From c99fd96b4467242051e974e746ce55916729989d Mon Sep 17 00:00:00 2001
From: "Vincent A. Arcila" <vincent.arcila@bsc.es>
Date: Thu, 29 May 2025 16:04:54 +0200
Subject: [PATCH] feat: Add csv output functionality

---
 bigotes.c | 336 +++++++++++++++++++++++++++++++++++++-----------------
 1 file changed, 230 insertions(+), 106 deletions(-)

diff --git a/bigotes.c b/bigotes.c
index b0859dd..c9235e4 100644
--- a/bigotes.c
+++ b/bigotes.c
@@ -17,11 +17,15 @@
 #include "stats.h"
 #include "common.h"
 
+
+#define STR_SIZE 8192
+
 static int read_from_stdin = 0;
 static int use_wall_clock = 0;
 static double min_wall_time = 30.0;
 static int use_exec = 0;
 static int use_machine_output = 0;
+static int use_csv_output = 0;
 static int be_quiet = 0;
 static int trim_outliers = 0;
 static long min_samples = 30;
@@ -158,26 +162,6 @@ cmp_double(const void *pa, const void *pb)
 		return 0;
 }
 
-/* Test for normality */
-static void
-shapiro_wilk_test(struct sampling *s)
-{
-	double W, p;
-	if (swilk(s->samples, s->n, &W, &p) != 0) {
-		err("swilk failed");
-		return;
-	}
-
-	if (use_machine_output) {
-		printf("%-10s %e\n", "sw_pvalue", p);
-		printf("%-10s %d\n", "sw_normal", (p >= 0.05));
-	} else {
-		const char *msg = (p < 0.05) ? "NOT normal" : "may be normal";
-		printf("    Shapiro-Wilk: W=%.2e, p-value=%.2e (%s)\n", W, p, msg);
-	}
-
-}
-
 static void
 stats(struct sampling *s)
 {
@@ -398,71 +382,239 @@ plot_histogram(struct sampling *s, int w, int h, int cut_outliers)
 	}
 }
 
+// Struct to hold all computed statistics
+struct summary_stats {
+	double mean, stdev, skewness, kurtosis;
+	double xmin, q1, median, q3, xmax;
+	double mad, sem;
+	double rel_stdev, rel_mad, rel_sem, rel_far;
+	long n, far;
+	double wall;
+	double W, p;  // Shapiro-Wilk results
+};
+
+// Helper to create a dynamic string buffer
+struct string_buffer {
+	char *str;
+	int offset;
+	int maxsize;
+};
+
+static struct string_buffer *
+create_string_buffer(void)
+{
+	struct string_buffer *buf = malloc(sizeof(struct string_buffer));
+	if (!buf) {
+		err("malloc failed");
+		exit(1);
+	}
+	
+	buf->str = malloc(STR_SIZE);
+	if (!buf->str) {
+		err("malloc failed");
+		exit(1);
+	}
+	
+	buf->offset = 0;
+	buf->maxsize = STR_SIZE;
+	return buf;
+}
+
 static void
-print_summary(struct sampling *s)
+free_string_buffer(struct string_buffer *buf)
+{
+	if (buf) {
+		free(buf->str);
+		free(buf);
+	}
+}
+
+// Modified snprintf_check to work with string_buffer
+#define snprintf_buf(buf, ...) \
+do { \
+	int ret = snprintf(buf->str + buf->offset, buf->maxsize - buf->offset, __VA_ARGS__); \
+	if (ret >= buf->maxsize - buf->offset) { \
+		int overflow = ret - (buf->maxsize - buf->offset); \
+		int newsize = buf->maxsize + (overflow < STR_SIZE ? STR_SIZE : overflow / STR_SIZE + STR_SIZE); \
+		char *newstr = malloc(sizeof(char) * newsize); \
+		if (newstr == NULL) { \
+			err("malloc failed"); \
+			exit(1); \
+		} \
+		memcpy(newstr, buf->str, buf->offset); \
+		free(buf->str); \
+		buf->str = newstr; \
+		buf->maxsize = newsize; \
+		ret = snprintf(buf->str + buf->offset, buf->maxsize - buf->offset, __VA_ARGS__); \
+	} \
+	buf->offset += ret; \
+} while (0)
+
+static void
+compute_summary_stats(struct sampling *s, struct summary_stats *stats)
 {
 	qsort(s->samples, s->n, sizeof(double), cmp_double);
 
-	double mean = stats_mean(s->samples, s->n);
-	double stdev = stats_stdev(s->samples, s->n, mean);
-	double skewness = stats_skewness(s->samples, s->n, mean, stdev);
-	double kurtosis = stats_kurtosis(s->samples, s->n, mean, stdev);
+	stats->mean = stats_mean(s->samples, s->n);
+	stats->stdev = stats_stdev(s->samples, s->n, stats->mean);
+	stats->skewness = stats_skewness(s->samples, s->n, stats->mean, stats->stdev);
+	stats->kurtosis = stats_kurtosis(s->samples, s->n, stats->mean, stats->stdev);
+
+	stats->xmin = stats_percentile(s->samples, s->n, 0.0);
+	stats->q1 = stats_percentile(s->samples, s->n, 0.25);
+	stats->median = stats_median(s->samples, s->n);
+	stats->q3 = stats_percentile(s->samples, s->n, 0.75);
+	stats->xmax = stats_percentile(s->samples, s->n, 1.0);
+
+	stats->mad = stats_mad(s->samples, s->n, stats->median);
+	stats->n = s->n;
+	stats->far = stats_outliers(s->samples, s->n, stats->q1, stats->q3, 3.0);
+	stats->wall = s->wall;
+
+	stats->sem = stats_sem(stats->stdev, s->n);
+	stats->rel_stdev = stats_percent(stats->stdev, stats->mean);
+	stats->rel_mad = stats_percent(stats->mad, stats->median);
+	stats->rel_sem = stats_percent(stats->sem * 1.96, stats->mean);
+	stats->rel_far = stats_percent(stats->far, s->n);
+
+	if (swilk(s->samples, s->n, &stats->W, &stats->p) != 0) {
+		err("swilk failed");
+		stats->W = NAN;
+		stats->p = NAN;
+	}
+}
 
-	double xmin = stats_percentile(s->samples, s->n, 0.0);
-	double q1 = stats_percentile(s->samples, s->n, 0.25);
-	double median = stats_median(s->samples, s->n);
-	double q3 = stats_percentile(s->samples, s->n, 0.75);
-	double xmax = stats_percentile(s->samples, s->n, 1.0);
+static void
+print_machine_output(struct summary_stats *stats, struct sampling *s)
+{
+	struct string_buffer *buf = create_string_buffer();
+
+	snprintf_buf(buf, "%-10s %e\n", "min", stats->xmin);
+	snprintf_buf(buf, "%-10s %e\n", "q1", stats->q1);
+	snprintf_buf(buf, "%-10s %e\n", "median", stats->median);
+	snprintf_buf(buf, "%-10s %e\n", "mean", stats->mean);
+	snprintf_buf(buf, "%-10s %e\n", "q3", stats->q3);
+	snprintf_buf(buf, "%-10s %e\n", "max", stats->xmax);
+	snprintf_buf(buf, "%-10s %ld\n", "samples", stats->n);
+	snprintf_buf(buf, "%-10s %e\n", "wall", stats->wall);
+	snprintf_buf(buf, "%-10s %e\n", "mad", stats->mad);
+	snprintf_buf(buf, "%-10s %e\n", "stdev", stats->stdev);
+	snprintf_buf(buf, "%-10s %e\n", "skewness", stats->skewness);
+	snprintf_buf(buf, "%-10s %e\n", "kurtosis", stats->kurtosis);
+	snprintf_buf(buf, "%-10s %ld\n", "far", stats->far);
+	snprintf_buf(buf, "%-10s %e\n", "rfar", stats->rel_far);
+	snprintf_buf(buf, "%-10s %e\n", "rmad", stats->rel_mad);
+	snprintf_buf(buf, "%-10s %e\n", "rstdev", stats->rel_stdev);
+	snprintf_buf(buf, "%-10s %e\n", "sem", stats->sem);
+	snprintf_buf(buf, "%-10s %e\n", "rsem", stats->rel_sem);
+	snprintf_buf(buf, "%-10s %e\n", "sw_pvalue", stats->p);
+	snprintf_buf(buf, "%-10s %d\n", "sw_normal", (stats->p >= 0.05));
+
+	const char *runmode = read_from_stdin ? "stdin" : "exec";
+	snprintf_buf(buf, "%-10s %s\n", "runmode", runmode);
 
-	double mad = stats_mad(s->samples, s->n, median);
-	long n = s->n;
-	long far = stats_outliers(s->samples, s->n, q1, q3, 3.0);
+	if (!read_from_stdin) {
+		snprintf_buf(buf, "%-10s ", "command");
+		for (char *const* p = s->cmd; *p; p++) {
+			snprintf_buf(buf, "%s ", *p);
+		}
+		snprintf_buf(buf, "\n");
+	}
 
-	double sem = stats_sem(stdev, s->n);
-	double rel_stdev = stats_percent(stdev, mean);
-	double rel_mad = stats_percent(mad, median);
-	double rel_sem = stats_percent(sem * 1.96, mean);
-	double rel_far = stats_percent(far, s->n);
+	printf("%s\n", buf->str);
+	free_string_buffer(buf);
+}
 
-	if (use_machine_output) {
-		printf("%-10s %e\n", "min", xmin);
-		printf("%-10s %e\n", "q1", q1);
-		printf("%-10s %e\n", "median", median);
-		printf("%-10s %e\n", "mean", mean);
-		printf("%-10s %e\n", "q3", q3);
-		printf("%-10s %e\n", "max", xmax);
-		printf("%-10s %ld\n", "samples", n);
-		printf("%-10s %e\n", "wall", s->wall);
-		printf("%-10s %e\n", "mad", mad);
-		printf("%-10s %e\n", "stdev", stdev);
-		printf("%-10s %e\n", "skewness", skewness);
-		printf("%-10s %e\n", "kurtosis", kurtosis);
-		printf("%-10s %ld\n", "far", far);
-		printf("%-10s %e\n", "rfar", rel_far);
-		printf("%-10s %e\n", "rmad", rel_mad);
-		printf("%-10s %e\n", "rstdev", rel_stdev);
-		printf("%-10s %e\n", "sem", sem);
-		printf("%-10s %e\n", "rsem", rel_sem);
+static void
+print_csv_output(struct summary_stats *stats, struct sampling *s)
+{
+	struct string_buffer *header = create_string_buffer();
+	struct string_buffer *values = create_string_buffer();
+
+	snprintf_buf(header, "median,mean,min,q1,q3,max,samples,wall,mad,stdev,skewness,kurtosis,far,rfar,rmad,rstdev,sem,rsem");
+	snprintf_buf(values, "%e,%e,%e,%e,%e,%e,%ld,%e,%e,%e,%e,%e,%ld,%e,%e,%e,%e,%e",
+			stats->median, stats->mean, stats->xmin, stats->q1, stats->q3, stats->xmax, 
+			stats->n, stats->wall, stats->mad, stats->stdev, stats->skewness, stats->kurtosis,
+			stats->far, stats->rel_far, stats->rel_mad, stats->rel_stdev, stats->sem, stats->rel_sem);
+
+	snprintf_buf(header, ",sw_pvalue,sw_normal");
+	snprintf_buf(values, ",%e,%d", stats->p, (stats->p >= 0.05));
+
+	const char *runmode = read_from_stdin ? "stdin" : "exec";
+	snprintf_buf(header, ",runmode");
+	snprintf_buf(values, ",%s", runmode);
+
+	if (!read_from_stdin) {
+		for (char *const* p = s->cmd; *p; p++) {
+			if (p == s->cmd) {
+				snprintf_buf(header, ",command");
+			} else {
+				snprintf_buf(header, ",arg%ld", p - s->cmd);
+			}
+			snprintf_buf(values, ",%s", *p);
+		}
+	}
+
+	printf("%s\n%s\n", header->str, values->str);
+	free_string_buffer(header);
+	free_string_buffer(values);
+}
+
+static void
+print_human_output(struct summary_stats *stats, struct sampling *s)
+{
+	struct string_buffer *buf = create_string_buffer();
+
+	snprintf_buf(buf, "\n");
+	snprintf_buf(buf, "%10s %10s %10s %10s %10s %10s\n",
+			"MIN", "Q1", "MEDIAN", "MEAN", "Q3", "MAX");
+	snprintf_buf(buf, "% 10.3e % 10.3e % 10.3e % 10.3e % 10.3e % 10.3e \n",
+			stats->xmin, stats->q1, stats->median, stats->mean, stats->q3, stats->xmax);
+	
+	snprintf_buf(buf, "\n");
+	snprintf_buf(buf, "%10s %10s %10s %10s %10s %10s\n",
+			"N", "WALL", "MAD", "STDEV", "SKEW", "KURTOSIS");
+	snprintf_buf(buf, "%10ld %10.1f % 10.3e % 10.3e % 10.3e % 10.3e\n",
+			stats->n, stats->wall, stats->mad, stats->stdev, stats->skewness, stats->kurtosis);
+	
+	snprintf_buf(buf, "\n");
+	snprintf_buf(buf, "%10s %10s %10s %10s %10s %10s\n",
+			"FAR", "%FAR", "%MAD", "%STDEV", "SEM", "%SEM");
+	snprintf_buf(buf, "%10ld %10.2f % 10.2f % 10.2f % 10.3e % 10.2f\n",
+			stats->far, stats->rel_far, stats->rel_mad, stats->rel_stdev, stats->sem, stats->rel_sem);
+	
+	snprintf_buf(buf, "\n");
+	const char *msg = (stats->p < 0.05) ? "NOT normal" : "may be normal";
+	snprintf_buf(buf, "    Shapiro-Wilk: W=%.2e, p-value=%.2e (%s)\n", stats->W, stats->p, msg);
+
+	if (read_from_stdin) {
+		snprintf_buf(buf, "    Read from stdin\n");
 	} else {
-		printf("\n");
-		printf("%10s %10s %10s %10s %10s %10s\n",
-				"MIN", "Q1", "MEDIAN", "MEAN", "Q3", "MAX");
-		printf("% 10.3e % 10.3e % 10.3e % 10.3e % 10.3e % 10.3e \n",
-				xmin, q1, median, mean, q3, xmax);
-		printf("\n");
-		printf("%10s %10s %10s %10s %10s %10s\n",
-				"N", "WALL", "MAD", "STDEV", "SKEW", "KURTOSIS");
-		printf("%10ld %10.1f % 10.3e % 10.3e % 10.3e % 10.3e\n",
-				n, s->wall, mad, stdev, skewness, kurtosis);
-		printf("\n");
-		printf("%10s %10s %10s %10s %10s %10s\n",
-				"FAR", "%FAR", "%MAD", "%STDEV", "SEM", "%SEM");
-		printf("%10ld %10.2f % 10.2f % 10.2f % 10.3e % 10.2f\n",
-				far, rel_far, rel_mad, rel_stdev, sem, rel_sem);
-		printf("\n");
+		snprintf_buf(buf, "    Cmd: ");
+		for (char *const* p = s->cmd; *p; p++) {
+			snprintf_buf(buf, "%s ", *p);
+		}
+		snprintf_buf(buf, "\n");
 	}
+
+	printf("%s", buf->str);
+	free_string_buffer(buf);
 }
 
+static void
+print_summary(struct sampling *s)
+{
+	struct summary_stats stats;
+	compute_summary_stats(s, &stats);
+
+	if (use_machine_output) {
+		print_machine_output(&stats, s);
+	} else if (use_csv_output) {
+		print_csv_output(&stats, s);
+	} else {
+		print_human_output(&stats, s);
+	}
+}
 
 /* Return -1 on error, 0 on success */
 static int
@@ -491,34 +643,6 @@ do_read(FILE *f, double *metric, int *end)
 	return 0;
 }
 
-static void
-print_command(struct sampling *s)
-{
-	if (use_machine_output) {
-		const char *runmode = read_from_stdin ? "stdin" : "exec";
-		printf("%-10s %s\n", "runmode", runmode);
-
-		if (!read_from_stdin) {
-			printf("%-10s ", "command");
-			for (char *const* p = s->cmd; *p; p++) {
-				printf("%s ", *p);
-			}
-			printf("\n");
-		}
-		return;
-	}
-
-	if (read_from_stdin) {
-		printf("    Read from stdin\n");
-	} else {
-		printf("    Cmd: ");
-		for (char *const* p = s->cmd; *p; p++) {
-			printf("%s ", *p);
-		}
-		printf("\n");
-	}
-}
-
 static int
 do_sample(char * const cmd[], char * const argv[], int argc)
 {
@@ -576,11 +700,8 @@ do_sample(char * const cmd[], char * const argv[], int argc)
 	}
 
 	print_summary(&s);
-	print_command(&s);
 
-	shapiro_wilk_test(&s);
-
-	if (!use_machine_output) {
+	if (!use_machine_output && !use_csv_output) {
 		printf("\n"); /* Leave one empty before histogram */
 		plot_histogram(&s, 64, 4, trim_outliers);
 		printf("\n"); /* Leave one empty after histogram */
@@ -607,7 +728,7 @@ main(int argc, char *argv[])
 	progname_set("bigotes");
 	int opt;
 
-	while ((opt = getopt(argc, argv, "imn:wo:qhXt:")) != -1) {
+	while ((opt = getopt(argc, argv, "cimn:wo:qhXt:")) != -1) {
 		switch (opt) {
 			case 'i':
 				read_from_stdin = 1;
@@ -624,6 +745,9 @@ main(int argc, char *argv[])
 			case 'm':
 				use_machine_output = 1;
 				break;
+			case 'c':
+				use_csv_output = 1;
+				break;
 			case 'n':
 				min_samples = atol(optarg);
 				break;