Uni- and bidirectional pseudo-generic shortest path finders in C89

Question

I have implemented two related shortest path algorithms for unweighted graphs in C89. My attempt was to learn some more idiomatic C constructs such as genericity (a client programmer should be able to plug in his/her graph data structures to the algorithms). Some questions:

1. Is it a good idea to embed the unit tests of a data type in the same *.c files?
2. Is it a good idea to do rather explicit type casting in order to make gcc -Wall -pendatic -ansi shut up?
3. Might the code below rely on an antipattern?

Code

bidirectional_breadth_first_search.c

#include "bidirectional_breadth_first_search.h"
#include "list.h"
#include "queue.h"
#include "utils.h"
#include <limits.h>

#define NEIGHBOR_NODE_ITERATOR_HAS_NEXT child_node_iterator_has_next
#define NEIGHBOR_NODE_ITERATOR_NEXT     child_node_iterator_next

typedef child_node_iterator neighbor_node_iterator;

list* bidirectional_breadth_first_search(void* source_node,
                                         void* target_node,
                                         child_node_iterator* child_iterator,
                                         parent_node_iterator* parent_iterator,
                                         size_t (*hash_function)(void*),
                                         int (*equals_function)(void*, void*))
{
    queue* queue_a;
    queue* queue_b;
    unordered_map* parents_a;
    unordered_map* parents_b;
    unordered_map* distance_a;
    unordered_map* distance_b;
    size_t dist_a;
    size_t dist_b;
    size_t best_cost;
    void* touch_node;
    void* current_node;
    void* child_node;
    void* parent_node;

    if (!source_node
        || !target_node
        || !child_iterator
        || !parent_iterator
        || !hash_function
        || !equals_function)
    {
        return NULL;
    }

    queue_a = queue_alloc();
    queue_b = queue_alloc();

    parents_a = unordered_map_alloc(10,
                                    1.0f,
                                    hash_function,
                                    equals_function);

    parents_b = unordered_map_alloc(10,
                                    1.0f,
                                    hash_function, 
                                    equals_function);

    distance_a = unordered_map_alloc(10,
                                     1.0f,
                                     hash_function,
                                     equals_function);

    distance_b = unordered_map_alloc(10,
                                     1.0f,
                                     hash_function,
                                     equals_function);

    queue_push_back(queue_a, source_node);
    queue_push_back(queue_b, target_node);
    unordered_map_put(parents_a, source_node, NULL);
    unordered_map_put(parents_b, target_node, NULL);
    unordered_map_put(distance_a, source_node, 0);
    unordered_map_put(distance_b, target_node, 0);

    best_cost = UINT_MAX;
    touch_node = NULL;

    while (queue_size(queue_a) > 0 && queue_size(queue_b) > 0)
    {
        dist_a = (size_t) unordered_map_get(distance_a, queue_front(queue_a));
        dist_b = (size_t) unordered_map_get(distance_a, queue_front(queue_b));

        if (touch_node && best_cost <= dist_a + dist_b)
        {
            return trace_back_path_bidirectional(touch_node,
                                                 parents_a,
                                                 parents_b);
        }

        current_node = queue_pop_front(queue_a);
        dist_a = (size_t) unordered_map_get(parents_a, current_node);
        dist_b = (size_t) unordered_map_get(parents_b, current_node);

        if (unordered_map_contains_key(parents_b, current_node)
            &&
            best_cost > dist_a + dist_b)
        {
            best_cost = dist_a + dist_b;
            touch_node = current_node;
        }

        child_iterator->child_node_iterator_init(child_iterator, 
                                                 current_node);

        while (child_iterator->child_node_iterator_has_next(child_iterator))
        {
            child_node = child_iterator->
                         child_node_iterator_next(child_iterator);

            if (!unordered_map_contains_key(parents_a, child_node))
            {
                unordered_map_put(parents_a, child_node, current_node);
                queue_push_back(queue_a, child_node);
            }
        }

        child_iterator->child_node_iterator_free(child_iterator);
        current_node = queue_pop_front(queue_b);
        dist_a = (size_t) unordered_map_get(distance_a, current_node);
        dist_b = (size_t) unordered_map_get(distance_b, current_node);

        if (unordered_map_contains_key(parents_a, current_node)
            &&
            best_cost > dist_a + dist_b)
        {
            best_cost = dist_a + dist_b;
            touch_node = current_node;
        }

        parent_iterator->parent_node_iterator_init(parent_iterator,
                                                   current_node);

        while (parent_iterator->
               parent_node_iterator_has_next(parent_iterator))
        {
            parent_node = parent_iterator->
                          parent_node_iterator_next(parent_iterator);

            if (!unordered_map_contains_key(parents_b, parent_node))
            {
                unordered_map_put(parents_b, parent_node, current_node);
                queue_push_back(queue_b, parent_node);
            }
        }

        parent_iterator->parent_node_iterator_free(parent_iterator);
    }

    return NULL;
}

breadth_first_search.c

#include "breadth_first_search.h"
#include "directed_graph_node.h"
#include "queue.h"
#include "list.h"
#include "unordered_map.h"
#include "unordered_set.h"

static list* trace_back_path(void* target_node, unordered_map* parents)
{
    list* path = list_alloc(10);
    void* node = target_node;

    while (node)
    {
        list_push_front(path, node);
        node = unordered_map_get(parents, node);
    }

    return path;
}

list* breadth_first_search(void* source_node, 
                           void* target_node,
                           child_node_iterator* child_iterator,
                           size_t (*hash_function)   (void*),
                           int    (*equals_function) (void*, void*))
{
    queue* q = queue_alloc();
    unordered_map* parent_map = 
        unordered_map_alloc(10,
                            1.0f, 
                            hash_function, 
                            equals_function);
    void* current_node;
    void* child_node;

    if (!source_node
        || !target_node
        || !child_iterator
        || !hash_function
        || !equals_function)
    {
        return NULL;
    }

    queue_push_back(q, source_node);
    unordered_map_put(parent_map, source_node, NULL);

    while (queue_size(q) > 0)
    {
        current_node = queue_pop_front(q);

        if (equals_function(current_node, target_node))
        {
            return trace_back_path(target_node, parent_map);
        }

        child_iterator->child_node_iterator_init(child_iterator, current_node);

        while (child_iterator->child_node_iterator_has_next(child_iterator))
        {
            child_node = child_iterator->
                         child_node_iterator_next(child_iterator);

            if (!unordered_map_contains_key(parent_map, child_node))
            {
                unordered_map_put(parent_map, child_node, current_node);
                queue_push_back(q, child_node);
            }
        }

        child_iterator->child_node_iterator_free(child_iterator);
    }

    return NULL;
}

(The entire repository is here.)

Answer 1

Distance?

In your bidirectional search you have the concepts of distance and cost, but nowhere do you actually insert any distance values to your unordered maps, except to initialize each endpoint with distance zero. Therefore, you will always retrieve a zero distance from distance_a and distance_b, either because you asked for the distance to the endpoint, or because you asked for the distance to any other node, which will not exist in the map (and you will get back NULL which is casted to a zero distance).

There are a couple of lines that are different than the others, but I think you made a copy paste error:

   dist_a = (size_t) unordered_map_get(parents_a, current_node);
   dist_b = (size_t) unordered_map_get(parents_b, current_node);

Here, you have the wrong maps. These maps should be distance_a and distance_b instead of parents_a and parents_b. Otherwise the distances you get back will be pointer values (i.e. random addresses) rather than distance values.

Your function still works because it will return the shortest path (by number of edges), simply because a breadth first search naturally does that. But you should either remove all the distance code if the graph is unweighted, or fix up the distance code if the graph is weighted.

Another copy paste error

I also think that this line is a copy paste error:

    dist_b = (size_t) unordered_map_get(distance_a, queue_front(queue_b));

I think distance_a should be distance_b here.

Memory leaks

None of the queues and maps that you allocate are ever freed, thus causing memory leaks.

size_t vs pointer

Your unordered maps use void * as the type of the value stored but you often cast the result of a get() to a size_t. According to this, size_t could be larger or smaller than a pointer, which could lead to truncation of values.