index : cgit

/* ui-stats.c: generate stats view
 *
 * Copyright (C) 2006-2014 cgit Development Team <cgit@lists.zx2c4.com>
 *
 * Licensed under GNU General Public License v2
 *   (see LICENSE.txt for full license text)
 */

#define USE_THE_REPOSITORY_VARIABLE

#include "cgit.h"
#include "ui-stats.h"
#include "html.h"
#include "ui-shared.h"

struct authorstat {
	long total;
	struct string_list list;
};

#define DAY_SECS (60 * 60 * 24)
#define WEEK_SECS (DAY_SECS * 7)

static void trunc_week(struct tm *tm)
{
	time_t t = timegm(tm);
	t -= ((tm->tm_wday + 6) % 7) * DAY_SECS;
	gmtime_r(&t, tm);
}

static void dec_week(struct tm *tm)
{
	time_t t = timegm(tm);
	t -= WEEK_SECS;
	gmtime_r(&t, tm);
}

static void inc_week(struct tm *tm)
{
	time_t t = timegm(tm);
	t += WEEK_SECS;
	gmtime_r(&t, tm);
}

static char *pretty_week(struct tm *tm)
{
	static char buf[10];

	strftime(buf, sizeof(buf), "W%V %G", tm);
	return buf;
}

static void trunc_month(struct tm *tm)
{
	tm->tm_mday = 1;
}

static void dec_month(struct tm *tm)
{
	tm->tm_mon--;
	if (tm->tm_mon < 0) {
		tm->tm_year--;
		tm->tm_mon = 11;
	}
}

static void inc_month(struct tm *tm)
{
	tm->tm_mon++;
	if (tm->tm_mon > 11) {
		tm->tm_year++;
		tm->tm_mon = 0;
	}
}

static char *pretty_month(struct tm *tm)
{
	static const char *months[] = {
		"Jan", "Feb", "Mar", "Apr", "May", "Jun",
		"Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
	};
	return cgit_fmt("%s %d", months[tm->tm_mon], tm->tm_year + 1900);
}

static void trunc_quarter(struct tm *tm)
{
	trunc_month(tm);
	while (tm->tm_mon % 3 != 0)
		dec_month(tm);
}

static void dec_quarter(struct tm *tm)
{
	dec_month(tm);
	dec_month(tm);
	dec_month(tm);
}

static void inc_quarter(struct tm *tm)
{
	inc_month(tm);
	inc_month(tm);
	inc_month(tm);
}

static char *pretty_quarter(struct tm *tm)
{
	return cgit_fmt("Q%d %d", tm->tm_mon / 3 + 1, tm->tm_year + 1900);
}

static void trunc_year(struct tm *tm)
{
	trunc_month(tm);
	tm->tm_mon = 0;
}

static void dec_year(struct tm *tm)
{
	tm->tm_year--;
}

static void inc_year(struct tm *tm)
{
	tm->tm_year++;
}

static char *pretty_year(struct tm *tm)
{
	return cgit_fmt("%d", tm->tm_year + 1900);
}

static const struct cgit_period periods[] = {
	{'w', "week", 12, 4, trunc_week, dec_week, inc_week, pretty_week},
	{'m', "month", 12, 4, trunc_month, dec_month, inc_month, pretty_month},
	{'q', "quarter", 12, 4, trunc_quarter, dec_quarter, inc_quarter, pretty_quarter},
	{'y', "year", 12, 4, trunc_year, dec_year, inc_year, pretty_year},
};

/* Given a period code or name, return a period index (1, 2, 3 or 4)
 * and update the period pointer to the correcsponding struct.
 * If no matching code is found, return 0.
 */
int cgit_find_stats_period(const char *expr, const struct cgit_period **period)
{
	size_t i;
	char code = '\0';

	if (!expr)
		return 0;

	if (strlen(expr) == 1)
		code = expr[0];

	for (i = 0; i < ARRAY_SIZE(periods); i++)
		if (periods[i].code == code || !strcmp(periods[i].name, expr)) {
			if (period)
				*period = &periods[i];
			return i + 1;
		}
	return 0;
}

const char *cgit_find_stats_periodname(int idx)
{
	if (idx > 0 && idx <= (int)ARRAY_SIZE(periods))
		return periods[idx - 1].name;
	else
		return "";
}

static void add_commit(struct string_list *authors, struct commitinfo *info,
	const struct cgit_period *period)
{
	struct string_list_item *author, *item;
	struct authorstat *authorstat;
	struct string_list *items;
	char *tmp;
	struct tm date;
	time_t t;
	uintptr_t *counter;

	/* A commit can lack an author header, so fall back rather than
	 * xstrdup(NULL). */
	tmp = xstrdup(info->author ? info->author : "(unknown)");
	author = string_list_insert(authors, tmp);
	if (!author->util)
		author->util = xcalloc(1, sizeof(struct authorstat));
	else
		free(tmp);
	authorstat = author->util;
	items = &authorstat->list;
	t = info->committer_date;
	// A crafted commit can carry a date outside the range gmtime_r can
	// represent, which would leave date uninitialized and later index
	// the month table out of bounds. Drop such a commit from the stats.
	if (!gmtime_r(&t, &date))
		return;
	period->trunc(&date);
	tmp = xstrdup(period->pretty(&date));
	item = string_list_insert(items, tmp);
	counter = (uintptr_t *)&item->util;
	if (*counter)
		free(tmp);
	(*counter)++;

	authorstat->total++;
}

static int cmp_total_commits(const void *a1, const void *a2)
{
	const struct string_list_item *i1 = a1;
	const struct string_list_item *i2 = a2;
	const struct authorstat *auth1 = i1->util;
	const struct authorstat *auth2 = i2->util;

	// Return the sign only, since a long difference truncated to int could
	// flip and leave the comparator inconsistent.
	if (auth2->total > auth1->total)
		return 1;
	if (auth2->total < auth1->total)
		return -1;
	return 0;
}

/* Walk the commit DAG once for the configured window. One walk, no
 * diffs, so the cost class matches rev-list and the page cache absorbs
 * repeat views. Merge commits are skipped.
 */
static struct string_list collect_stats(const struct cgit_period *period)
{
	struct string_list authors;
	struct rev_info rev;
	struct commit *commit;
	const char *argv[] = {NULL, ctx.qry.head, NULL, NULL};
	int argc = 2;
	time_t now, since;
	long i;
	struct tm tm;

	time(&now);
	gmtime_r(&now, &tm);
	period->trunc(&tm);
	for (i = 1; i < period->count; i++)
		period->dec(&tm);
	since = timegm(&tm);

	if (ctx.qry.path) {
		argv[2] = "--";
		argv[3] = ctx.qry.path;
		argc += 2;
	}
	repo_init_revisions(the_repository, &rev, NULL);
	rev.abbrev = DEFAULT_ABBREV;
	rev.commit_format = CMIT_FMT_DEFAULT;
	rev.max_parents = 1;
	rev.verbose_header = 1;
	rev.show_root_diff = 0;
	setup_revisions(argc, argv, &rev, NULL);
	// Prune the walk to the displayed window instead of traversing the
	// whole history and discarding older commits. The in-process check
	// below still bounds the period edge exactly.
	rev.max_age = since;
	prepare_revision_walk(&rev);
	memset(&authors, 0, sizeof(authors));
	while ((commit = get_revision(&rev)) != NULL) {
		struct commitinfo *info = cgit_parse_commit(commit);

		if ((time_t)info->committer_date >= since)
			add_commit(&authors, info, period);

		cgit_free_commitinfo(info);
		release_commit_memory(the_repository->parsed_objects, commit);
		commit->parents = NULL;
	}
	return authors;
}

/*
 * The column labels for the displayed window, oldest first. Every table below
 * walks the same period, and each step of that walk converts a struct tm to a
 * time_t and back, so the labels are worked out once and shared. It also frees
 * the callers from pretty() handing back a buffer it reuses.
 */
static struct string_list build_period_labels(const struct cgit_period *period)
{
	struct string_list labels = STRING_LIST_INIT_DUP;
	struct tm tm;
	time_t now;
	int i;

	time(&now);
	gmtime_r(&now, &tm);
	period->trunc(&tm);
	for (i = 1; i < period->count; i++)
		period->dec(&tm);
	for (i = 0; i < period->count; i++) {
		string_list_append(&labels, period->pretty(&tm));
		period->inc(&tm);
	}
	return labels;
}

/* Sum one row across a run of authors. Taking a count rather than an end index
 * keeps an empty author list from describing a run that wraps. */
static void print_combined_authorrow(struct string_list *authors, size_t from,
				     size_t count, const char *name,
				     const char *leftclass,
				     const char *centerclass,
				     const char *rightclass,
				     const struct string_list *labels)
{
	struct string_list_item *author;
	struct authorstat *authorstat;
	struct string_list *items;
	struct string_list_item *date;
	size_t i, j;
	long total, subtotal;

	total = 0;
	htmlf("<tr><td class='%s'>%s</td>", leftclass,
		cgit_fmt(name, (long)count));
	for (j = 0; j < labels->nr; j++) {
		subtotal = 0;
		for (i = from; i < from + count; i++) {
			author = &authors->items[i];
			authorstat = author->util;
			items = &authorstat->list;
			date = string_list_lookup(items, labels->items[j].string);
			if (date)
				subtotal += (uintptr_t)date->util;
		}
		htmlf("<td class='%s'>%ld</td>", centerclass, subtotal);
		total += subtotal;
	}
	htmlf("<td class='%s'>%ld</td></tr>", rightclass, total);
}

static void print_authors(struct string_list *authors, int requested_top,
			  const struct string_list *labels)
{
	struct string_list_item *author;
	struct authorstat *authorstat;
	struct string_list *items;
	struct string_list_item *date;
	size_t i, j, top;
	long total;

	html("<table class='stats'><tr><th>Author</th>");
	for (j = 0; j < labels->nr; j++)
		htmlf("<th>%s</th>", labels->items[j].string);
	html("<th>Total</th></tr>\n");

	// The author count arrives through ofs, which carries -1 for "all".
	top = (requested_top <= 0 || (size_t)requested_top > authors->nr)
		? authors->nr : (size_t)requested_top;

	for (i = 0; i < top; i++) {
		author = &authors->items[i];
		html("<tr><td class='left'>");
		html_txt(author->string);
		html("</td>");
		authorstat = author->util;
		items = &authorstat->list;
		total = 0;
		for (j = 0; j < labels->nr; j++) {
			date = string_list_lookup(items, labels->items[j].string);
			if (!date)
				html("<td>0</td>");
			else {
				htmlf("<td>%lu</td>", (uintptr_t)date->util);
				total += (uintptr_t)date->util;
			}
		}
		htmlf("<td class='sum'>%ld</td></tr>", total);
	}

	if (top < authors->nr)
		print_combined_authorrow(authors, top, authors->nr - top,
			"Others (%ld)", "left", "", "sum", labels);

	print_combined_authorrow(authors, 0, authors->nr, "Total",
		"total", "sum", "sum", labels);
	html("</table>");
}

/* Bytes at HEAD per language, judged by file extension. One recursive
 * tree read, sizes come from object headers without loading content. */
static const struct {
	const char *ext;
	const char *label;
} lang_map[] = {
	{"c", "C"}, {"h", "C"},
	{"cpp", "C++"}, {"cc", "C++"}, {"cxx", "C++"},
	{"hpp", "C++"}, {"hh", "C++"},
	{"js", "JavaScript"}, {"mjs", "JavaScript"},
	{"ts", "TypeScript"}, {"tsx", "TypeScript"},
	{"py", "Python"}, {"lua", "Lua"},
	{"sh", "Shell"}, {"bash", "Shell"},
	{"go", "Go"}, {"rs", "Rust"}, {"zig", "Zig"},
	{"java", "Java"}, {"kt", "Kotlin"}, {"cs", "C#"}, {"swift", "Swift"},
	{"rb", "Ruby"}, {"pl", "Perl"}, {"pm", "Perl"}, {"php", "PHP"},
	{"hs", "Haskell"}, {"el", "Lisp"}, {"ml", "OCaml"},
	{"css", "CSS"}, {"scss", "CSS"},
	{"html", "HTML"}, {"htm", "HTML"}, {"xml", "XML"},
	{"md", "Markdown"}, {"rst", "Text"}, {"txt", "Text"},
	{"json", "JSON"}, {"yml", "YAML"}, {"yaml", "YAML"}, {"toml", "TOML"},
	{"mk", "Make"}, {"tex", "TeX"}, {"sql", "SQL"}, {"vim", "Vimscript"},
};

struct lang_walk_ctx {
	struct string_list langs;
	unsigned long total;
};

static int lang_walk_cb(const struct object_id *oid, struct strbuf *base,
			const char *pathname, unsigned mode, void *cbdata)
{
	struct lang_walk_ctx *lw = cbdata;
	struct string_list_item *item;
	const char *ext, *label = NULL;
	unsigned long size;
	int i;

	if (S_ISDIR(mode))
		return READ_TREE_RECURSIVE;
	if (!S_ISREG(mode))
		return 0;
	if (odb_read_object_info(the_repository->objects, oid, &size) != OBJ_BLOB
	    || !size)
		return 0;

	ext = strrchr(pathname, '.');
	if (ext && ext != pathname && ext[1]) {
		for (i = 0; i < (int)ARRAY_SIZE(lang_map); i++)
			if (!strcasecmp(ext + 1, lang_map[i].ext)) {
				label = lang_map[i].label;
				break;
			}
	} else if (!strcmp(pathname, "Makefile")) {
		label = "Make";
	}
	if (!label)
		label = "Other";
	item = string_list_insert(&lw->langs, label);
	item->util = (void *)((uintptr_t)item->util + size);
	lw->total += size;
	return 0;
}

static int cmp_lang_bytes(const void *a1, const void *a2)
{
	const struct string_list_item *i1 = a1;
	const struct string_list_item *i2 = a2;
	uintptr_t b1 = (uintptr_t)i1->util;
	uintptr_t b2 = (uintptr_t)i2->util;

	return b1 < b2 ? 1 : b1 > b2 ? -1 : 0;
}

static void summarize_tree(struct lang_walk_ctx *lw)
{
	struct pathspec paths = { .nr = 0 };
	struct object_id oid;
	struct commit *commit;

	if (repo_get_oid(the_repository, ctx.qry.head, &oid))
		return;
	commit = lookup_commit_reference(the_repository, &oid);
	if (!commit || repo_parse_commit(the_repository, commit))
		return;

	read_tree(the_repository, repo_get_commit_tree(the_repository, commit),
		  &paths, lang_walk_cb, lw);
	qsort(lw->langs.items, lw->langs.nr, sizeof(struct string_list_item),
	      cmp_lang_bytes);
}

static void print_language_row(const char *label, unsigned long bytes,
			       unsigned long total)
{
	struct strbuf size = STRBUF_INIT;

	strbuf_humanise_bytes(&size, bytes);
	html("<tr><td class='left'>");
	html_txt(label);
	html("</td><td>");
	html_txt(size.buf);
	htmlf("</td><td>%.1f%%</td></tr>\n", 100.0 * bytes / total);
	strbuf_release(&size);
}

/* Bordered like the commits-per-author table. */
static void print_languages(struct lang_walk_ctx *lw)
{
	unsigned long other;
	size_t i;
	int shown;

	if (!lw->total)
		return;

	/* Show up to six named rows, everything else folds into an
	 * Other row at the end. */
	other = 0;
	shown = 0;
	for (i = 0; i < lw->langs.nr; i++) {
		if (shown < 6 && strcmp(lw->langs.items[i].string, "Other")) {
			shown++;
			continue;
		}
		other += (uintptr_t)lw->langs.items[i].util;
	}

	html("<h2>Languages</h2>");
	html("<table class='stats'>");
	html("<tr><th>Language</th><th>Size</th><th>Share</th></tr>\n");
	shown = 0;
	for (i = 0; i < lw->langs.nr && shown < 6; i++) {
		if (!strcmp(lw->langs.items[i].string, "Other"))
			continue;
		print_language_row(lw->langs.items[i].string,
				   (uintptr_t)lw->langs.items[i].util,
				   lw->total);
		shown++;
	}
	if (other)
		print_language_row("Other", other, lw->total);
	html("</table>");
}

/* Create a sorted string_list with one entry per author. The util-field
 * for each author is another string_list which is used to calculate the
 * number of commits per time-interval.
 */
void cgit_show_stats(void)
{
	struct string_list authors;
	struct lang_walk_ctx lw;
	const struct cgit_period *period;
	int top, i;
	const char *code = "w";

	if (ctx.qry.period)
		code = ctx.qry.period;

	i = cgit_find_stats_period(code, &period);
	if (!i) {
		cgit_print_error_page(404, "Not found",
			"Unknown statistics type: %c", code[0]);
		return;
	}
	if (ctx.repo->max_stats && i > ctx.repo->max_stats) {
		cgit_print_error_page(400, "Bad request",
			"Statistics type disabled: %s", period->name);
		return;
	}
	/* Walk the tree before the history walk. Releasing commit memory
	 * during that walk resets each commit's slab index, and a later
	 * lookup through the commit graph would then read another
	 * commit's slot and walk the wrong tree. */
	memset(&lw, 0, sizeof(lw));
	summarize_tree(&lw);

	authors = collect_stats(period);
	qsort(authors.items, authors.nr, sizeof(struct string_list_item),
		cmp_total_commits);

	top = ctx.qry.ofs;
	if (!top)
		top = 10;

	cgit_print_layout_start();

	/* The options panel floats right of the page top, the same spot
	 * the diff controls occupy on the diff pages. */
	html("<div class='cgit-panel'>");
	html("<b>stat options</b>");
	html("<form method='get'>");
	cgit_add_hidden_formfields(1, 0, "stats");
	html("<table><tr><td colspan='2'/></tr>");
	if (!ctx.repo->max_stats || ctx.repo->max_stats > 1) {
		int nperiods = ctx.repo->max_stats ?
			ctx.repo->max_stats : (int)ARRAY_SIZE(periods);
		html("<tr><td class='label'>Period:</td>");
		html("<td class='ctrl'><select name='period' data-autosubmit='1'>");
		for (i = 0; i < nperiods; i++)
			html_option(cgit_fmt("%c", periods[i].code),
				    periods[i].name, cgit_fmt("%c", period->code));
		html("</select></td></tr>");
	}
	html("<tr><td class='label'>Authors:</td>");
	html("<td class='ctrl'><select name='ofs' data-autosubmit='1'>");
	html_intoption(10, "10", top);
	html_intoption(25, "25", top);
	html_intoption(50, "50", top);
	html_intoption(100, "100", top);
	html_intoption(-1, "all", top);
	html("</select></td></tr>");
	html("<tr><td/><td class='ctrl'>");
	html("<noscript><input type='submit' value='Reload'/></noscript>");
	html("</td></tr></table>");
	html("</form>");
	html("</div>");

	htmlf("<h2>Commits per author per %s", period->name);
	if (ctx.qry.path) {
		html(" (path '");
		html_txt(ctx.qry.path);
		html("')");
	}
	html("</h2>");
	{
		struct string_list labels = build_period_labels(period);
		print_authors(&authors, top, &labels);
		string_list_clear(&labels, 0);
	}

	print_languages(&lw);
	string_list_clear(&lw.langs, 0);

	cgit_print_layout_end();
}