[PATCH 0/5] xdiff: replace emit_func() with type-safe

[PATCH 0/5] xdiff: replace emit_func() with type-safe

[René Scharfe]

Commit ef2e62fe added the ability to bypass the creation of textual diffs
and to access diff data directly, through the added emit_func member of
xdemitconf_t.  This function pointer has a type of "void (*)()", because
"int (*)(xdfenv_t *, xdchange_t *, xdemitcb_t *, xdemitconf_t const *)"
(its real type) would be difficult to export due to the many internal
types in that signature.

This feature is currently used twice in git blame, through the function
xdi_diff_hunks().  Both cases only need to know which lines are affected
by the different hunks of a diff, i.e. the numbers shown in hunk headers
of unified diffs.

This patch series adds a simpler mechanism to accommodate these two
callers, in a type-safe way, without exporting too many internal types
of libxdiff.  The last patch removes the old functions.

  xdiff: add hunk_func()
  blame: use hunk_func(), part 1
  blame: use hunk_func(), part 2
  blame: factor out helper for calling xdi_diff()
  xdiff: remove emit_func() and xdi_diff_hunks()

 builtin/blame.c   |   53 +++++++++++++++++++++++++++++++----------------------
 xdiff-interface.c |   44 --------------------------------------------
 xdiff-interface.h |    4 ----
 xdiff/xdiff.h     |    6 +++++-
 xdiff/xdiffi.c    |   17 +++++++++++++++--
 5 files changed, 51 insertions(+), 73 deletions(-)

-- 
1.7.10.1

[PATCH 1/5] xdiff: add hunk_func()

[René Scharfe]

Add a way to register a callback function that is gets passed the
start line and line count of each hunk of a diff.  Only standard
types are used.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
---
 xdiff/xdiff.h  |    5 +++++
 xdiff/xdiffi.c |   17 +++++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/xdiff/xdiff.h b/xdiff/xdiff.h
index 09215af..33c010b 100644
--- a/xdiff/xdiff.h
+++ b/xdiff/xdiff.h
@@ -86,6 +86,10 @@ typedef struct s_xdemitcb {
 
 typedef long (*find_func_t)(const char *line, long line_len, char *buffer, long buffer_size, void *priv);
 
+typedef int (*xdl_emit_hunk_consume_func_t)(long start_a, long count_a,
+					    long start_b, long count_b,
+					    void *cb_data);
+
 typedef struct s_xdemitconf {
 	long ctxlen;
 	long interhunkctxlen;
@@ -93,6 +97,7 @@ typedef struct s_xdemitconf {
 	find_func_t find_func;
 	void *find_func_priv;
 	void (*emit_func)();
+	xdl_emit_hunk_consume_func_t hunk_func;
 } xdemitconf_t;
 
 typedef struct s_bdiffparam {
diff --git a/xdiff/xdiffi.c b/xdiff/xdiffi.c
index bc889e8..4d671f4 100644
--- a/xdiff/xdiffi.c
+++ b/xdiff/xdiffi.c
@@ -538,6 +538,20 @@ void xdl_free_script(xdchange_t *xscr) {
 	}
 }
 
+static int xdl_call_hunk_func(xdfenv_t *xe, xdchange_t *xscr, xdemitcb_t *ecb,
+			      xdemitconf_t const *xecfg)
+{
+	xdchange_t *xch, *xche;
+
+	for (xch = xscr; xch; xch = xche->next) {
+		xche = xdl_get_hunk(xch, xecfg);
+		if (xecfg->hunk_func(xch->i1, xche->i1 + xche->chg1 - xch->i1,
+				     xch->i2, xche->i2 + xche->chg2 - xch->i2,
+				     ecb->priv) < 0)
+			return -1;
+	}
+	return 0;
+}
 
 int xdl_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
 	     xdemitconf_t const *xecfg, xdemitcb_t *ecb) {
@@ -546,6 +560,9 @@ int xdl_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
 	emit_func_t ef = xecfg->emit_func ?
 		(emit_func_t)xecfg->emit_func : xdl_emit_diff;
 
+	if (xecfg->hunk_func)
+		ef = xdl_call_hunk_func;
+
 	if (xdl_do_diff(mf1, mf2, xpp, &xe) < 0) {
 
 		return -1;
-- 
1.7.10.1

[PATCH 2/5] blame: use hunk_func(), part 1

[René Scharfe]

Use blame_chunk_cb() directly as hunk_func() callback, without detour
through xdi_diff_hunks().

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
---
 builtin/blame.c |   17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/builtin/blame.c b/builtin/blame.c
index 324d476..9ae5cf3 100644
--- a/builtin/blame.c
+++ b/builtin/blame.c
@@ -759,12 +759,14 @@ struct blame_chunk_cb_data {
 	long tlno;
 };
 
-static void blame_chunk_cb(void *data, long same, long p_next, long t_next)
+static int blame_chunk_cb(long start_a, long count_a,
+			  long start_b, long count_b, void *data)
 {
 	struct blame_chunk_cb_data *d = data;
-	blame_chunk(d->sb, d->tlno, d->plno, same, d->target, d->parent);
-	d->plno = p_next;
-	d->tlno = t_next;
+	blame_chunk(d->sb, d->tlno, d->plno, start_b, d->target, d->parent);
+	d->plno = start_a + count_a;
+	d->tlno = start_b + count_b;
+	return 0;
 }
 
 /*
@@ -781,6 +783,8 @@ static int pass_blame_to_parent(struct scoreboard *sb,
 	struct blame_chunk_cb_data d;
 	xpparam_t xpp;
 	xdemitconf_t xecfg;
+	xdemitcb_t ecb;
+
 	memset(&d, 0, sizeof(d));
 	d.sb = sb; d.target = target; d.parent = parent;
 	last_in_target = find_last_in_target(sb, target);
@@ -795,7 +799,10 @@ static int pass_blame_to_parent(struct scoreboard *sb,
 	xpp.flags = xdl_opts;
 	memset(&xecfg, 0, sizeof(xecfg));
 	xecfg.ctxlen = 0;
-	xdi_diff_hunks(&file_p, &file_o, blame_chunk_cb, &d, &xpp, &xecfg);
+	xecfg.hunk_func = blame_chunk_cb;
+	memset(&ecb, 0, sizeof(ecb));
+	ecb.priv = &d;
+	xdi_diff(&file_p, &file_o, &xpp, &xecfg, &ecb);
 	/* The rest (i.e. anything after tlno) are the same as the parent */
 	blame_chunk(sb, d.tlno, d.plno, last_in_target, target, parent);
 
-- 
1.7.10.1

[PATCH 3/5] blame: use hunk_func(), part 2

[René Scharfe]

Use handle_split_cb() directly as hunk_func() callback, without going
through xdi_diff_hunks().

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
---
 builtin/blame.c |   18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/builtin/blame.c b/builtin/blame.c
index 9ae5cf3..c83fc7c 100644
--- a/builtin/blame.c
+++ b/builtin/blame.c
@@ -906,12 +906,15 @@ struct handle_split_cb_data {
 	long tlno;
 };
 
-static void handle_split_cb(void *data, long same, long p_next, long t_next)
+static int handle_split_cb(long start_a, long count_a,
+			   long start_b, long count_b, void *data)
 {
 	struct handle_split_cb_data *d = data;
-	handle_split(d->sb, d->ent, d->tlno, d->plno, same, d->parent, d->split);
-	d->plno = p_next;
-	d->tlno = t_next;
+	handle_split(d->sb, d->ent, d->tlno, d->plno, start_b, d->parent,
+		     d->split);
+	d->plno = start_a + count_a;
+	d->tlno = start_b + count_b;
+	return 0;
 }
 
 /*
@@ -931,6 +934,8 @@ static void find_copy_in_blob(struct scoreboard *sb,
 	struct handle_split_cb_data d;
 	xpparam_t xpp;
 	xdemitconf_t xecfg;
+	xdemitcb_t ecb;
+
 	memset(&d, 0, sizeof(d));
 	d.sb = sb; d.ent = ent; d.parent = parent; d.split = split;
 	/*
@@ -954,8 +959,11 @@ static void find_copy_in_blob(struct scoreboard *sb,
 	xpp.flags = xdl_opts;
 	memset(&xecfg, 0, sizeof(xecfg));
 	xecfg.ctxlen = 1;
+	xecfg.hunk_func = handle_split_cb;
+	memset(&ecb, 0, sizeof(ecb));
+	ecb.priv = &d;
 	memset(split, 0, sizeof(struct blame_entry [3]));
-	xdi_diff_hunks(file_p, &file_o, handle_split_cb, &d, &xpp, &xecfg);
+	xdi_diff(file_p, &file_o, &xpp, &xecfg, &ecb);
 	/* remainder, if any, all match the preimage */
 	handle_split(sb, ent, d.tlno, d.plno, ent->num_lines, parent, split);
 }
-- 
1.7.10.1

[PATCH 4/5] blame: factor out helper for calling xdi_diff()

[René Scharfe]

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
---
 builtin/blame.c |   38 ++++++++++++++++----------------------
 1 file changed, 16 insertions(+), 22 deletions(-)

diff --git a/builtin/blame.c b/builtin/blame.c
index c83fc7c..778d661 100644
--- a/builtin/blame.c
+++ b/builtin/blame.c
@@ -88,6 +88,20 @@ struct origin {
 	char path[FLEX_ARRAY];
 };
 
+static int diff_hunks(mmfile_t *file_a, mmfile_t *file_b, long ctxlen,
+		      xdl_emit_hunk_consume_func_t hunk_func, void *cb_data)
+{
+	xpparam_t xpp = {0};
+	xdemitconf_t xecfg = {0};
+	xdemitcb_t ecb = {0};
+
+	xpp.flags = xdl_opts;
+	xecfg.ctxlen = ctxlen;
+	xecfg.hunk_func = hunk_func;
+	ecb.priv = cb_data;
+	return xdi_diff(file_a, file_b, &xpp, &xecfg, &ecb);
+}
+
 /*
  * Prepare diff_filespec and convert it using diff textconv API
  * if the textconv driver exists.
@@ -781,9 +795,6 @@ static int pass_blame_to_parent(struct scoreboard *sb,
 	int last_in_target;
 	mmfile_t file_p, file_o;
 	struct blame_chunk_cb_data d;
-	xpparam_t xpp;
-	xdemitconf_t xecfg;
-	xdemitcb_t ecb;
 
 	memset(&d, 0, sizeof(d));
 	d.sb = sb; d.target = target; d.parent = parent;
@@ -795,14 +806,7 @@ static int pass_blame_to_parent(struct scoreboard *sb,
 	fill_origin_blob(&sb->revs->diffopt, target, &file_o);
 	num_get_patch++;
 
-	memset(&xpp, 0, sizeof(xpp));
-	xpp.flags = xdl_opts;
-	memset(&xecfg, 0, sizeof(xecfg));
-	xecfg.ctxlen = 0;
-	xecfg.hunk_func = blame_chunk_cb;
-	memset(&ecb, 0, sizeof(ecb));
-	ecb.priv = &d;
-	xdi_diff(&file_p, &file_o, &xpp, &xecfg, &ecb);
+	diff_hunks(&file_p, &file_o, 0, blame_chunk_cb, &d);
 	/* The rest (i.e. anything after tlno) are the same as the parent */
 	blame_chunk(sb, d.tlno, d.plno, last_in_target, target, parent);
 
@@ -932,9 +936,6 @@ static void find_copy_in_blob(struct scoreboard *sb,
 	int cnt;
 	mmfile_t file_o;
 	struct handle_split_cb_data d;
-	xpparam_t xpp;
-	xdemitconf_t xecfg;
-	xdemitcb_t ecb;
 
 	memset(&d, 0, sizeof(d));
 	d.sb = sb; d.ent = ent; d.parent = parent; d.split = split;
@@ -955,15 +956,8 @@ static void find_copy_in_blob(struct scoreboard *sb,
 	 * file_o is a part of final image we are annotating.
 	 * file_p partially may match that image.
 	 */
-	memset(&xpp, 0, sizeof(xpp));
-	xpp.flags = xdl_opts;
-	memset(&xecfg, 0, sizeof(xecfg));
-	xecfg.ctxlen = 1;
-	xecfg.hunk_func = handle_split_cb;
-	memset(&ecb, 0, sizeof(ecb));
-	ecb.priv = &d;
 	memset(split, 0, sizeof(struct blame_entry [3]));
-	xdi_diff(file_p, &file_o, &xpp, &xecfg, &ecb);
+	diff_hunks(file_p, &file_o, 1, handle_split_cb, &d);
 	/* remainder, if any, all match the preimage */
 	handle_split(sb, ent, d.tlno, d.plno, ent->num_lines, parent, split);
 }
-- 
1.7.10.1

[PATCH 5/5] xdiff: remove emit_func() and xdi_diff_hunks()

[René Scharfe]

The functions are unused now, remove them.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
---
 xdiff-interface.c |   44 --------------------------------------------
 xdiff-interface.h |    4 ----
 xdiff/xdiff.h     |    1 -
 xdiff/xdiffi.c    |    6 +-----
 4 files changed, 1 insertion(+), 54 deletions(-)

diff --git a/xdiff-interface.c b/xdiff-interface.c
index 0e2c169..ecfa05f 100644
--- a/xdiff-interface.c
+++ b/xdiff-interface.c
@@ -156,50 +156,6 @@ int xdi_diff_outf(mmfile_t *mf1, mmfile_t *mf2,
 	return ret;
 }
 
-struct xdiff_emit_hunk_state {
-	xdiff_emit_hunk_consume_fn consume;
-	void *consume_callback_data;
-};
-
-static int process_diff(xdfenv_t *xe, xdchange_t *xscr, xdemitcb_t *ecb,
-			xdemitconf_t const *xecfg)
-{
-	long s1, s2, same, p_next, t_next;
-	xdchange_t *xch, *xche;
-	struct xdiff_emit_hunk_state *state = ecb->priv;
-	xdiff_emit_hunk_consume_fn fn = state->consume;
-	void *consume_callback_data = state->consume_callback_data;
-
-	for (xch = xscr; xch; xch = xche->next) {
-		xche = xdl_get_hunk(xch, xecfg);
-
-		s1 = XDL_MAX(xch->i1 - xecfg->ctxlen, 0);
-		s2 = XDL_MAX(xch->i2 - xecfg->ctxlen, 0);
-		same = s2 + XDL_MAX(xch->i1 - s1, 0);
-		p_next = xche->i1 + xche->chg1;
-		t_next = xche->i2 + xche->chg2;
-
-		fn(consume_callback_data, same, p_next, t_next);
-	}
-	return 0;
-}
-
-int xdi_diff_hunks(mmfile_t *mf1, mmfile_t *mf2,
-		   xdiff_emit_hunk_consume_fn fn, void *consume_callback_data,
-		   xpparam_t const *xpp, xdemitconf_t *xecfg)
-{
-	struct xdiff_emit_hunk_state state;
-	xdemitcb_t ecb;
-
-	memset(&state, 0, sizeof(state));
-	memset(&ecb, 0, sizeof(ecb));
-	state.consume = fn;
-	state.consume_callback_data = consume_callback_data;
-	xecfg->emit_func = (void (*)())process_diff;
-	ecb.priv = &state;
-	return xdi_diff(mf1, mf2, xpp, xecfg, &ecb);
-}
-
 int read_mmfile(mmfile_t *ptr, const char *filename)
 {
 	struct stat st;
diff --git a/xdiff-interface.h b/xdiff-interface.h
index 49d1116..eff7762 100644
--- a/xdiff-interface.h
+++ b/xdiff-interface.h
@@ -4,15 +4,11 @@
 #include "xdiff/xdiff.h"
 
 typedef void (*xdiff_emit_consume_fn)(void *, char *, unsigned long);
-typedef void (*xdiff_emit_hunk_consume_fn)(void *, long, long, long);
 
 int xdi_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp, xdemitconf_t const *xecfg, xdemitcb_t *ecb);
 int xdi_diff_outf(mmfile_t *mf1, mmfile_t *mf2,
 		  xdiff_emit_consume_fn fn, void *consume_callback_data,
 		  xpparam_t const *xpp, xdemitconf_t const *xecfg);
-int xdi_diff_hunks(mmfile_t *mf1, mmfile_t *mf2,
-		   xdiff_emit_hunk_consume_fn fn, void *consume_callback_data,
-		   xpparam_t const *xpp, xdemitconf_t *xecfg);
 int parse_hunk_header(char *line, int len,
 		      int *ob, int *on,
 		      int *nb, int *nn);
diff --git a/xdiff/xdiff.h b/xdiff/xdiff.h
index 33c010b..219a3bb 100644
--- a/xdiff/xdiff.h
+++ b/xdiff/xdiff.h
@@ -96,7 +96,6 @@ typedef struct s_xdemitconf {
 	unsigned long flags;
 	find_func_t find_func;
 	void *find_func_priv;
-	void (*emit_func)();
 	xdl_emit_hunk_consume_func_t hunk_func;
 } xdemitconf_t;
 
diff --git a/xdiff/xdiffi.c b/xdiff/xdiffi.c
index 4d671f4..1b7012a 100644
--- a/xdiff/xdiffi.c
+++ b/xdiff/xdiffi.c
@@ -557,11 +557,7 @@ int xdl_diff(mmfile_t *mf1, mmfile_t *mf2, xpparam_t const *xpp,
 	     xdemitconf_t const *xecfg, xdemitcb_t *ecb) {
 	xdchange_t *xscr;
 	xdfenv_t xe;
-	emit_func_t ef = xecfg->emit_func ?
-		(emit_func_t)xecfg->emit_func : xdl_emit_diff;
-
-	if (xecfg->hunk_func)
-		ef = xdl_call_hunk_func;
+	emit_func_t ef = xecfg->hunk_func ? xdl_call_hunk_func : xdl_emit_diff;
 
 	if (xdl_do_diff(mf1, mf2, xpp, &xe) < 0) {
 
-- 
1.7.10.1

Re: [PATCH 0/5] xdiff: replace emit_func() with type-safe

[Thomas Rast]

René Scharfe <rene.scharfe@lsrfire.ath.cx> writes:

> Commit ef2e62fe added the ability to bypass the creation of textual diffs
> and to access diff data directly, through the added emit_func member of
> xdemitconf_t.  This function pointer has a type of "void (*)()", because
> "int (*)(xdfenv_t *, xdchange_t *, xdemitcb_t *, xdemitconf_t const *)"
> (its real type) would be difficult to export due to the many internal
> types in that signature.
>
> This feature is currently used twice in git blame, through the function
> xdi_diff_hunks().  Both cases only need to know which lines are affected
> by the different hunks of a diff, i.e. the numbers shown in hunk headers
> of unified diffs.
>
> This patch series adds a simpler mechanism to accommodate these two
> callers, in a type-safe way, without exporting too many internal types
> of libxdiff.  The last patch removes the old functions.

Cool, thanks a lot, this will come in handy when I finally get around to
finishing log -L.

-- 
Thomas Rast
trast@{inf,student}.ethz.ch