Removing LibC dependency, improving search algorithms, simplifying API

[ashvardanian]

This is a large PR, but most importantly, it removes the LibC dependency 🥳

#include <ctype.h>  // `tolower`
#include <search.h> // `qsort_s`
#include <stddef.h> // `sz_size_t`
#include <stdint.h> // `uint8_t`
#include <stdlib.h> // `qsort_r`
#include <string.h> // `memcpy`

All of those headers are gone.

// Define a type for the comparison function, depending on the platform.
#if defined(WIN32) || defined(_WIN32) || defined(__WIN32__) || defined(__NT__) || defined(__APPLE__)
typedef int (*sz_qsort_comparison_func_t)(void *, void const *, void const *);
#else
typedef int (*sz_qsort_comparison_func_t)(void const *, void const *, void *);
#endif


inline static int _sz_sort_sequence_strncmp(
#if defined(WIN32) || defined(_WIN32) || defined(__WIN32__) || defined(__NT__) || __APPLE__
    void *sequence_raw, void const *a_raw, void const *b_raw
#else
    void const *a_raw, void const *b_raw, void *sequence_raw
#endif
) {
    // https://man.freebsd.org/cgi/man.cgi?query=qsort_s&sektion=3&n=1
    // https://www.man7.org/linux/man-pages/man3/strcmp.3.html
    sz_sequence_t *sequence = (sz_sequence_t *)sequence_raw;
    sz_size_t a = *(sz_size_t *)a_raw;
    sz_size_t b = *(sz_size_t *)b_raw;
    sz_size_t a_len = sequence->get_length(sequence->handle, a);
    sz_size_t b_len = sequence->get_length(sequence->handle, b);
    int res = strncmp( //
        sequence->get_start(sequence->handle, a),
        sequence->get_start(sequence->handle, b),
        a_len > b_len ? b_len : a_len);
    return res ? res : a_len - b_len;
}

inline static int _sz_sort_sequence_strncasecmp(
#if defined(WIN32) || defined(_WIN32) || defined(__WIN32__) || defined(__NT__) || __APPLE__
    void *sequence_raw, void const *a_raw, void const *b_raw
#else
    void const *a_raw, void const *b_raw, void *sequence_raw
#endif
) {
    // https://man.freebsd.org/cgi/man.cgi?query=qsort_s&sektion=3&n=1
    // https://www.man7.org/linux/man-pages/man3/strcmp.3.html
    sz_sequence_t *sequence = (sz_sequence_t *)sequence_raw;
    sz_size_t a = *(sz_size_t *)a_raw;
    sz_size_t b = *(sz_size_t *)b_raw;
    sz_size_t a_len = sequence->get_length(sequence->handle, a);
    sz_size_t b_len = sequence->get_length(sequence->handle, b);
    int res = strncasecmp( //
        sequence->get_start(sequence->handle, a),
        sequence->get_start(sequence->handle, b),
        a_len > b_len ? b_len : a_len);
    return res ? res : a_len - b_len;
}

...

        // Perform sorts on smaller chunks instead of the whole handle
#if defined(WIN32) || defined(_WIN32) || defined(__WIN32__) || defined(__NT__)
        // https://stackoverflow.com/a/39561369
        // https://learn.microsoft.com/en-us/cpp/c-runtime-library/reference/qsort-s?view=msvc-170
        qsort_s(sequence->order, split, sizeof(sz_size_t), qsort_comparator, (void *)sequence);
        qsort_s(sequence->order + split,
                sequence->count - split,
                sizeof(sz_size_t),
                qsort_comparator,
                (void *)sequence);
#elif __APPLE__
        qsort_r(sequence->order, split, sizeof(sz_size_t), (void *)sequence, qsort_comparator);
        qsort_r(sequence->order + split,
                sequence->count - split,
                sizeof(sz_size_t),
                (void *)sequence,
                qsort_comparator);
#else
        // https://linux.die.net/man/3/qsort_r
        qsort_r(sequence->order, split, sizeof(sz_size_t), qsort_comparator, (void *)sequence);
        qsort_r(sequence->order + split,
                sequence->count - split,
                sizeof(sz_size_t),
                qsort_comparator,
                (void *)sequence);
#endif

...

The ugly qsort_r and qsort_s mess is also gone!

[ashvardanian]

🎉 This PR is included in version 2.0.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀