Skip to content

CarlosLunaMota/CLM_LIBS

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

51 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation


CLM_LIBS

CONTENT: The CLM_LIBS code-generating macros

AUTHOR: Carlos Luna-Mota

SOURCE: https://github.com/CarlosLunaMota/CLM_LIBS

TABLE OF CONTENTS:


INTRODUCTION

This is a collection of code-generating macros for the C preprocessor.

The main motivation to write these macros is to put together all the functions that I frequently use in my C projects so I can test them properly, document them properly and trust them hereafter.

I am using the code-generating macro approach for writting these libraries because I wanted to provide generic data structures without using void pointers.

Although this is a personal project, I am trying to document everything so others can use this code too. For the same reason, I release the code into the Public Domain using the Unlicense (see LICENSE below).


CODE CONVENTIONS

Each library is just a code generating macro whose name starts with IMPORT_CLM_. The C preprocessor will substitute each macro call with a set of functions and type definitions before compiling the code, so you just get the same behavior that you would get if you coded these auxilary functions and data types by hand (including type-safety).

The name of all types and functions starts with the name of the library (macro) they belong to. The user can add an optional prefix to all the functions and data types when calling a macro to avoid conflicts with existing code or different calls to the same macro.

All CLM_LIBS functions are declared as static inline and are mutually independent (no CLM_LIBS function calls any other CLM_LIBS function) although a few of them are recursive (they call themselves). This self-imposed constraint hurts modularity and readability but simplifies the task of versioning and reusing the code: You could erase or substitute any function without worrying about dependencies.

I am willing to sacrfice efficiency for practical reasons like safety, versatility or interface coherence. I will probably be the main user of these libraries so I want them to be easy to use rather than fast or easy to code. If performance is your main concern, there might be better libraries out there (that will probably be WAY more difficult to use). This being said, I will try to provide code that is efficient in terms of time and space. I just will not agonize about it.

I am also not targeting outdated C compilers or embeded systems. I make use of C99 features and manage the memory allocations with the standard malloc and free functions. I am writing these libraries to be easy to read and modify so anyone can adapt them to their needs, though.

Finally, regarding data types:

  • double is the default floating point type
  • size_t is the default index and size type (transdichotomous model)
  • char * is the default \0-ended string type
  • bool is the default boolean type (rather than int)
  • All container types are named as the macro they belong to and are pointers to the corresponding structs (same name with _s suffix).

CODE DEPENDENCIES

The file CLM_LIBS.h contains the code, the documentation and the license. The other files that you may find in the repository are used for the developement of CLM_LIBS.h and are not needed in practice:

  • README.md - Contains the documentation in GitHub's MarkDown format
  • CHEATSHEET.md - Contains a quick-reference cheatsheet
  • UpdateDocs.py - Contains a script for generating the documentation
  • LICENSE - Contains another copy of the same license (see below)
  • makefile - Contains a script for automating the test process
  • test.c - Contains a unit-test suite for each library

CLM_LIBS.h relies only on the C standard library and compiles without warnings using the GCC parameters -std=c99 -Wall -Wextra -pedantic -Wformat-overflow=0. More over, the code has been tested using valgrind --leak-check=full reporting no memory leaks.

The standard libraries used are (in alphabetical order):

#include <assert.h>
#include <float.h>
#include <limits.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>

CODE LICENSE

This is free and unencumbered software released into the public domain.

Anyone is free to copy, modify, publish, use, compile, sell, or distribute this software, either in source code form or as a compiled binary, for any purpose, commercial or non-commercial, and by any means.

In jurisdictions that recognize copyright laws, the author or authors of this software dedicate any and all copyright interest in the software to the public domain. We make this dedication for the benefit of the public at large and to the detriment of our heirs and successors. We intend this dedication to be an overt act of relinquishment in perpetuity of all present and future rights to this software under copyright law.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

For more information, please refer to http://unlicense.org/


THE MACROS

Each macro generates a collection of related functions and typedefs.

All the macros have an optional parameter prefix that allows the user to change the name of the functions and data types so they don't crash with other user-defined functions/types or with different calls to the same macro.

Some of the macros require two mandatory parameters (type & less) that are used to build generic functions tailored to that data type.

The type parameter could be any standard or user-defined data type.

The less parameter should be the name (NOT A POINTER, JUST THE NAME) of a comparing function/macro for the corresponding data type.

The bool less(const type x, const type y) function must accept two const type values and return:

  • less(x, y) == true if x < y
  • less(x, y) == false if x >= y

Example:

#include "CLM_LIBS.h"

#define LESS_NUM(i, j) ((i) < (j))

typedef struct pair { int i; double d; } pair_t;

bool less_pair(const pair_t p1, pair_t p2) {
    return (p1.i < p2.i || (p1.i == p2.i && p1.d < p2.d));
}

IMPORT_CLM_RAND()
IMPORT_CLM_ARRAY(int,    LESS_NUM, int_)
IMPORT_CLM_ARRAY(double, LESS_NUM, dbl_)
IMPORT_CLM_STREE(pair_t, less_pair,    )

int main(void) {

    pair_t    pair;
    size_t    i, N = 10;
    stree     tree = NULL;
    int_array A    = int_array_new(N);
    dbl_array B    = dbl_array_new(N);

    printf("\nRANDOM:\n");
    for (i = 0; i < N; i++) {
        A[i] = rand_int(100);
        B[i] = rand_double(100);
        printf("\t(%d, %f)\n", A[i], B[i]);
    }

    printf("\nLEXICOGRAPHICALLY SORTED:\n");
    for (i = 0; i < N; i++) {
        pair.i = A[i];
        pair.d = B[i];
        stree_insert(&tree, pair);
    }
    stree_min(&tree);
    while (tree) {
        pair = stree_pop(&tree);
        printf("\t(%d, %f)\n", pair.i, pair.d);
    }

    printf("\nINDEPENDENTLY SORTED:\n");
    int_array_sort(A, N);
    dbl_array_sort(B, N);
    for (i = 0; i < N; i++) { printf("\t(%d, %f)\n", A[i], B[i]); }

    free(A);
    free(B);
}

CLM_LIBS

#define CLM_LIBS 20240815

Contains the version number (= date) of this release of CLM_LIBS.


CLM_TIME

#define IMPORT_CLM_TIME(prefix)

A set of time-related and date-related functions.


time_elapsed

static inline double time_elapsed(const clock_t crono);

Returns a double with the number of seconds elapsed since the last time you executed: crono = clock();

Example: Compare the performance of two functions with:

clock_t crono = clock();
Function_1();
printf("Function 1 = %.3f sec.\n", time_elapsed(crono));

crono = clock();
Function_2();
printf("Function 2 = %.3f sec.\n", time_elapsed(crono));

time_pause

static inline void time_pause(const char* message);

Flushes the stdin, prints message in the stdout and then reads the stdin until a \n or a EOF appears.

Warning: If message == NULL prints the default message:

">> Press return to resume <<\n"

Example: Pause the program until the user hits return with:

show_results();
time_pause("\t\tPAUSE: Press return to resume.\n");
show_more_results();

time_stamp

static inline char *time_stamp(const int format);

Returns a \0 ended string with a timestamp formated as follows:

  • The Year will be included if the 6th bit of format is 1.
  • The Month will be included if the 5th bit of format is 1.
  • The Day will be included if the 4th bit of format is 1.
  • The hour will be included if the 3rd bit of format is 1.
  • The minute will be included if the 2nd bit of format is 1.
  • The second will be included if the 1st bit of format is 1.
  • The output will be "machine friendly" if format >= 0.
  • The output will be "human friendly" if format < 0.

Some useful examples:

  • get_timestamp( 32+16+8+4+2+1); returns "YYYYMMDDhhmmss"
  • get_timestamp( 32+16+8+4+2); returns "YYYYMMDDhhmm"
  • get_timestamp( 32+16+8); returns "YYYYMMDD"
  • get_timestamp( 4+2+1); returns "hhmmss"
  • get_timestamp(-32-16-8-4-2-1); returns "YYYY/MM/DD hh:mm:ss"
  • get_timestamp(-32-16-8-4-2); returns "YYYY/MM/DD hh:mm"
  • get_timestamp(-32-16-8); returns "YYYY/MM/DD"
  • get_timestamp(-32-16); returns "YYYY/MM"
  • get_timestamp( -16-8-4-2-1); returns "MM/DD hh:mm:ss"
  • get_timestamp( -16-8-4-2); returns "MM/DD hh:mm"
  • get_timestamp( -16-8); returns "MM/DD"
  • get_timestamp( -4-2-1); returns "hh:mm:ss"
  • get_timestamp( -4-2); returns "hh:mm"

Warning: The user should free the string after using it.

Warning: The case format == 0 defaults to format == 63.

Example: Output MyString in a timestamped file with:

char filename[36];
char *timestamp = time_stamp(0);

sprintf(filename, "output_%s.txt", timestamp);
FILE *MyOutput = fopen(filename, "a");
fprintf(MyOutput, MyString);

fclose(MyOutput);
free(timestamp);

CLM_RAND

#define IMPORT_CLM_RAND(prefix)

A set of random and quasi-random number generators.


rand_int

static inline int rand_int(const int n);

Returns an uniformly random int in the range [0, n).

Warning: Parameter n must satisfy RAND_MAX >= n > 0.

Example: Print 10 random int in the range [A, B] with:

size_t i;
for (i = 0; i < 10; i++) { printf("%d ", A+rand_int(B-A+1)); }

rand_size_t

static inline size_t rand_size_t(const size_t n);

Returns an uniformly random size_t in the range [0, n).

Warning: Parameter n must satisfy RAND_MAX >= n > 0.

Example: Select a random element from an array of size N with:

type elem = MyArray[rand_size_t(N)];

rand_double

static inline double rand_double(const double n);

Returns an uniformly random int in the range [0.0, n).

Warning: Parameter n must satisfy n > DLB_EPSILON.

Example: Print 10 random double in the range [A, B) with:

size_t i;
for (i = 0; i < 10; i++) { printf("%f ", A+rand_double(B-A)); }

rand_halton

static inline double rand_halton(const size_t base, size_t n);

Returns a quasi-random number in the range [0.0, 1.0).

More specifically, returns the n-th element of the Halton low-discrepancy sequence that has the base as a base.

In general it is advisable to use small prime numbers as base. In particular, if you need to produce quasi-random D-dimensional points, you should use D coprime numbers as 'base', one for each dimension (see example below).

Warning: Parameter base must satisfy base > 1.

Warning: The algorithm is deterministic, which means that for a fixed (base, n) input, it will always provide the same output.

Example: Get 100 quasi-random 3D points with:

double *X = (double *) malloc(100 * sizeof(double));
double *Y = (double *) malloc(100 * sizeof(double));
double *Z = (double *) malloc(100 * sizeof(double));
if (X && Y && Z) {
    for (size_t i = 0; i < 100; i++) {
        X[i] = x_min + (x_max-x_min)*rand_halton(2, i);
        Y[i] = y_min + (y_max-y_min)*rand_halton(3, i);
        Z[i] = z_min + (z_max-z_min)*rand_halton(5, i);
    }
}

rand_color

static inline void rand_color(int *r, int *g, int *b);

Returns a random RGB color with the same Saturation and Brightness (but different Hue) than the RGB color provided.

More specifically, generates a low discrepancy sequence of colors based on the additive recurrence:

Hue[n] = (Hue[n-1] + (phi-1)) % 1.0

where Hue is taken in the [0.0, 1.0) range and:

phi = (1.0 + sqrt(5.0)) / 2.0 = 1.61803398874989484820458683436

Warning: r, g and b must be integers in [0, 255].

Warning: The algorithm is deterministic, which means that for a fixed input color, it will always provide the same output color.

Example: Generate 5 random colors with:

int r = 255;
int g = 135;
int b =   0;
int R[5]; G[5]; B[5];
for (size_t i = 0; i < 5; i++) {
    rand_color(&r, &g, &b);
    R[i] = r; G[i] = g; B[i] = b;
}

CLM_PRINTF

#define IMPORT_CLM_PRINTF(prefix)

A set of formatting functions for the standard terminal output.


printf_reset

static inline void printf_reset();

Resets all formating parameters of printf to their default values.

Example: Give your command line a "retro look" with:

printf_reset();
printf_set_back_color(0x00, 0x00, 0x00);
printf_set_text_color(0x33, 0xFF, 0x00);

printf_delete

static inline void printf_delete();

Deletes everything from the cursor position to the end of the line.

Example: Remove the last 2 lines with:

printf_move(0,-1);
printf_delete();
printf_move(0,-1);
printf_delete();

Example: Erase the last 5 characters with:

printf_move(-5,0);
printf_delete();

printf_set_bold

static inline void printf_set_bold();

Sets the printf function in boldface mode.

Example: Print the string txt in boldface with:

printf_set_bold();
printf("%s", txt);
printf_reset();

printf_set_light

static inline void printf_set_light();

Sets the printf function in lightface mode.

Example: Print the string txt in lightface with:

printf_set_light();
printf("%s", txt);
printf_reset();

printf_set_strike

static inline void printf_set_strike();

Sets the printf function in strikethrough mode.

Example: Print the string txt striked with:

printf_set_strike();
printf("%s", txt);
printf_reset();

printf_set_underline

static inline void printf_set_underline();

Sets the printf function in underline mode.

Example: Print the string txt underlined with:

printf_set_underline();
printf("%s", txt);
printf_reset();

printf_move

static inline void printf_move(const int dx, const int dy);

Moves the cursor to the relative position defined by (dx,dy).

Warning: Text will be overwritten from that position.

Warning: Negative movement goes towards the upper-left corner.

Example: Remove the last 2 lines with:

printf_move(0,-1);
printf_delete();
printf_move(0,-1);
printf_delete();

Example: Erase the last 5 characters with:

printf_move(-5,0);
printf_delete();

printf_set_text_grey

static inline void printf_set_text_grey(const int grey);

Sets a greyscale text color for the printf function.

The greyscale palette goes from black (0) to white (23).

Warning: grey must be a integer in the range [0, 23].

Example: Print the string txt in white with:

printf_set_text_grey(23);
printf("%s", txt);
printf_reset();

printf_set_back_grey

static inline void printf_set_back_grey(const int grey);

Sets a greyscale background color for the printf function.

The greyscale palette goes from black (0) to white (23).

Warning: grey must be a integer in the range [0, 23].

Example: Print the string txt on a white background with:

printf_set_back_grey(23);
printf("%s", txt);
printf_reset();

printf_set_text_color

static inline void printf_set_text_color(const int r, const int g, const int b);

Sets the color of the text for the printf function.

The actual color of the text will be the closest color to (r,g,b) in the web-safe palette: [0x00, 0x33, 0x66, 0x99, 0xCC, 0xFF]^3

Warning: r, g and b must be integers in [0, 255].

Example: Print the string txt in red with:

printf_set_text_color(0xFF, 0x00, 0x00);
printf("%s", txt);
printf_reset();

printf_set_back_color

static inline void printf_set_back_color(const int r, const int g, const int b);

Sets the color of the background for the printf function.

The actual background color will be the closest color to (r,g,b) in the web-safe palette: [0x00, 0x33, 0x66, 0x99, 0xCC, 0xFF]^3

Warning: r, g and b must be integers in [0, 255].

Example: Print the string txt on a blue background with:

printf_set_back_color(0, 0, 255);
printf("%s", txt);
printf_reset();

CLM_HASH

#define IMPORT_CLM_HASH(prefix)

A set of reversible bit-mixer functions for unsigned integers.


hash_mix32

static inline uint32_t hash_mix32(uint32_t h);

Given a uint32_t value returns another uint32_t value that can be used as a (non-cryptographic) hash of the first value.

On average, changing a single bit in the input changes each bit of the output with probability 1/2, so, this deterministic function could be used as a non-cryptographic pseudorandom number generator.

This function is based on triple32 by Chris Wellons, see: https://nullprogram.com/blog/2018/07/31/

Warning: Zero is a fixed point by design.

Warning: This function is reversible by design.

Example: Find the index of elem in a 2^K hash table with:

int index = hash_mix32(elem) & ((1 << K)-1);

Example: Generate 5 pseudorandom uint32_t numbers with:

uint32_t N[5];
uint32_t SEED   = (uint32_t) 314159285;         // change it!
uint32_t PHI_32 = (uint32_t) 0x9e3779b9;        // 2^32 * Phi
for (int i = 0; i < 5; i++) {
    N[i] = hash_mix32(SEED + (i+1) * PHI_32);
}

hash_unmix32

static inline uint32_t hash_unmix32(uint32_t h);

Given a uint32_t value returns another uint32_t value that can be used as a (non-cryptographic) hash of the first value.

This function is the inverse of hash_mix32 and has the very same properties. See hash_mix32 documentation for the details.

Warning: Zero is a fixed point by design.

Warning: This function is reversible by design.

Example: Reverse a hash operation with:

assert(hash_unmix32(hash_mix32(i)) == i);
assert(hash_mix32(hash_unmix32(i)) == i);

hash_mix64

static inline uint64_t hash_mix64(uint64_t h);

Given a uint64_t value returns another uint64_t value that can be used as a (non-cryptographic) hash of the first value.

On average, changing a single bit in the input changes each bit of the output with probability 1/2, so, this deterministic function could be used as a non-cryptographic pseudorandom number generator.

This function is based on splitmix64 by Sebastiano Vigna, see: http://xoshiro.di.unimi.it/splitmix64.c

Warning: Zero is a fixed point by design.

Warning: This function is reversible by design.

Example: Find the index of elem in a 2^K hash table with:

size_t index = hash_mix64(elem) & ((1 << K)-1);

Example: Generate 5 pseudorandom uint64_t numbers with:

uint64_t N[5];
uint64_t SEED   = (uint64_t) 314159285;          // change it!
uint64_t PHI_64 = (uint64_t) 0x9e3779b97f4a7c15; // 2^64 * Phi
for (int i = 0; i < 5; i++) {
    N[i] = hash_mix64(SEED + (i+1) * PHI_64);
}

hash_unmix64

static inline uint64_t hash_unmix64(uint64_t h);

Given a uint64_t value returns another uint64_t value that can be used as a (non-cryptographic) hash of the first value.

This function is the inverse of hash_mix64 and has the very same properties. See hash_mix64 documentation for the details.

Warning: Zero is a fixed point by design.

Warning: This function is reversible by design.

Example: Reverse a hash operation with:

assert(hash_unmix64(hash_mix64(i)) == i);
assert(hash_mix64(hash_unmix64(i)) == i);

CLM_DAMM

#define IMPORT_CLM_DAMM(prefix)

A set of check digit functions using the Damm algorithm.


damm_dec

static inline char damm_dec(const char *txt);

Returns a the checksum character of the \0-terminated numeric string txt.

The algorithm guarantees that:

damm_dec(txt ++ damm_dec(txt)) == '0'

where txt ++ c is the concatenation of txt and c.

Warning: All non-numeric characters will be ignored.

Example: Add a checksum character to the string s with:

size_t length    = strlen(s);
char *s_checksum = (char *) malloc((length+2) * sizeof(char));
sprintf(s_checksum, "%s%c", txt, damm_dec(txt));
assert(damm_dec(s_checksum) == '0');

damm_hex

static inline char damm_hex(const char *txt);

Returns a the checksum character of the \0-terminated hexadecimal string txt.

The algorithm guarantees that:

damm_hex(txt ++ damm_hex(txt)) == '0'

where txt ++ c is the concatenation of txt and c.

Warning: All non-hexadecimal characters will be ignored.

Warning: The algorithm is case insensitive.

Example: Add a checksum character to the string s with:

size_t length    = strlen(s);
char *s_checksum = (char *) malloc((length+2) * sizeof(char));
sprintf(s_checksum, "%s%c", txt, damm_hex(txt));
assert(damm_hex(s_checksum) == '0');

CLM_ARC4

#define IMPORT_CLM_ARC4(prefix)

A set of functions related to the ARC4 stream cypher.

All these functions assume (and enforce through assertions) that the char type has a width of (exactly) 8 bits. This is quite common, but not necessarily true in all systems. This macro cannot be used in systems where that condition is not satisfied.


arc4_hash

static inline char *arc4_hash(const char *txt, const size_t length, const size_t drop);

Returns a hash of the txt string.

The output is formated as a \0 terminated array of char of twice the specified length. Each character is in hexadecimal format (i.e. all characters belong to [0,...,9,A,...,F]).

Parameter drop specifies how many pseudo-random bytes will be discarded before computing the hash. It is recomended to select a multiple of 256 as a drop (classic ARC4 defaults drop to 0).

Warning: ARC4 IS NO LONGER CONSIDERED CRYPTOGRAPHICALLY SAFE.

Warning: The txt parameter must be a \0 terminated string.

Example: Get N pseudo-random bytes with:

size_t i, ii;
char   c;
char  *seed   = "MySeed";
char  *hash   = arc4_hash(seed, N, 0);
char  *stream = (char *) malloc(N * sizeof(char));
assert(hash != NULL && stream != NULL);
for (i = ii = 0; i < N; i++) {
    c = hash[ii++];
    if ('0' <= c && c <= '9') { stream[i] = (c-'0')<<4; }
    else                      { stream[i] = (c-'A')<<4; }
    c = hash[ii++];
    if ('0' <= c && c <= '9') { stream[i] += (c-'0'); }
    else                      { stream[i] += (c-'A'); }
}
free(hash);

arc4_encrypt

static inline char *arc4_encrypt(const char *txt, const char *key, const size_t drop);

Encrypts the txt string using the key string.

The output is formated as a \0 terminated array of char that is twice as long as txt. Each character is in hexadecimal format (i.e. all characters belong to [0,...,9,A,...,F]).

Parameter drop specifies how many pseudo-random bytes will be discarded before encrypting txt. It is recomended to select a multiple of 256 as a drop (classic ARC4 defaults drop to 0).

Classic ARC4 requires 0 < strlen(key) <= 256. This version accepts smaller or larger key strings but is guaranteed to produce the same output as the classic ARC4 if the key satisfies the 0 < strlen(key) <= 256 condition.

Warning: ARC4 IS NO LONGER CONSIDERED CRYPTOGRAPHICALLY SAFE.

Warning: The key parameter must be a \0 terminated string.

Warning: The txt parameter must be a \0 terminated string.

Example: Encrypt Attack at dawn using Secret as key with:

char *code = arc4_encrypt("Attack at dawn", "Secret", 0);
if (code != NULL) { printf("Encrypted: '%s'\n", code); }
free(code);

arc4_decrypt

static inline char *arc4_decrypt(const char *txt, const char *key, const size_t drop);

Decrypts the txt string using the key string.

The txt must be formated as a \0 terminated array of char in hexadecimal format (all characters belong to [0,...,9,A,...,F]).

Parameter drop specifies how many pseudo-random bytes will be discarded before encrypting txt. It is recomended to select a multiple of 256 as a drop (classic ARC4 defaults drop to 0).

Classic ARC4 requires 0 < strlen(key) <= 256. This version accepts smaller or larger key strings but is guaranteed to produce the same output as the classic ARC4 if the key satisfies the 0 < strlen(key) <= 256 condition.

Warning: ARC4 IS NO LONGER CONSIDERED CRYPTOGRAPHICALLY SAFE.

Warning: The key parameter must be a \0 terminated string.

Warning: The txt parameter must be a \0 terminated string.

Example: Decrypt the coded message using Secret as key with:

char *code = "45A01F645FC35B383552544B9BF5"
char *txt  = arc4_decrypt(code, "Secret", 0);
if (txt != NULL) { printf("Decrypted: '%s'\n", txt); }
free(txt);

CLM_ITER

#define IMPORT_CLM_ITER(prefix)

A set of iterators to work with some generic combinatorial objects:

Products: all mixed-radix tuples of fixed length

  prod(3, (2,3,4)) = {(0,0,0), (0,0,1), (0,0,2), (0,0,3),
                      (0,1,0), (0,1,1), (0,1,2), (0,1,3),
                      (0,2,0), (0,2,1), (0,2,2), (0,2,3),
                      (1,0,0), (1,0,1), (1,0,2), (1,0,3),
                      (1,1,0), (1,1,1), (1,1,2), (1,1,3),
                      (1,2,0), (1,2,1), (1,2,2), (1,2,3)}

Permutations: all tuples of fixed length whose elements are all different and taken from a set of size base

  perm(3, 4, false) = {(0,1,2), (0,2,1), (1,0,2), (1,2,0),
                       (2,0,1), (2,1,0), (0,1,3), (0,3,1),
                       (1,0,3), (1,3,0), (3,0,1), (3,1,0),
                       (0,2,3), (0,3,2), (2,0,3), (2,3,0),
                       (3,0,2), (3,2,0), (1,2,3), (1,3,2),
                       (2,1,3), (2,3,1), (3,1,2), (3,2,1)}

Permutations with replacement: all tuples of fixed length whose elements are taken with replacement from a set of size base

  perm(3, 4, true) = {(0,0,0), (0,0,1), (0,0,2), (0,0,3),
                      (0,1,0), (0,1,1), (0,1,2), (0,1,3),
                      (0,2,0), (0,2,1), (0,2,2), (0,2,3),
                      (0,3,0), (0,3,1), (0,3,2), (0,3,3),
                      (1,0,0), (1,0,1), (1,0,2), (1,0,3),
                      (1,1,0), (1,1,1), (1,1,2), (1,1,3),
                      (1,2,0), (1,2,1), (1,2,2), (1,2,3),
                      (1,3,0), (1,3,1), (1,3,2), (1,3,3),
                      (2,0,0), (2,0,1), (2,0,2), (2,0,3),
                      (2,1,0), (2,1,1), (2,1,2), (2,1,3),
                      (2,2,0), (2,2,1), (2,2,2), (2,2,3),
                      (2,3,0), (2,3,1), (2,3,2), (2,3,3),
                      (3,0,0), (3,0,1), (3,0,2), (3,0,3),
                      (3,1,0), (3,1,1), (3,1,2), (3,1,3),
                      (3,2,0), (3,2,1), (3,2,2), (3,2,3),
                      (3,3,0), (3,3,1), (3,3,2), (3,3,3)}

Combinations: all sorted tuples of fixed length whose elements are all different and taken from a set of size base

  comb(3, 4, false) = {(0,1,2), (0,1,3), (0,2,3), (1,2,3)}

Combinations with replacement: all sorted tuples of fixed length whose elements are taken with replacement from a set of size base

  comb(3, 4, true) = {(0,0,0), (0,0,1), (0,0,2), (0,0,3),
                      (0,1,1), (0,1,2), (0,1,3), (0,2,2),
                      (0,2,3), (0,3,3), (1,1,1), (1,1,2),
                      (1,1,3), (1,2,2), (1,2,3), (1,3,3),
                      (2,2,2), (2,2,3), (2,3,3), (3,3,3)}

For each combinatorial object two functions are provided:

  • iter_rand_ returns an uniformly random element of the object.
  • iter_next_ given an element of the object, returns the next one.

The elements of a combinatorial object are always represented as size_t arrays that contain indexes of the corresponding base set (i.e. 0 <= array[i] < base) so you can use them to reference arbitrary sets of base elements.


iter_rand_prod

static inline void iter_rand_prod(size_t **prod, const size_t length, const size_t *base);

Generates an uniformly random length-sized tuple such that:

0 <= prod[i] < base[i]

If *prod == NULL the function tries to allocate a new tuple. Otherwise, it overwrites the information stored in prod.

Warning: Parameter length must satisfy lenght > 0.

Warning: Parameter base must satisfy base != NULL and also RAND_MAX >= base[i] > 0 for al i = 0...length-1.

Example: Generate three random elements of [0,5)x[0,3) with:

size_t length  = 2;
size_t base[2] = {5, 3};
size_t *prod   = NULL;
iter_rand_prod(&prod, length, base);
for (size_t i = 0; i < 3; i++) {
    do_something(prod);
    iter_rand_prod(&prod, length, base);
}

iter_next_prod

static inline void iter_next_prod(size_t **prod, const size_t length, const size_t *base);

Updates *prod to hold the next length-sized tuple such that:

0 <= prod[i] < base[i]

If *prod == NULL the function tries to allocate the first tuple. Once it reaches the last tuple, the function frees & nulls *prod.

Warning: Parameter length must satisfy lenght > 0.

Warning: Parameter base must satisfy base != NULL and also base[i] > 0 for all i = 0...length-1.

Example: Iterate over all the elements of [0,5)x[0,3) with:

size_t length  = 2;
size_t base[2] = {5, 3};
size_t *prod   = NULL;
iter_next_prod(&prod, length, base);
while (prod) {
    do_something(prod);
    iter_next_prod(&prod, length, base);
}

iter_rand_perm

static inline void iter_rand_perm(size_t **perm, const size_t length, const size_t base, const bool rep);

Generates an uniformly random length-sized tuple such that:

0 <= perm[i] < base

If *perm == NULL the function tries to allocate a new tuple. Otherwise, it overwrites the information stored in perm.

If rep == false all the elements of the tuple will be different.

Warning: Parameter length must satisfy lenght > 0.

Warning: Parameter base must satisfy base <= RAND_MAX.

Warning: If rep == true then base > 0 is required.

Warning: If rep == false then base >= length is required.

Example: Get a random element of [0,5)^3 with:

size_t *perm = NULL;
iter_rand_perm(&perm, 3, 5, true);

Example: Get a random 3-permutation of [0,5) with:

size_t *perm = NULL;
iter_rand_perm(&perm, 3, 5, false);

iter_next_perm

static inline void iter_next_perm(size_t **perm, const size_t length, const size_t base, const bool rep);

Updates *perm to hold the next length-sized tuple such that:

0 <= perm[i] < base[i]

If *perm == NULL the function tries to allocate the first tuple. Once it reaches the last tuple, the function frees & nulls *perm.

If rep == false all the elements of the tuple will be different.

Warning: Parameter length must satisfy lenght > 0.

Warning: If rep == true then base > 0 is required.

Warning: If rep == false then base >= length is required.

Example: Iterate over all the elements of [0,5)^3 with:

size_t *perm = NULL;
iter_next_perm(&perm, 3, 5, true);
while (perm) {
    do_something(perm);
    iter_next_perm(&perm, 3, 5, true);
}

Example: Iterate over all the 3-permutations of [0,5) with:

size_t *perm = NULL;
iter_next_perm(&perm, 3, 5, false);
while (perm) {
    do_something(perm);
    iter_next_perm(&perm, 3, 5, false);
}

iter_rand_comb

static inline void iter_rand_comb(size_t **comb, const size_t length, const size_t base, const bool rep);

Generates an uniformly random length-sized tuple such that:

0 <= comb[i] < base    &&    comb[i] <= comb[i+1]

If *comb == NULL the function tries to allocate a new tuple. Otherwise, it overwrites the information stored in comb.

If rep == false all the elements of the tuple will be different.

Warning: Parameter length must satisfy lenght > 0.

Warning: Parameter base must satisfy base <= RAND_MAX.

Warning: If rep == true then base > 0 is required.

Warning: If rep == false then base >= length is required.

Example: Get a random 3-combination of [0,5) with:

size_t *comb = NULL;
iter_rand_comb(&comb, 3, 5, false);

Example: Get a random 3-comb-with-rep of [0,5) with:

size_t *comb = NULL;
iter_rand_comb(&comb, 3, 5, true);

iter_next_comb

static inline void iter_next_comb(size_t **comb, const size_t length, const size_t base, const bool rep);

Updates *comb to hold the next length-sized tuple such that:

0 <= comb[i] < base    &&    comb[i] <= comb[i+1]

If *comb == NULL the function tries to allocate the first tuple. Once it reaches the last tuple, the function frees & nulls *comb.

If rep == false all the elements of the tuple will be different.

Warning: Parameter length must satisfy lenght > 0.

Warning: If rep == true then base > 0 is required.

Warning: If rep == false then base >= length is required.

Example: Iterate over all the 3-combinations of [0,5) with:

size_t *comb  = NULL;
iter_next_comb(&comb, 3, 5, false);
while (comb) {
    do_something(comb);
    iter_next_comb(&comb, 3, 5, false);
}

Example: Iterate over all the 3-comb-with-rep of [0,5) with:

size_t *comb  = NULL;
iter_next_comb(&comb, 3, 5, true);
while (comb) {
    do_something(comb);
    iter_next_comb(&comb, 3, 5, true);
}

CLM_FRACTAL

#define IMPORT_CLM_FRACTAL(prefix)

A set of functions to work with some well known space-filling curves.


fractal_van_der_corput

static inline size_t fractal_van_der_corput(const size_t base, size_t digits, size_t n);

Returns the n-th element of the Van der Corput sequence with the specified base and number of digits.

Warning: Parameter base must satisfy base > 1.

Warning: Parameter digits must satisfy digits > 0.

Warning: n can be bigger, but only the last digits digits of n will be used.

Warning: This operations is its own inverse.

Example: Print a permutation of the first K^D integers with:

size_t pow = 1;
for (size_t i = 0; i < D; i++) { pow *= K; }
for (size_t N = 0; N < pow; N++) {
    printf("%zu ", fractal_van_der_corput(K, D, N));
}

fractal_lebesgue_coord

static inline void fractal_lebesgue_coord(const size_t dim, const size_t bits, size_t *L);

Returns the dim dimensional coordinates of the N-th point of the Lebesgue (Z-index) space-filling curve of bits levels.

The 64 less significant bits of N should be stored in L[0], the next 64 less significant bits in L[1], and so on.

Warning: Parameter dim must satisfy dim > 0.

Warning: Parameter bits must satisfy bits > 0.

Warning: A size_t varible must have at least bits bits.

Example: Visit the 4096 points of the [0,16)^3 3D cube with:

size_t coord[3];
for (size_t i = 0; i < 4096; i++) {
    coord[0] = i; coord[1] = coord[2] = 0;
    fractal_lebesgue_coord(3, 4, coord);
    do_something(coord);
}

fractal_lebesgue_index

static inline void fractal_lebesgue_index(const size_t dim, const size_t bits, size_t *L);

Returns the index of the dim dimensional point of the given coordinates in a Lebesgue (Z-index) space-filling curve of bits levels.

The 64 less significant bits of the index will be stored in L[0], the next 64 less significant bits in L[1], and so on.

Warning: Parameter dim must satisfy dim > 0.

Warning: Parameter bits must satisfy bits > 0.

Warning: A size_t varible must have at least bits bits.

Example: Compare two points of the [0,16)^3 3D cube with:

size_t P[3] = {5, 3, 14};
size_t Q[3] = {3, 5, 14};
fractal_lebesgue_index(3, 4, P);
fractal_lebesgue_index(3, 4, Q);
if (P[0] < Q[0]) { printf("P goes before Q.\n"); }
else             { printf("Q goes before P.\n"); }

fractal_hilbert_coord

static inline void fractal_hilbert_coord(const size_t dim, const size_t bits, size_t *H);

Returns the dim dimensional coordinates of the N-th point of the Hilbert space-filling curve of bits levels.

The 64 less significant bits of N should be stored in L[0], the next 64 less significant bits in L[1], and so on.

Warning: Parameter dim must satisfy dim > 0.

Warning: Parameter bits must satisfy bits > 0.

Warning: A size_t varible must have at least bits bits.

Example: Visit the 4096 points of the [0,16)^3 3D cube with:

size_t coord[3];
for (size_t i = 0; i < 4096; i++) {
    coord[0] = i; coord[1] = coord[2] = 0;
    fractal_hilbert_coord(3, 4, coord);
    do_something(coord);
}

fractal_coord_to_h

static inline void fractal_hilbert_index(const size_t dim, const size_t bits, size_t *H);

Returns the index of the dim dimensional point of the given coordinates in a Hilbert space-filling curve of bits levels.

The 64 less significant bits of the index will be stored in L[0], the next 64 less significant bits in L[1], and so on.

Warning: Parameter dim must satisfy dim > 0.

Warning: Parameter bits must satisfy bits > 0.

Warning: A size_t varible must have at least bits bits.

Example: Compare two points of the [0,16)^3 3D cube with:

size_t P[3] = {5, 3, 14};
size_t Q[3] = {3, 5, 14};
fractal_hilbert_index(3, 4, P);
fractal_hilbert_index(3, 4, Q);
if (P[0] < Q[0]) { printf("P goes before Q.\n"); }
else             { printf("Q goes before P.\n"); }

CLM_DISJOINT

#define IMPORT_CLM_DISJOINT(prefix)

A disjoint-sets (union-find) data structure implemented as a size_t array of indices that represents a rooted forest (each element points to its parent, roots point to themselves).

Merging and root-finding trigger path compressions but the algorithm does not keep track of sizes, heights or ranks and always merges by pointing the root with the larger index to the root with the smallest index. This merging strategy is fast enough for pretty much all known applications and requires half of the space.

Path compressions are performed using a double leaf-to-root pass. The first pass is used to find the root of the set and the second one is used to make all the elements in the path point directly to the root.


disjoint_new

static inline size_t *disjoint_new(const size_t max_size);

Allocates max_size disjoint sets in a size_t array.

Warning: Parameter max_size must satisfy max_size > 0.

Warning: It might return NULL if there is not enough memory.

Example: Allocate N disjoint sets with:

size_t *sets = disjoint_new(N);
do_something(sets);
free(sets);

Example: Determine the number of disjoint sets with:

size_t i, num_sets = 0;
for (i = 0; i < N; i++) { if (sets[i] == i) { num_sets++; } }

disjoint_root

static inline size_t disjoint_root(size_t *sets, size_t i);

Returns the root of the set that contains the element i.

Warning: User must ensure that i < max_size.

Example: Determine if i and j belong to the same set with:

if (disjoint_root(sets, i) == disjoint_root(sets, j)) {
    printf("%zu and %zu belong to the same set!\n", i, j);
} else {
    printf("%zu and %zu belong to different sets!\n", i, j);
}

disjoint_merge

static inline size_t disjoint_merge(size_t *sets, size_t i, size_t j);

Merges the sets that contain the elements i and j.

Returns 0 if i and j belonged to the same set before calling this function and 1 otherwise (so you can easily keep track of the current number of sets).

Warning: User must ensure that i < max_size && j < max_size.

Example: Make K random merges with:

size_t *sets = disjoint_new(N);
size_t i, j, num_sets = N;
while (num_sets > N-K) {
    i = rand_size_t(N);
    j = rand_size_t(N);
    num_sets -= disjoint_merge(sets, i, j);
}

CLM_ARRAY

#define IMPORT_CLM_ARRAY(type, less, prefix)

A set of functions to work with arrays.


array

typedef type *array;

Definition of the array data type.


array_new

static inline array array_new(const size_t length);

Allocates an array of the given length.

Warning: Parameter length must satisfy length > 0.

Warning: It might return NULL if there is not enough memory.

Example: Allocate an array of length N with:

array A = array_new(N);
do_something(A);
free(A);

array_sort

static inline void array_sort(const array A, const size_t length);

Sorts the first length elements of the array A in-place using shellsort with a growing factor of 2.25.

Warning: The parameter A must satisfy: A != NULL.

Warning: This sorting algorithm is NOT stable.

Example: Sort the first N elements of MyArray with:

array_sort(MyArray, N);

Example: Sort the K smallest elements of MyArray with:

array_select(MyArray, N, K);
array_sort(MyArray, K-1);

Example: Sort the K biggest elements of MyArray with:

array_select(MyArray, N, N-K-1)
array_sort(MyArray+N-K, N-K);

array_shuffle

static inline void array_shuffle(const array A, const size_t length);

Shuffles the first length elements of the array A in-place in O(length) time.

Warning: The parameter A must satisfy: A != NULL.

Example: Shuffle the first N elements of MyArray with:

array_shuffle(MyArray, N);

array_select

static inline type array_select(const array A, const size_t length, const size_t rank);

Uses Quickselect to find and return the rank-th element of an array of size length in O(length) expected time.

The algorithm is in-place, i.e. the original order of the array can be disturbed during this operation.

In particular, the k-th element of the array will be in its correct place after doing array_select(A, N, k).

Moreover, A[0:k-1] will contain the k-1 elements of A that are smaller or equal than A[k] and A[k+1:length-1] will contain the length-k elements of A that are bigger than A[k].

Warning: The parameter A must satisfy: A != NULL.

Warning: The parameters must satisfy: rank < length.

Example: Find the median of an unsorted array of doubles with:

double median = array_select(MyArray, N, N/2);

Example: Sort the K smallest elements of MyArray with:

array_select(MyArray, N, K);
array_sort(MyArray, K-1);

Example: Sort the K biggest elements of MyArray with:

array_select(MyArray, N, N-K-1)
array_sort(MyArray+N-K, N-K);

array_bisect

static inline size_t array_bisect(const array A, const size_t length, const type data);

Returns the rightmost insertion point for data in the sorted array A in O(log(length)) time using binary search.

In particular:

  • data >= A[i] in the interval [0, array_bisect(A, N, data))
  • data < A[i] in the interval [array_bisect(A, N, data), N)

Warning: The parameter A must satisfy: A != NULL.

Warning: If A is not sorted, the behavior is undefined.

Example: Find MyData in MyArray with:

size_t i = array_bisect(MyArray, N, MyData);
if (i > 0 && MyArray[i-1] == MyData) { printf("Found");     }
else                                 { printf("Not found"); }

Example: Insertion-sort MyArray with:

for (size_t i = 1; i < length; ++i) {
     type   t = MyArray[i];
     size_t j = array_bisect(MyArray, N, t);
     memmove(A+j+1, A+j, (i-j)*sizeof(type));
     A[j] = t;
}

CLM_CLIST

#define IMPORT_CLM_CLIST(type, prefix)

A set of data types and functions related to the clist container.

The clist container is implemented as a circular singly-linked list of dinamically allocated nodes.

The clist container can be used to implement either a stack or a queue but is unable to perform the pop_back operation efficiently so it is not a fully operational deque.


clist

typedef struct clist_s {
    struct clist_s *next;
    type            data;
} clist_s, *clist;

Definition of the basic clist container data type.

Warning: An empty clist is just a NULL pointer.

Example: Initialize an empty clist with:

clist MyList = NULL;

Example: Test a MyList for non-emptyness with:

if (MyList) { printf("MyList is not empty"); }
else        { printf("MyList is empty");     }

Example: Compute the length of MyList with:

size_t size = 0;
if (MyList) {
    clist itr = MyList;
    do { itr = itr->next; size++; } while (itr != MyList);
}

Example: Iterate over all the elements of MyList with:

if (MyList) {
    clist itr = MyList;
    do {itr = itr->next;
        do_something(itr->data);
    } while (itr != MyList);
}

clist_push_front

static inline bool clist_push_front(clist *list, const type data);

Adds an elemenet at the beginning of list in O(1) time.

Warning: Returns true if successful, and false otherwise.

Example: Add data to the beginning of MyList with:

bool success = clist_push_front(&MyList, data);
if (success) { assert(clist_front(&MyList) == data); }

clist_push_back

static inline bool clist_push_back(clist *list, const type data);

Adds an elemenet at the end of list in O(1) time.

Warning: Returns true if successful, and false otherwise.

Example: Add data to the end of MyList with:

bool success = clist_push_back(&MyList, data);
if (success) { assert(clist_back(&MyList) == data); }

clist_pop_front

static inline type clist_pop_front(clist *list);

Removes and returns the first element of list in O(1) time.

Warning: Although you can add elements to either end of the list they will always exit the list from the beginning.

The combination of push_front & pop_front behaves as a LIFO stack whereas push_back & pop_front behaves as a FIFO queue.

Warning: The parameter list must satisfy *list != NULL.

Example: Empty MyList with:

while (MyList) { free(clist_pop_front(&MyList)); }

clist_front

static inline type clist_front(clist *list);

Returns the first element of a non-empty list in O(1) time.

Warning: The parameter list must satisfy *list != NULL.

Example: Retrieve the next element to exit MyList with:

type next = clist_front(&MyList);

clist_back

static inline type clist_back(clist *list);

Returns the last element of a non-empty list in O(1) time.

Warning: The parameter list must satisfy *list != NULL.

Example: Retrieve the last element to exit MyList with:

type last = clist_back(&MyList);

CLM_STREE

#define IMPORT_CLM_STREE(type, less, prefix)

A set of data types and functions related to the stree container.

The stree container is implemented as a splay tree of dinamically allocated nodes with top-down simple-splaying updates.

Splay trees were defined in the paper "Self-Adjusting Binary Search Trees" by Daniel D. Sleator & Robert E. Tarjan published in 1985 in the Journal of the ACM 32 (3): 652–686. doi:10.1145/3828.3835

Splay trees are quite different from other binary search trees so, it is advisable to learn about them at Wikipedia or other sources before using them.

All the stree operations (except stree_root that always takes O(1) time) are efficient in an amortized sense. This means that a sequence of N of these operations is guaranteed to take O(N*log(N)) time to complete, with some of them requiring as few as O(1) operations and some of them taking O(N) time to complete depending on the particular sequence of operations and the particular values inserted/searched/removed from the stree.

In practice, Splay trees are a little bit slower than the equivalent AVL tree, Red-Black tree or Weight-Balanced tree implementation but use less memory and show an adaptive behavior that might benefit the user in many common applications (e.g. inserting N sorted elements takes just O(N) time rather than O(N*log(N)) time).

Incidentaly, Splay Trees also solve the "failed queries" problem of type-safe search trees (i.e. what to return when a find or delete query fails) in a very natural way:

  • The usual find operation is broken in two steps: first ask if data is in the stree with a stree_find operation and then recover it with a stree_root operation. Since stree_find puts data in the root of the stree the second operation only takes O(1) and no time is wasted to gain safety.
  • The next and prev operations are also broken in two steps: first move the next/prev element to the root and then retrieve it with a stree_root operation. Again, since we are moving the elements to the root it just takes O(1) time to retrieve them.
  • All other data-retrieving operations (pop, min & max) also end up working with the root node of the stree and will not fail unless the stree is empty, which is a condition that can be checked easily in O(1) time with a tree != NULL test.

stree

typedef struct stree_s {
    struct stree_s *left;
    struct stree_s *right;
    type            data;
} stree_s, *stree;

Definition of the basic splay tree container type stree.

Warning: An empty stree is just a NULL pointer.

Example: Initialize an empty stree with:

stree MyTree = NULL;

Example: Test for non-emptyness of MyTree with:

if (MyTree) { printf("MyTree is not empty"); }
else        { printf("MyTree is empty");     }

stree_root

static inline type stree_root(stree *tree);

Returns the element stored at the current root node of tree in O(1) time.

Warning: The parameter tree must satisfy *tree != NULL.

Example: Retrieve the current root element of MyTree with:

if (MyTree) { root = stree_root(&MyTree); }

stree_min

static inline type stree_min(stree *tree);

Moves the smallest element to the root of tree and returns it.

Warning: The parameter tree must satisfy *tree != NULL.

Example: Find the smallest element of MyTree with:

if (MyTree) { smallest = stree_min(&MyTree); }

stree_max

static inline type stree_max(stree *tree);

Moves the biggest element to the root of tree and returns it.

Warning: The parameter tree must satisfy *tree != NULL.

Example: Find the biggest element of MyTree with:

if (MyTree) { biggest = stree_max(&MyTree); }

stree_pop

static inline type stree_pop(stree *tree);

Removes the current root node from tree and returns its content.

It guarantees that, if the current root-node element is either the biggest or the smallest in the tree, then, after performing this operation, the root node element will still be the biggest or the smallest element in the resulting tree (respectively).

Warning: The parameter tree must satisfy *tree != NULL.

Example: Remove data from MyTree (if present) with:

if (stree_find(&MyTree, data)) {
    free(stree_pop(&MyTree));
}

Example: Remove all elements from MyTree with:

while (MyTree) { free(stree_pop(&MyTree)); }

Example: Consume MyTree forwards with:

if (MyTree) {
    stree_min(&MyTree);
    while (MyTree) {
        do_something(stree_pop(&MyTree));
    }
}

Example: Consume MyTree backwards with:

if (MyTree) {
    stree_max(&MyTree);
    while (MyTree) {
        do_something(stree_pop(&MyTree));
    }
}

stree_next

static inline bool stree_next(stree *tree);

If tree is empty or the current root is the biggest element it just returs false. Otherwise, it moves to the root the smallest element that is bigger than the current root and returns true.

Example: Iterate forwards over the elements of MyTree with:

if (MyTree) {
    do_something(stree_min(&MyTree));
    while (stree_next(&MyTree)) {
        do_something(stree_root(&MyTree));
    }
}

stree_prev

static inline bool stree_prev(stree *tree);

If tree is empty or the current root is the smallest element it just returs false. Otherwise, it moves to the root the biggest element that is smaller than the current root and returns true.

Example: Iterate backwards over the elements of MyTree with:

if (MyTree) {
    do_something(stree_max(&MyTree));
    while (stree_prev(&MyTree)) {
        do_something(stree_root(&MyTree));
    }
}

stree_find

static inline bool stree_find(stree *tree, const type data);

Looks for data in the tree. If data is in the tree it moves it to the root node and returns true. Otherwise returns false.

Example: Find whether data is in MyTree with:

if (stree_find(&MyTree, data)) {
    printf("data is in MyTree.");
    do_something(stree_root(&MyTree));
} else {
    printf("data is not in MyTree.");
}

stree_insert

static inline bool stree_insert(stree *tree, const type data);

Inserts data in tree.

If tree already contains data, it gets overwritten and the function returns true. Otherwise the function insert data in tree and returns true if possible and false otherwise.

Example: Insert data in MyTree (without overwriting) with:

if (stree_find(&MyTree, data)) {
    printf("data was in MyTree already.");
} else {
    stree_insert(&MyTree, data);
}

CLM_WTREE

#define IMPORT_CLM_WTREE(type, less, prefix)

A set of data types and functions related to the wtree container.

The wtree container is implemented as a weight-balanced tree of dinamically allocated nodes with rebalancing constants (2.5, 1.5).

Weight-balanced trees were defined in the paper "Binary search trees of bounded balance" by J. Nievergelt & E. M. Reingold in 1972 in the Proceedings of the Fourth Annual ACM Symposium on Theory of Computing 137–142. doi:10.1145/800152.804906

Weight-balanced trees are among the easiest balanced trees to code, specially if you allow yourself to implement the insert and delete functions recursively. Using the constants (1+sqrt(2), sqrt(2)) for rebalancing operations guarantee a maximum tree height of, at most, 2 log(size), which is the same upper bound that the more popular Red-Black trees provide.

In practice, however, it is advisable to use integer or fractional constants such as (3, 2) or (2.5, 1.5) to avoid "integer by floating point" multiplications or squaring subtree sizes.

A wtree with rebalancing constants (X, Y) guarantees a maximum height of log(size)/log((X+1)/X). The constants (2.5, 1.5) used in this implementation guarantee a maximum height of 2.06 log(size) but usually behave much better.

There is a really thorough study on how to select these constants correctly that was published in the paper "Balancing weight-balanced trees" by Yoichi Hirai & Kazuhiko Yamamoto in 2011 in the Journal of Functional Programming 21 (3): 137–142. doi:10.1017/S0956796811000104

Another good reason to choose weight-balanced trees as the default balanced tree container is that they use the size field to maintain balance. This field is useful on its own because it allows us to perform the order-statistic queries rank and select efficiently. Given the effect of memory alignement of structs, a size_t field like size often takes as much space as a simple bool field like color so you better use the whole thing to gain functionality.

Finally, when you work with type-safe generic containers like this one you have to decide what happens when a query fails. I.e. what happens if the user ask you to recover an element that is not there.

Different libraries use different approaches to solve this problem:

  • Make the user responsible of making queries that never fail. So the user should ask if the element is there before recovering it.
  • Returning a pointer to a tree node instead of a value. Meaning that user should extract the value afterwards.
  • Some default value (usually 0) is hardcoded. Meaning that the user cannot store that value in the container.
  • The container is not fully implemented. Allowing the user to define what happens when a query fails.

I want to provide a complete generic container implementation that is able to hold any value of the selected type and I do not want the user to work with individual tree nodes, iterators or extractor functions for safety reasons. Therefore, the only remaining option is to make sure that the user never makes a failing query and using an order statistics tree as a generic container makes easy to enforce the fail-proof queries approach:

  • Making the find function return the rank of the element (or 0 if the element was not there) rather than the element itself. This allows the implementation to issue an error message (0) with reasonable semantics (if (wtree_find(*tree, data)) {...}) when the query fails and provides the reference number (rank) required to recover (select) or delete (remove) that element.
  • Using the select function (which depends on rank) to recover information rather than find, min, max, next or prev functions (which depend on data) prevents many failed queries by encouraging the user to ask for the rank of data first.
  • Make the remove function depend on the rank of the removed element rather than on the element itself. This makes the user ask for the rank of an element before trying to delete it, preventing failed deleting queries.
  • Implementing a constant-time wtree_size function allows the user to ensure the validity of any query depending on rank values with just a simple 1 <= rank && rank <= wtree_size(&tree) test.
  • Making the insert function return the rank of the inserted data not only allow the user to know whether an insertion failed (0) but also to get for free the information that other operations over the same data might require.

wtree

typedef struct wtree_s {
  struct wtree_s *left;
  struct wtree_s *right;
  size_t          size;
  type            data;
} wtree_s, *wtree;

Definition of the weight-balanced tree container type wtree.

Warning: An empty wtree is just a NULL pointer.

Example: Initialize an empty wtree with:

wtree MyTree = NULL;

Example: Test for non-emptyness of MyTree with:

if (MyTree) { printf("MyTree is not empty"); }
else        { printf("MyTree is empty");     }

wtree_size

static inline size_t wtree_size(const wtree *tree);

Returns the number of elements stored in tree in O(1) time.

Example: Copy all the elements of tree into an array with:

size_t N = wtree_size(&MyTree);
type  *A = NULL;
if (N) { A = (type *) malloc(N * sizeof(type)); }
if (A != NULL) {
    for (size_t i = 0; i < N; i++) {
        A[i] = wtree_select(&MyTree, i+1);
    }
}

wtree_find

static inline size_t wtree_find(const wtree *tree, const type data);

Looks for data in the tree in O(log(size)) time. If data is in the tree it returns its rank. Otherwise it returns 0.

Warning: The returned rank is 1-based (1 <= rank <= size).

Example: Find whether data is in MyTree with:

size_t rank = wtree_find(&MyTree, data);
if (rank) {
    printf("data is in MyTree.");
    do_something(wtree_select(&MyTree, rank));
} else {
    printf("data is not in MyTree.");
}

wtree_insert

static inline size_t wtree_insert(wtree *tree, const type data);

Inserts data in tree in O(log(size)) time.

If tree already contains data, it gets overwritten and the function returns the current rank of data in the tree. Otherwise the function tries to insert data in tree and returns its current rank if possible and 0 otherwise.

Example: Insert data in MyTree (without overwriting) with:

if (wtree_find(&MyTree, data)) {
    printf("data was in MyTree already.");
} else {
    wtree_insert(&MyTree, data);
}

Example: Implement a bisect function for wtrees with:

size_t wtree_bisect(wtree *tree, type data) {
    size_t rank = wtree_find(tree, data);
    if (rank == 0) {
        rank = wtree_insert(tree, data);
        wtree_remove(tree, rank);
    }
    return rank;
}

wtree_select

static inline type wtree_select(const wtree *tree, const size_t rank);

Returns the element of the tree that has a given rank in O(log(size)) time.

Warning: The parameter rank must satisfy 1 <= rank <= size

Example: Find the smallest element of MyTree with:

type smallest = wtree_select(&MyTree, 1);

Example: Find the biggest element of MyTree with:

type biggest = wtree_select(&MyTree, wtree_size(&MyTree));

Example: Iterate over the elements of MyTree with:

size_t size = wtree_size(&MyTree);
for (size_t i = 1; i <= size; i++) {
    do_something(wtree_select(&MyTree, i));
}

wtree_remove

static inline type wtree_remove(wtree *tree, const size_t rank);

Removes and returns the element of the tree that has a given rank in O(log(size)) time.

Warning: The parameter tree must satisfy *tree != NULL.

Warning: The parameter rank must satisfy 1 <= rank <= size

Example: Remove data from MyTree (if present) with:

size_t rank = wtree_find(&MyTree, data);
if (rank) { free(wtree_remove(&MyTree, rank)); }

Example: Remove all elements from MyTree with:

while (MyTree) { free(wtree_remove(&MyTree, 1)); }