#include <stdlib.hfa>
#include <math.h>
#include <limits.h>
#include <unistd.h>
#include <string.h>

// U(0,1)
static double U() {
    return (double)rand() / (double)INT_MAX;
}

// generate random draws from a geometric distribution of the given mean
// https://math.stackexchange.com/questions/485448/prove-the-way-to-generate-geometrically-distributed-random-numbers
static double denom;
static void initialize(int mean) {
    srand(getpid());
    double p = 1.0 / (double) mean;
    denom = log(1-p);
}
static int nextGeoRand() {
    // ret = ⌊ln(U)/ln(1−p)⌋ where U ~ U(0, 1)
    return 1 + (int) (log(U()) / denom);
}

// write a randomly generated alphabetic string whose length is adjused from a draw of the above distribution
static void emit1( int offset, double mcfreq, char mchar ) {
    int lim = offset + nextGeoRand();
    // printf("==%d\n", lim);
    for (i; lim) {
        char emit;
        if (U() < mcfreq) emit = mchar;
        else emit = 'a' + (rand() % ('z'-'a'));
        printf("%c", emit);
    }
    printf("\n");
}

// usage: ./make-corpus toGen mean [offset=0] [mcfreq=0.0] [mchar='-']
//
// Outputs alphabetic (plus magic-char) strings, one per line.
// toGen: number of strings (lines)
// 
// generated length ~  offset + geo(mean)
//                  >= 1
//
// offset=0,  mean=1:  constant length 1
// offset=0,  mean=2:  lengths go like number of coin tosses it takes to get heads
// offset=0,  mean=6:  lengths go like number of cube die rolls it takes to get :::
// offset=15, mean=1:  constant length 16
// offset=15, mean=2:  population's minimum is 16 and mean is 17
//
// Magic Char (mc) does not affect these lengths.  Any mc occurrence replaces an alphabetic char.
// mcfreq: (in [0,1]) expected fraction of the characters output that are mchar
//
int main(int argc, char ** argv) {

    int toGen;
    int mean;
    int offset = 0;
    double mcfreq = 0.0;
    char mchar = '-';

    assert(argc >= 3 && argc <= 6);
    switch(argc) {
        case 6:
            assert(strlen(argv[5]) == 0);
            mchar = argv[5][0];
        case 5:
            mcfreq = atof(argv[4]);
            assert(mcfreq >= 0.0 && mcfreq <= 1.0);
        case 4:
            offset = atoi(argv[3]);
            assert(offset >= 0 && offset < 10000);
        default:
            mean = atoi(argv[2]);
            assert(mean > 0);
            assert(mean < 1000);
            toGen = atoi(argv[1]);
            assert(toGen > 0);
            assert(toGen < 1000000);
    }

    initialize(mean);
    for( i; toGen ) {
        emit1(offset, mcfreq, mchar);
    }
}