log: allow users to follow a file

Teach the "log" UI to behave in the same way as "git log --follow", when
given a suitable instruction by the user.  The default behaviour remains
to show the log without following renames, but the follow behaviour can
be activated by following a link in the page header.

Follow is not the default because outputting merges in follow mode is
tricky ("git log --follow" will not show merges).  We also disable the
graph in follow mode because the commit graph is not simplified so we
end up with frequent gaps in the graph and many lines that do not
connect with any commits we're actually showing.

We also teach the "diff" and "commit" UIs to respect the follow flag on
URLs, causing the single-file version of these UIs to detect renames.
This feature is needed only for commits that rename the path we're
interested in.

For commits before the file has been renamed (i.e. that appear later in
the log list) we change the file path in the links from the log to point
to the old name; this means that links to commits always limit by the
path known to that commit.  If we didn't do this we would need to walk
down the log diff'ing every commit whenever we want to show a commit.
The drawback is that the "Log" link in the top bar of such a page links
to the log limited by the old name, so it will only show pre-rename
commits.  I consider this a reasonable trade-off since the "Back" button
still works and the log matches the path displayed in the top bar.

Since following renames requires running diff on every commit we
consider, I've added a knob to the configuration file to globally
enable/disable this feature.  Note that we may consider a large number
of commits the revision walking machinery no longer performs any path
limitation so we have to examine every commit until we find a page full
of commits that affect the target path or something related to it.

Suggested-by: René Neumann <necoro@necoro.eu>
Signed-off-by: John Keeping <john@keeping.me.uk>

10 files changed, 194 insertions, 18 deletions

diff --git a/cgit.c b/cgit.c
index baad1c8..d84b4be 100644
--- a/cgit.c
+++ b/cgit.c
@@ -152,6 +152,8 @@ static void config_cb(const char *name, const char *value)
 		ctx.cfg.snapshots = cgit_parse_snapshots_mask(value);
 	else if (!strcmp(name, "enable-filter-overrides"))
 		ctx.cfg.enable_filter_overrides = atoi(value);
+	else if (!strcmp(name, "enable-follow-links"))
+		ctx.cfg.enable_follow_links = atoi(value);
 	else if (!strcmp(name, "enable-http-clone"))
 		ctx.cfg.enable_http_clone = atoi(value);
 	else if (!strcmp(name, "enable-index-links"))
@@ -333,6 +335,8 @@ static void querystring_cb(const char *name, const char *value)
 		ctx.qry.context = atoi(value);
 	} else if (!strcmp(name, "ignorews")) {
 		ctx.qry.ignorews = atoi(value);
+	} else if (!strcmp(name, "follow")) {
+		ctx.qry.follow = atoi(value);
 	}
 }
 
diff --git a/cgit.h b/cgit.h
index b2253d2..3120562 100644
--- a/cgit.h
+++ b/cgit.h
@@ -179,6 +179,7 @@ struct cgit_query {
 	int show_all;
 	int context;
 	int ignorews;
+	int follow;
 	char *vpath;
 };
 
@@ -221,6 +222,7 @@ struct cgit_config {
 	int case_sensitive_sort;
 	int embedded;
 	int enable_filter_overrides;
+	int enable_follow_links;
 	int enable_http_clone;
 	int enable_index_links;
 	int enable_index_owner;
diff --git a/cgitrc.5.txt b/cgitrc.5.txt
index e21ece9..759f353 100644
--- a/cgitrc.5.txt
+++ b/cgitrc.5.txt
@@ -150,6 +150,10 @@ enable-filter-overrides::
 	Flag which, when set to "1", allows all filter settings to be
 	overridden in repository-specific cgitrc files. Default value: none.
 
+enable-follow-links::
+	Flag which, when set to "1", allows users to follow a file in the log
+	view.  Default value: "0".
+
 enable-http-clone::
 	If set to "1", cgit will act as an dumb HTTP endpoint for git clones.
 	You can add "http://$HTTP_HOST$SCRIPT_NAME/$CGIT_REPO_URL" to clone-url
diff --git a/ui-diff.c b/ui-diff.c
index 1cf2ce0..caebd5d 100644
--- a/ui-diff.c
+++ b/ui-diff.c
@@ -36,6 +36,7 @@ static struct fileinfo {
 
 static int use_ssdiff = 0;
 static struct diff_filepair *current_filepair;
+static const char *current_prefix;
 
 struct diff_filespec *cgit_get_current_old_file(void)
 {
@@ -132,11 +133,30 @@ static void count_diff_lines(char *line, int len)
 	}
 }
 
+static int show_filepair(struct diff_filepair *pair)
+{
+	/* Always show if we have no limiting prefix. */
+	if (!current_prefix)
+		return 1;
+
+	/* Show if either path in the pair begins with the prefix. */
+	if (starts_with(pair->one->path, current_prefix) ||
+	    starts_with(pair->two->path, current_prefix))
+		return 1;
+
+	/* Otherwise we don't want to show this filepair. */
+	return 0;
+}
+
 static void inspect_filepair(struct diff_filepair *pair)
 {
 	int binary = 0;
 	unsigned long old_size = 0;
 	unsigned long new_size = 0;
+
+	if (!show_filepair(pair))
+		return;
+
 	files++;
 	lines_added = 0;
 	lines_removed = 0;
@@ -279,6 +299,9 @@ static void filepair_cb(struct diff_filepair *pair)
 	int binary = 0;
 	linediff_fn print_line_fn = print_line;
 
+	if (!show_filepair(pair))
+		return;
+
 	current_filepair = pair;
 	if (use_ssdiff) {
 		cgit_ssdiff_header_begin();
@@ -365,6 +388,18 @@ void cgit_print_diff(const char *new_rev, const char *old_rev,
 	const unsigned char *old_tree_sha1, *new_tree_sha1;
 	diff_type difftype;
 
+	/*
+	 * If "follow" is set then the diff machinery needs to examine the
+	 * entire commit to detect renames so we must limit the paths in our
+	 * own callbacks and not pass the prefix to the diff machinery.
+	 */
+	if (ctx.qry.follow && ctx.cfg.enable_follow_links) {
+		current_prefix = prefix;
+		prefix = "";
+	} else {
+		current_prefix = NULL;
+	}
+
 	if (!new_rev)
 		new_rev = ctx.qry.head;
 	if (get_sha1(new_rev, new_rev_sha1)) {
diff --git a/ui-log.c b/ui-log.c
index 8028b27..ff832ce 100644
--- a/ui-log.c
+++ b/ui-log.c
@@ -12,7 +12,7 @@
 #include "ui-shared.h"
 #include "argv-array.h"
 
-static int files, add_lines, rem_lines;
+static int files, add_lines, rem_lines, lines_counted;
 
 /*
  * The list of available column colors in the commit graph.
@@ -67,7 +67,7 @@ void show_commit_decorations(struct commit *commit)
 			strncpy(buf, deco->name + 11, sizeof(buf) - 1);
 			cgit_log_link(buf, NULL, "branch-deco", buf, NULL,
 				      ctx.qry.vpath, 0, NULL, NULL,
-				      ctx.qry.showmsg);
+				      ctx.qry.showmsg, 0);
 		}
 		else if (starts_with(deco->name, "tag: refs/tags/")) {
 			strncpy(buf, deco->name + 15, sizeof(buf) - 1);
@@ -84,7 +84,7 @@ void show_commit_decorations(struct commit *commit)
 			cgit_log_link(buf, NULL, "remote-deco", NULL,
 				      sha1_to_hex(commit->object.sha1),
 				      ctx.qry.vpath, 0, NULL, NULL,
-				      ctx.qry.showmsg);
+				      ctx.qry.showmsg, 0);
 		}
 		else {
 			strncpy(buf, deco->name, sizeof(buf) - 1);
@@ -98,6 +98,74 @@ next:
 	html("</span>");
 }
 
+static void handle_rename(struct diff_filepair *pair)
+{
+	/*
+	 * After we have seen a rename, we generate links to the previous
+	 * name of the file so that commit & diff views get fed the path
+	 * that is correct for the commit they are showing, avoiding the
+	 * need to walk the entire history leading back to every commit we
+	 * show in order detect renames.
+	 */
+	if (0 != strcmp(ctx.qry.vpath, pair->two->path)) {
+		free(ctx.qry.vpath);
+		ctx.qry.vpath = xstrdup(pair->two->path);
+	}
+	inspect_files(pair);
+}
+
+static int show_commit(struct commit *commit, struct rev_info *revs)
+{
+	struct commit_list *parents = commit->parents;
+	struct commit *parent;
+	int found = 0, saved_fmt;
+	unsigned saved_flags = revs->diffopt.flags;
+
+
+	/* Always show if we're not in "follow" mode with a single file. */
+	if (!ctx.qry.follow)
+		return 1;
+
+	/*
+	 * In "follow" mode, we don't show merges.  This is consistent with
+	 * "git log --follow -- <file>".
+	 */
+	if (parents && parents->next)
+		return 0;
+
+	/*
+	 * If this is the root commit, do what rev_info tells us.
+	 */
+	if (!parents)
+		return revs->show_root_diff;
+
+	/* When we get here we have precisely one parent. */
+	parent = parents->item;
+	parse_commit(parent);
+
+	files = 0;
+	add_lines = 0;
+	rem_lines = 0;
+
+	DIFF_OPT_SET(&revs->diffopt, RECURSIVE);
+	diff_tree_sha1(parent->tree->object.sha1,
+		       commit->tree->object.sha1,
+		       "", &revs->diffopt);
+	diffcore_std(&revs->diffopt);
+
+	found = !diff_queue_is_empty();
+	saved_fmt = revs->diffopt.output_format;
+	revs->diffopt.output_format = DIFF_FORMAT_CALLBACK;
+	revs->diffopt.format_callback = cgit_diff_tree_cb;
+	revs->diffopt.format_callback_data = handle_rename;
+	diff_flush(&revs->diffopt);
+	revs->diffopt.output_format = saved_fmt;
+	revs->diffopt.flags = saved_flags;
+
+	lines_counted = 1;
+	return found;
+}
+
 static void print_commit(struct commit *commit, struct rev_info *revs)
 {
 	struct commitinfo *info;
@@ -177,7 +245,8 @@ static void print_commit(struct commit *commit, struct rev_info *revs)
 		cgit_print_age(commit->date, TM_WEEK * 2, FMT_SHORTDATE);
 	}
 
-	if (ctx.repo->enable_log_filecount || ctx.repo->enable_log_linecount) {
+	if (!lines_counted && (ctx.repo->enable_log_filecount ||
+			       ctx.repo->enable_log_linecount)) {
 		files = 0;
 		add_lines = 0;
 		rem_lines = 0;
@@ -325,7 +394,17 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 			}
 		}
 	}
-	if (commit_graph) {
+
+	if (!path || !ctx.cfg.enable_follow_links) {
+		/*
+		 * If we don't have a path, "follow" is a no-op so make sure
+		 * the variable is set to false to avoid needing to check
+		 * both this and whether we have a path everywhere.
+		 */
+		ctx.qry.follow = 0;
+	}
+
+	if (commit_graph && !ctx.qry.follow) {
 		argv_array_push(&rev_argv, "--graph");
 		argv_array_push(&rev_argv, "--color");
 		graph_set_column_colors(column_colors_html,
@@ -337,6 +416,8 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 	else if (commit_sort == 2)
 		argv_array_push(&rev_argv, "--topo-order");
 
+	if (path && ctx.qry.follow)
+		argv_array_push(&rev_argv, "--follow");
 	argv_array_push(&rev_argv, "--");
 	if (path)
 		argv_array_push(&rev_argv, path);
@@ -347,10 +428,17 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 	rev.verbose_header = 1;
 	rev.show_root_diff = 0;
 	rev.ignore_missing = 1;
+	rev.simplify_history = 1;
 	setup_revisions(rev_argv.argc, rev_argv.argv, &rev, NULL);
 	load_ref_decorations(DECORATE_FULL_REFS);
 	rev.show_decorations = 1;
 	rev.grep_filter.regflags |= REG_ICASE;
+
+	rev.diffopt.detect_rename = 1;
+	rev.diffopt.rename_limit = ctx.cfg.renamelimit;
+	if (ctx.qry.ignorews)
+		DIFF_XDL_SET(&rev.diffopt, IGNORE_WHITESPACE);
+
 	compile_grep_patterns(&rev.grep_filter);
 	prepare_revision_walk(&rev);
 
@@ -368,11 +456,12 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 		cgit_log_link(ctx.qry.showmsg ? "Collapse" : "Expand", NULL,
 			      NULL, ctx.qry.head, ctx.qry.sha1,
 			      ctx.qry.vpath, ctx.qry.ofs, ctx.qry.grep,
-			      ctx.qry.search, ctx.qry.showmsg ? 0 : 1);
+			      ctx.qry.search, ctx.qry.showmsg ? 0 : 1,
+			      ctx.qry.follow);
 		html(")");
 	}
 	html("</th><th class='left'>Author</th>");
-	if (commit_graph)
+	if (rev.graph)
 		html("<th class='left'>Age</th>");
 	if (ctx.repo->enable_log_filecount) {
 		html("<th class='left'>Files</th>");
@@ -388,13 +477,30 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 		ofs = 0;
 
 	for (i = 0; i < ofs && (commit = get_revision(&rev)) != NULL; i++) {
+		if (show_commit(commit, &rev))
+			i++;
 		free_commit_buffer(commit);
 		free_commit_list(commit->parents);
 		commit->parents = NULL;
 	}
 
 	for (i = 0; i < cnt && (commit = get_revision(&rev)) != NULL; i++) {
-		print_commit(commit, &rev);
+		/*
+		 * In "follow" mode, we must count the files and lines the
+		 * first time we invoke diff on a given commit, and we need
+		 * to do that to see if the commit touches the path we care
+		 * about, so we do it in show_commit.  Hence we must clear
+		 * lines_counted here.
+		 *
+		 * This has the side effect of avoiding running diff twice
+		 * when we are both following renames and showing file
+		 * and/or line counts.
+		 */
+		lines_counted = 0;
+		if (show_commit(commit, &rev)) {
+			i++;
+			print_commit(commit, &rev);
+		}
 		free_commit_buffer(commit);
 		free_commit_list(commit->parents);
 		commit->parents = NULL;
@@ -406,7 +512,8 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 			cgit_log_link("[prev]", NULL, NULL, ctx.qry.head,
 				      ctx.qry.sha1, ctx.qry.vpath,
 				      ofs - cnt, ctx.qry.grep,
-				      ctx.qry.search, ctx.qry.showmsg);
+				      ctx.qry.search, ctx.qry.showmsg,
+				      ctx.qry.follow);
 			html("</li>");
 		}
 		if ((commit = get_revision(&rev)) != NULL) {
@@ -414,14 +521,16 @@ void cgit_print_log(const char *tip, int ofs, int cnt, char *grep, char *pattern
 			cgit_log_link("[next]", NULL, NULL, ctx.qry.head,
 				      ctx.qry.sha1, ctx.qry.vpath,
 				      ofs + cnt, ctx.qry.grep,
-				      ctx.qry.search, ctx.qry.showmsg);
+				      ctx.qry.search, ctx.qry.showmsg,
+				      ctx.qry.follow);
 			html("</li>");
 		}
 		html("</ul>");
 	} else if ((commit = get_revision(&rev)) != NULL) {
 		htmlf("<tr class='nohover'><td colspan='%d'>", columns);
 		cgit_log_link("[...]", NULL, NULL, ctx.qry.head, NULL,
-			      ctx.qry.vpath, 0, NULL, NULL, ctx.qry.showmsg);
+			      ctx.qry.vpath, 0, NULL, NULL, ctx.qry.showmsg,
+			      ctx.qry.follow);
 		html("</td></tr>\n");
 	}
 
diff --git a/ui-refs.c b/ui-refs.c
index d3d71dd..73a187b 100644
--- a/ui-refs.c
+++ b/ui-refs.c
@@ -63,7 +63,7 @@ static int print_branch(struct refinfo *ref)
 		return 1;
 	html("<tr><td>");
 	cgit_log_link(name, NULL, NULL, name, NULL, NULL, 0, NULL, NULL,
-		      ctx.qry.showmsg);
+		      ctx.qry.showmsg, 0);
 	html("</td><td>");
 
 	if (ref->object->type == OBJ_COMMIT) {
diff --git a/ui-repolist.c b/ui-repolist.c
index 2453a7f..edefc4c 100644
--- a/ui-repolist.c
+++ b/ui-repolist.c
@@ -330,7 +330,7 @@ void cgit_print_repolist(void)
 			html("<td>");
 			cgit_summary_link("summary", NULL, "button", NULL);
 			cgit_log_link("log", NULL, "button", NULL, NULL, NULL,
-				      0, NULL, NULL, ctx.qry.showmsg);
+				      0, NULL, NULL, ctx.qry.showmsg, 0);
 			cgit_tree_link("tree", NULL, "button", NULL, NULL, NULL);
 			html("</td>");
 		}
diff --git a/ui-shared.c b/ui-shared.c
index 4f84b7c..6be0c2e 100644
--- a/ui-shared.c
+++ b/ui-shared.c
@@ -303,7 +303,8 @@ void cgit_plain_link(const char *name, const char *title, const char *class,
 
 void cgit_log_link(const char *name, const char *title, const char *class,
 		   const char *head, const char *rev, const char *path,
-		   int ofs, const char *grep, const char *pattern, int showmsg)
+		   int ofs, const char *grep, const char *pattern, int showmsg,
+		   int follow)
 {
 	char *delim;
 
@@ -332,6 +333,11 @@ void cgit_log_link(const char *name, const char *title, const char *class,
 	if (showmsg) {
 		html(delim);
 		html("showmsg=1");
+		delim = "&amp;";
+	}
+	if (follow) {
+		html(delim);
+		html("follow=1");
 	}
 	html("'>");
 	html_txt(name);
@@ -373,6 +379,10 @@ void cgit_commit_link(char *name, const char *title, const char *class,
 		html("ignorews=1");
 		delim = "&amp;";
 	}
+	if (ctx.qry.follow) {
+		html(delim);
+		html("follow=1");
+	}
 	html("'>");
 	if (name[0] != '\0')
 		html_txt(name);
@@ -429,6 +439,10 @@ void cgit_diff_link(const char *name, const char *title, const char *class,
 		html("ignorews=1");
 		delim = "&amp;";
 	}
+	if (ctx.qry.follow) {
+		html(delim);
+		html("follow=1");
+	}
 	html("'>");
 	html_txt(name);
 	html("</a>");
@@ -469,7 +483,7 @@ static void cgit_self_link(char *name, const char *title, const char *class)
 			      ctx.qry.has_sha1 ? ctx.qry.sha1 : NULL,
 			      ctx.qry.path, ctx.qry.ofs,
 			      ctx.qry.grep, ctx.qry.search,
-			      ctx.qry.showmsg);
+			      ctx.qry.showmsg, ctx.qry.follow);
 	else if (!strcmp(ctx.qry.page, "commit"))
 		cgit_commit_link(name, title, class, ctx.qry.head,
 				 ctx.qry.has_sha1 ? ctx.qry.sha1 : NULL,
@@ -945,7 +959,7 @@ void cgit_print_pageheader(void)
 			       ctx.qry.sha1, NULL);
 		cgit_log_link("log", NULL, hc("log"), ctx.qry.head,
 			      NULL, ctx.qry.vpath, 0, NULL, NULL,
-			      ctx.qry.showmsg);
+			      ctx.qry.showmsg, ctx.qry.follow);
 		cgit_tree_link("tree", NULL, hc("tree"), ctx.qry.head,
 			       ctx.qry.sha1, ctx.qry.vpath);
 		cgit_commit_link("commit", NULL, hc("commit"),
@@ -993,6 +1007,14 @@ void cgit_print_pageheader(void)
 		html("<div class='path'>");
 		html("path: ");
 		cgit_print_path_crumbs(ctx.qry.vpath);
+		if (ctx.cfg.enable_follow_links && !strcmp(ctx.qry.page, "log")) {
+			html(" (");
+			ctx.qry.follow = !ctx.qry.follow;
+			cgit_self_link(ctx.qry.follow ? "follow" : "unfollow",
+					NULL, NULL);
+			ctx.qry.follow = !ctx.qry.follow;
+			html(")");
+		}
 		html("</div>");
 	}
 	html("<div class='content'>");
diff --git a/ui-shared.h b/ui-shared.h
index 43d0fa6..788b1bc 100644
--- a/ui-shared.h
+++ b/ui-shared.h
@@ -31,7 +31,7 @@ extern void cgit_plain_link(const char *name, const char *title,
 extern void cgit_log_link(const char *name, const char *title,
 			  const char *class, const char *head, const char *rev,
 			  const char *path, int ofs, const char *grep,
-			  const char *pattern, int showmsg);
+			  const char *pattern, int showmsg, int follow);
 extern void cgit_commit_link(char *name, const char *title,
 			     const char *class, const char *head,
 			     const char *rev, const char *path);
diff --git a/ui-tree.c b/ui-tree.c
index bbc468e..c8d24f6 100644
--- a/ui-tree.c
+++ b/ui-tree.c
@@ -166,7 +166,7 @@ static int ls_item(const unsigned char *sha1, struct strbuf *base,
 	html("<td>");
 	cgit_log_link("log", NULL, "button", ctx.qry.head,
 		      walk_tree_ctx->curr_rev, fullpath.buf, 0, NULL, NULL,
-		      ctx.qry.showmsg);
+		      ctx.qry.showmsg, 0);
 	if (ctx.repo->max_stats)
 		cgit_stats_link("stats", NULL, "button", ctx.qry.head,
 				fullpath.buf);