szaydel
11/13/2019 - 10:36 PM
histogram utility
OS := $(shell uname -s)
ifeq ($(OS), Darwin)
CC = clang
LIBS += -lgsl
endif
ifeq ($(OS), Linux)
CC = gcc
LIBS += -lgsl -lgslcblas -lm
endif
histogram: histogram.o
$(CC) -Wall -g3 -o histogram histogram.o $(LIBS)
histogram.o: histogram.c
$(CC) -Wall -g3 -c histogram.c
clean:
rm -rf *.o histogram
#ifdef __MACH__
#include <mach/mach_time.h>
#endif
#ifdef __linux
#include <time.h>
#endif
#include <math.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <sys/ioctl.h>
#include <gsl/gsl_histogram.h>
#include <gsl/gsl_matrix.h>
#include <gsl/gsl_rstat.h>
#include <gsl/gsl_vector.h>
typedef struct prog_cfg {
int symbol;
bool base10bins;
bool log2scale;
} prog_cfg_t;
typedef struct time_it {
uint64_t start;
uint64_t elapsed;
uint32_t scale;
} time_it_t;
enum e_args {
SOLID,
HASH,
CIRCLE,
SQUARE,
LOG,
};
uint64_t timestamp(void) {
#ifdef __MACH__ // OSX does not have clock_gettime, use mach_absolute_time.
return mach_absolute_time();
#else
struct timespec ts;
clock_gettime(CLOCK_MONOTONIC, &ts);
return ts.tv_nsec + (ts.tv_sec * 1000000000LL);
#endif
}
#define NUM_BINS_BASE2 34
#define NUM_BINS_BASE10 22
int main(int argc, char **argv) {
char *symbols[] = {"▇", "▉", "◉", "▨"};
char *block = symbols[0];
prog_cfg_t config = {
.symbol = SOLID,
.base10bins = false,
};
// Process command line arguments, if multiple styles listed, last one wins!
for (size_t i = 0; i < argc; i++) {
if (strcmp(argv[i], "-hash") == 0) {
config.symbol = HASH;
} else if (strcmp(argv[i], "-circle") == 0) {
config.symbol = CIRCLE;
} else if (strcmp(argv[i], "-square") == 0) {
config.symbol = SQUARE;
} else if (strcmp(argv[i], "-log") == 0) {
config.log2scale = true;
} else if (strcmp(argv[i], "-base10bins") == 0) {
config.base10bins = true;
}
}
time_it_t how_long = {
.start = 0,
.elapsed = 0,
};
double hist_bins[NUM_BINS_BASE2] = {0,
(long)2 << 0,
(long)2 << 1,
(long)2 << 2,
(long)2 << 3,
(long)2 << 4,
(long)2 << 5,
(long)2 << 6,
(long)2 << 7,
(long)2 << 8,
(long)2 << 9,
(long)2 << 10,
(long)2 << 11,
(long)2 << 12,
(long)2 << 13,
(long)2 << 14,
(long)2 << 15,
(long)2 << 16,
(long)2 << 17,
(long)2 << 18,
(long)2 << 19,
(long)2 << 20,
(long)2 << 21,
(long)2 << 22,
(long)2 << 23,
(long)2 << 24,
(long)2 << 25,
(long)2 << 26,
(long)2 << 27,
(long)2 << 28,
(long)2 << 29,
(long)2 << 30,
(long)2 << 31,
(long)2 << 32};
double hist_bins_base10[NUM_BINS_BASE10] = {
0, 1, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9, 1e10,
1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19, 1e20,
};
double x;
double largest = 0;
size_t maxblks = 0;
gsl_histogram *h;
// Allocate a new histogram
if (config.base10bins) {
h = gsl_histogram_alloc(NUM_BINS_BASE10 - 1);
if (!h) {
perror("error in gsl_histogram");
exit(1);
}
gsl_histogram_set_ranges(h, hist_bins_base10, NUM_BINS_BASE10);
} else {
h = gsl_histogram_alloc(NUM_BINS_BASE2 - 1);
if (!h) {
perror("error in gsl_histogram");
exit(1);
}
gsl_histogram_set_ranges(h, hist_bins, NUM_BINS_BASE2);
}
// Allocate workspace for running stats
gsl_rstat_workspace *ws = gsl_rstat_alloc();
if (!ws) {
perror("error in gsl_rstat_alloc");
exit(1);
}
gsl_rstat_quantile_workspace *qws_25 = gsl_rstat_quantile_alloc(0.25);
gsl_rstat_quantile_workspace *qws_50 = gsl_rstat_quantile_alloc(0.5);
gsl_rstat_quantile_workspace *qws_75 = gsl_rstat_quantile_alloc(0.75);
if (!qws_25 || !qws_50 || !qws_75) {
perror("error in gsl_rstat_quantile_alloc");
exit(1);
}
how_long.start = timestamp();
while (fscanf(stdin, "%lg", &x) == 1) {
if (x > largest)
largest = x;
(void)gsl_rstat_add(x, ws);
(void)gsl_rstat_quantile_add(x, qws_25);
(void)gsl_rstat_quantile_add(x, qws_50);
(void)gsl_rstat_quantile_add(x, qws_75);
(void)gsl_histogram_increment(h, x);
}
how_long.elapsed = timestamp() - how_long.start;
// Count number of bins that actually have values
double nonempty_bins = 0;
size_t limit =
config.base10bins ? (NUM_BINS_BASE10 - 1) : (NUM_BINS_BASE2 - 1);
for (size_t i = 0; i < limit; i++) {
if (gsl_histogram_get(h, i) > 0) {
nonempty_bins++;
}
}
// Compute number of blocks available to print from number of available
// columns in this terminal window.
struct winsize winsz;
ioctl(STDOUT_FILENO, TIOCGWINSZ, &winsz);
maxblks = winsz.ws_col - 20;
// Set block to first enabled
block = symbols[config.symbol];
// Compute scale
double scale = 0;
if (config.log2scale) {
scale = maxblks / log2(gsl_histogram_max_val(h));
} else {
scale = maxblks / gsl_histogram_max_val(h);
}
printf("\n");
double upper, lower;
how_long.start = timestamp();
for (size_t i = 0; i < limit; i++) {
if (gsl_histogram_get(h, i) == 0 &&
(gsl_histogram_bins(h) > i + 1 && gsl_histogram_get(h, i + 1) == 0))
continue;
(void)gsl_histogram_get_range(h, i, &lower, &upper);
if (lower > largest) {
goto done;
}
printf("%.3e |", lower);
double freq = gsl_histogram_get(h, i) / gsl_rstat_n(ws);
double end = scale * (config.log2scale ? log2(gsl_histogram_get(h, i))
: gsl_histogram_get(h, i));
for (size_t j = 0; j < end; j++) {
printf("%s", block);
}
if (!freq) {
printf("\n");
continue;
}
printf(" %ld\n", (long)gsl_histogram_get(h, i));
}
done:
how_long.elapsed += timestamp() - how_long.start;
uint64_t msecs = how_long.elapsed / 1000000;
printf("\n | quantiles\n");
printf(" spent(ms) %8.2g |\n", (double)(msecs));
printf(" min %8.2g | 25%% %.2g\n", gsl_rstat_min(ws),
gsl_rstat_quantile_get(qws_25));
printf(" max %8.2g | 50%% %.2g\n", gsl_rstat_max(ws),
gsl_rstat_quantile_get(qws_50));
printf(" mean(μ) %8.2g | 75%% %.2g\n", gsl_rstat_mean(ws),
gsl_rstat_quantile_get(qws_75));
printf(" stddev(σ) %8.2g |\n", gsl_rstat_sd(ws));
gsl_histogram_free(h);
gsl_rstat_free(ws);
gsl_rstat_quantile_free(qws_25);
gsl_rstat_quantile_free(qws_50);
gsl_rstat_quantile_free(qws_75);
return 0;
}
#!/usr/bin/env python
from random import randrange
from sys import argv
end = 1
if len(argv[:]) > 1:
end = int(argv[1])
for i in range(0, end):
print(randrange(0, 1<<32))