;      0;       "bonjour";

The header line describes the parameters, namely x, y, and my_string. What preceedes their names are their types. All following lines describe the settings of those parameters. A generated iterator it in C may then reproduce the content of this table easily by

myIterator_t* it = myIterator_new();

while( myIterator_next(it) ) {
    printf("%i; %i; \"%s\";", it->x, it->y, it->my_string);

Real tests may use the parameters provided by it to do some serious testing. The syntax of ‘selections’ may be used to define a set of values for a parameter to chose. Selections are defined as comma separated lists of values surrounded by [ and ] brackets. Using selections the aforementioned example shrinks down to

 int x;  int y;       char* my_string;
[0,10];  [0,1];  ["hello", "bonjour"];

This means that x can take the values 0 and 10, y can take the values 0 and 1, and my_string can be set to "hello" and "bonjour". HWUT automatically iterates over all permutations of theses settings without changing any line in the code that uses the iterator.

Every line following the header line is a permutation specification. HWUT can provide iterators that seeminglessly iterate over multiple permutation sets. For example, consider to test a function that tells whether the product of two numbers is even or odd. Math tells us that if a number is multiplied with an even number, then the result is even, if both operands are odd, the result is odd.

int x;          int y;           bool even;
## Odd  x Odd  => Odd
[1,3,5,7,9];    [1,3,5,7,9];     false;
## Even x Odd  => Even
[0,2,4,6,8,10]; [1,3,5,7,9];     true;
## Odd  x Even => Even
[1,3,5,7,9];    [0,2,4,6,8,10];  true;
## Even x Even => Even
[0,2,4,6,8,10]; [0,2,4,6,8,10];  \

Each of the four lines after the header line describes a iterator. Lines starting or ending with ## are comment lines and semantically ignored. The backslash \ allows to continue a line after line break, as shown in the last line of the fragment. Since the numbers for x and y in the first line are all odd, it can be said that any permutation of them results in products which are odd. All other iterators result in products which are even. The test code in C would be

myIterator_t   it;

while( myIterator_next(it) ) {
    assert( product_is_even(it->x, it->y) == it->result );

The above table is a transparent description of a large number of tests, namely: 5x5 + 6x5 + 5x6 + 6x6 = 121 tests. Ranges by be used to make things even more concise. They are expressions following the pattern:




where first is a numeric value that defines the first number in the range and last defines the last number in the range. Optionally, the step size in the range may be given by step_size. Using ‘ranges’ allows for even more concise descriptions.

int x;     int y;    bool even;
## Odd x Odd         => Odd
|1:9:2|;   |1:9:2|;  false;
## Even x Anything (and vice versa) => Even
|0:10:2|;  |0:10|;   true;
|0:10|;    |0:10:2|; true;

It is even possible to define settings of parameters in terms of other parameters using the ‘focus’. A focus description using one of the following syntaxes:

@ focus
@ focus | reach |
@ focus | reach : step |
@ focus | reach : step : min |
@ focus | reach : step : min : max |

where focus may either be the name of another factor or some source code provided in between {^ and ^} brackets. The reach defines the extend around the given focus. The default value is 1. step defines the step size inside the reach. The set of addmissible values can be restricted by min (first addmissible) and max (last addmissible).

Using focusses makes it very easy to write tests that trigger specific conditions inside the function, e.g. the if condition in

void some_func(uint8_t x, uint8_t y) {
    if( x < y ) {

can be checked in all its variations by

int x;      int y;
|0:0xFF|;      @x;

more complex conditions are only slightly complexer to test. For example the if condition in

void some_func(uint8_t x, uint8_t y) {
    if( x + z < y ) {

may be explored by

int x;      int y;
|0:0xFF|;   @{^ x + z ^};

This is so, since the any code may appear inside the {^ and ^} brackets. The next two sections discuss in detail the types of parameters and the way that ranges can be specified. The following table summarizes the different way to define possible parameter values.

  Syntax Meaning
Selection [a,b,c, ...] Any value from a, b, c, ...
Range |a:b| Numbers from a to b including a and b. Numbers have a distance of 1.
... with step |a:b:c| Like previous where numbers have a distance of c.
Focus @x|a| Numbers in a reach of a around parameter x. Numbers have a distance of 1.
... with step @x|a:b| Like previous where numbers have a distance of b.
... with step and border @x|a:b:c:d| Like previous where numbers are greater or equal c and lesser or equal d.

The role of parameter x as focus may be played by expressions in {^ and ^} for more complex focus definitions.


Some elegant constants in line

a set of tests quite well define.

A nicer choice for your collection

comes with the feature called ‘selection’.

And, to avoid much hokus pocus

instruct your par’meters to ‘focus’.

Value Types

HWUT defines elementary types to describe parameter values. Those value types are HWUT-intern and not directly related to the concrete types as they are mentioned in the header line of the iterator description. The value of a parameter will be squeezed into the parameter’s concrete type once code is generated.

The HWUT value types are the following:


Values of positive and negative integers. Accepted integer formats are

Numeric Base Pattern Example
Decimal [0-9]+ 4711
Hexadecimal 0x[0-9a-fA-F.]+ 0xC0.FFEE.BABA
Octal 0o[0-9a-fA-F.]+ 0o731
Binary 0b[0-1.]+ 0b111.0100.1.10.101

The ‘.’ may be used in prefixed numbers as a redundant marker to help reading the number itself. This comes handy in larger bit frames for examples, where it would be difficult otherwise to identify specific bits.

INTEGER values can be used as constants, in selections, in ranges, and in focus ranges.


A FLOAT is a finite number of digits following the decimal point. In the description of INTEGER the ‘.’ has been disallowed as marker. This happened because the ‘.’ following the decimal number pattern results in the FLOAT pattern.



FLOAT values can be used as constants, in selections, in ranges, and in focus ranges.


A sequence of characters in " quotes is interpreted as a string. A backslashed " is replaced by a quote. A backslashed backslash is replaced by a backslash.


"Hello World!"

STRING values can only be used as constants and in selections.


An array is a sequence of integers bracket by { and } brackets. Elements of the array are separated by commas.


{ 0x01, 0x04, 0xFE, 0xE4 }

IARRAY objects can only be used as constants and in selections.


An array is a sequence of floats. They follow the syntax of IARRAY except that the numbers in brackets are floats.


{ 3.14159, 2.71828, 1.61803, 2.58498 }

FARRAY objects can only be used as constants and in selections.



allow your scalars to promote.

If arrays enter then the play

embrace IARRAY and F-ARRAY!

Those names you talk then a la mode

but never write them in your code.

The terms INTEGER, FLOAT, STRING, IARRAY, FARRAY never appear in any iterator specification. A parameter’s type is determined through their pattern and syntax. So, the user needs somehow be aware about what type is his specifying.

Minimalist Example

This section provides a complete example using generators in C. It is a good idea to keep the iterator definition in the test file, so that all test related data is in one place. The iterator code must be setup in a region which is semantically indifferent to the compiler. This can be a long /* to */ comment or a region spun by #if 0 and #endif. Consider the following file test-it.c

#if 0
<<hwut-iterator:  myIterator>>
<<hwut-file:      myIterator>>
#include <stdint.h>
    int Case;    int x;    int y;
           0;  |1:9:2|;  |1:9:2|;
           1; |0:10:2|;   |0:10|;

#include "hwut_unit.h"
#include "myIterator.h"

int main(int argc, char** argv)
    myIterator_t it;

    hwut_info("Check product of even and odd;");


    while( myIterator_next(&it) ) {
        if( it->Case == 0 ) {
           // Odd x Odd == Odd
           assert( my_product(it->x, it->y) % 2 != 0 );
        } else if( it->Case == 0 ) {
           // Even x Anything == Even
           assert( my_product(it->x, it->y) % 2 == 0 );
           assert( my_product(it->y, it->x) % 2 == 0 );

The marker <<hwut-iterator: myIterator>> tells the parser that from here on a iterator definition follows. The second marker <<hwut-file: myIterator>> defines the file stem of the output files. The section following the first dashed line contains source code to be pasted into the iterator header iterator. What follows the second dashed line is the definition of the iterator sections. The last dashed line signalizes the end of the iterator. Precisely an iterator definition consists of the following elements:

# A <<hwut-iterator ...>> tag possibly followed by a <<hwut-file ...>>

# A dashed line.

# Some source header content to be pasted in from of the iterator’s header.

# A dashed line.

# The iterator definition consisting of:

# A header line defining the name of the elements and possibly their type.

# A list of lines which define permutation sections.

# A dashed line.

Multiple iterators may be specified in the same file. For this, simply a new <<<hwut-iterator ...>> tag appear somewhere else–followed by the aforementioned sequence of definitions.


There is a family of hwut_verify_verbose_it...-functions which take use of the iterators. They check a condition and print the setting of the iterator directly, in case of error. Namely, those functions are:

hwut_verify_verbose_it(it, condition);
hwut_verify_it(it, condition);
hwut_verify_verbose_it(it, annotation, condition);
hwut_verify_it(it, annotation, condition);

All functions expect a pointer to an iterator as first argument. The condition argument specifies the condition to be tested. The annotation is some additional message to be printed on screen when this test is considered. They are good candidates to replace the assert statements during the test.

In the C code the new header file myIterator.h needs to included so that the names of the iterator are known. Later, the linker needs to link the test application against a compiled version myIterator.c which provides the functionality.

Hwut needs to called with gen as first argument in order to trigger iterator code generation. This is best done in a Makefile.

test-it.exe: test-it.c
    hwut gen test-it.c
    $(CC) -I$(HWUT_PATH)/support/C             \
          $(HWUT_PATH)/support/C/hwut_cursor.c \
          test-it.c                            \
          myIterator.c                         \
          -o test-it.c

Since myIterator.c needs to be redone each time that test-it.c is modified no separate rule is required for the hwut gen command. It can be pasted immediately before the CC compile command. The HWUT iterators use hwut_cursor.c which has to be specified also on the command line. The include path -I$(HWUT_PATH)/support/C is required so that HWUT’s headers may be found. This test can be build with:

> make test-it.exe

and executed by:

> ./test-it.exe