[PATCH v2 0/3] fast-export: speed improvements

[PATCH v2 0/3] fast-export: speed improvements

[Felipe Contreras]

Hi,

Second try, dropped one fix, and address a few comments.

Parsing the marks of an import of the emacs repository moves fast-export to a
crawl. It takes 14 minutes in my setup, after these patches, it takes 1 second.

The important patches are #2 and #3, the rest are niceities.

Felipe Contreras (3):
  fast-{import,export}: use get_sha1_hex() directly
  fast-export: improve speed by skipping blobs
  fast-export: don't parse all the commits

 builtin/fast-export.c | 22 +++++++++++++++-------
 fast-import.c         | 10 +++++-----
 2 files changed, 20 insertions(+), 12 deletions(-)

-- 
1.8.3.rc0.401.g45bba44

[PATCH v2 1/3] fast-{import,export}: use get_sha1_hex() directly

[Felipe Contreras]

It's wrong to call get_sha1() if they should be SHA-1s, plus
inefficient.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
---
 builtin/fast-export.c |  2 +-
 fast-import.c         | 10 +++++-----
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/builtin/fast-export.c b/builtin/fast-export.c
index d60d675..a4dee14 100644
--- a/builtin/fast-export.c
+++ b/builtin/fast-export.c
@@ -621,7 +621,7 @@ static void import_marks(char *input_file)
 
 		mark = strtoumax(line + 1, &mark_end, 10);
 		if (!mark || mark_end == line + 1
-			|| *mark_end != ' ' || get_sha1(mark_end + 1, sha1))
+			|| *mark_end != ' ' || get_sha1_hex(mark_end + 1, sha1))
 			die("corrupt mark line: %s", line);
 
 		if (last_idnum < mark)
diff --git a/fast-import.c b/fast-import.c
index 5f539d7..e02f212 100644
--- a/fast-import.c
+++ b/fast-import.c
@@ -1822,7 +1822,7 @@ static void read_marks(void)
 		*end = 0;
 		mark = strtoumax(line + 1, &end, 10);
 		if (!mark || end == line + 1
-			|| *end != ' ' || get_sha1(end + 1, sha1))
+			|| *end != ' ' || get_sha1_hex(end + 1, sha1))
 			die("corrupt mark line: %s", line);
 		e = find_object(sha1);
 		if (!e) {
@@ -2490,7 +2490,7 @@ static void note_change_n(struct branch *b, unsigned char *old_fanout)
 		if (commit_oe->type != OBJ_COMMIT)
 			die("Mark :%" PRIuMAX " not a commit", commit_mark);
 		hashcpy(commit_sha1, commit_oe->idx.sha1);
-	} else if (!get_sha1(p, commit_sha1)) {
+	} else if (!get_sha1_hex(p, commit_sha1)) {
 		unsigned long size;
 		char *buf = read_object_with_reference(commit_sha1,
 			commit_type, &size, commit_sha1);
@@ -2604,7 +2604,7 @@ static int parse_from(struct branch *b)
 			free(buf);
 		} else
 			parse_from_existing(b);
-	} else if (!get_sha1(from, b->sha1))
+	} else if (!get_sha1_hex(from, b->sha1))
 		parse_from_existing(b);
 	else
 		die("Invalid ref name or SHA1 expression: %s", from);
@@ -2632,7 +2632,7 @@ static struct hash_list *parse_merge(unsigned int *count)
 			if (oe->type != OBJ_COMMIT)
 				die("Mark :%" PRIuMAX " not a commit", idnum);
 			hashcpy(n->sha1, oe->idx.sha1);
-		} else if (!get_sha1(from, n->sha1)) {
+		} else if (!get_sha1_hex(from, n->sha1)) {
 			unsigned long size;
 			char *buf = read_object_with_reference(n->sha1,
 				commit_type, &size, n->sha1);
@@ -2792,7 +2792,7 @@ static void parse_new_tag(void)
 		oe = find_mark(from_mark);
 		type = oe->type;
 		hashcpy(sha1, oe->idx.sha1);
-	} else if (!get_sha1(from, sha1)) {
+	} else if (!get_sha1_hex(from, sha1)) {
 		struct object_entry *oe = find_object(sha1);
 		if (!oe) {
 			type = sha1_object_info(sha1, NULL);
-- 
1.8.3.rc0.401.g45bba44

Re: [PATCH v2 1/3] fast-{import,export}: use get_sha1_hex() directly

[Junio C Hamano]

Felipe Contreras <felipe.contreras@gmail.com> writes:

> It's wrong to call get_sha1() if they should be SHA-1s, plus
> inefficient.
>
> Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
> ---

It appears that "they should be SHA-1s" assumption does not hold;
this patch breaks at least 3303, 9020, and 9300.

Also assuming these are always 40-hex goes directly against what is
documented in Documentation/git-fast-import.txt (look for "Here
committish is any of the following").  My bad while reviewing the
earlier round.

I've redone 'pu' (which was failing the test last night) after
dropping this and keeping only patches 2 and 3 from the series.

Re: [PATCH v2 1/3] fast-{import,export}: use get_sha1_hex() directly

[Felipe Contreras]

On Tue, May 7, 2013 at 9:38 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Felipe Contreras <felipe.contreras@gmail.com> writes:
>
>> It's wrong to call get_sha1() if they should be SHA-1s, plus
>> inefficient.
>>
>> Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
>> ---
>
> It appears that "they should be SHA-1s" assumption does not hold;
> this patch breaks at least 3303, 9020, and 9300.
>
> Also assuming these are always 40-hex goes directly against what is
> documented in Documentation/git-fast-import.txt (look for "Here
> committish is any of the following").  My bad while reviewing the
> earlier round.
>
> I've redone 'pu' (which was failing the test last night) after
> dropping this and keeping only patches 2 and 3 from the series.

Turns out most of the get_sha1() calls were correct; this does the trick:

diff --git a/builtin/fast-export.c b/builtin/fast-export.c
index 18fdfb3..d1d68e9 100644
--- a/builtin/fast-export.c
+++ b/builtin/fast-export.c
@@ -623,7 +623,7 @@ static void import_marks(char *input_file)

                mark = strtoumax(line + 1, &mark_end, 10);
                if (!mark || mark_end == line + 1
-                       || *mark_end != ' ' || get_sha1(mark_end + 1, sha1))
+                       || *mark_end != ' ' || get_sha1_hex(mark_end + 1, sha1))
                        die("corrupt mark line: %s", line);

                if (last_idnum < mark)
diff --git a/fast-import.c b/fast-import.c
index 5f539d7..3f32149 100644
--- a/fast-import.c
+++ b/fast-import.c
@@ -1822,7 +1822,7 @@ static void read_marks(void)
                *end = 0;
                mark = strtoumax(line + 1, &end, 10);
                if (!mark || end == line + 1
-                       || *end != ' ' || get_sha1(end + 1, sha1))
+                       || *end != ' ' || get_sha1_hex(end + 1, sha1))
                        die("corrupt mark line: %s", line);
                e = find_object(sha1);
                if (!e) {

-- 
Felipe Contreras

Re: [PATCH v2 1/3] fast-{import,export}: use get_sha1_hex() directly

[Junio C Hamano]

Felipe Contreras <felipe.contreras@gmail.com> writes:

> Turns out most of the get_sha1() calls were correct; this does the trick:
>
> diff --git a/builtin/fast-export.c b/builtin/fast-export.c
> index 18fdfb3..d1d68e9 100644
> --- a/builtin/fast-export.c
> +++ b/builtin/fast-export.c
> @@ -623,7 +623,7 @@ static void import_marks(char *input_file)
>
>                 mark = strtoumax(line + 1, &mark_end, 10);
>                 if (!mark || mark_end == line + 1
> -                       || *mark_end != ' ' || get_sha1(mark_end + 1, sha1))
> +                       || *mark_end != ' ' || get_sha1_hex(mark_end + 1, sha1))
>                         die("corrupt mark line: %s", line);
>
>                 if (last_idnum < mark)
> diff --git a/fast-import.c b/fast-import.c
> index 5f539d7..3f32149 100644
> --- a/fast-import.c
> +++ b/fast-import.c
> @@ -1822,7 +1822,7 @@ static void read_marks(void)
>                 *end = 0;
>                 mark = strtoumax(line + 1, &end, 10);
>                 if (!mark || end == line + 1
> -                       || *end != ' ' || get_sha1(end + 1, sha1))
> +                       || *end != ' ' || get_sha1_hex(end + 1, sha1))

This is where --import-marks is handled, and we should be seeing

	:markid SHA-1

one per each line (according to Documentation/git-fast-import.txt).
So this one should be get_sha1_hex().

The other one in fast-export.c would be the same.

The other ones in the original patch were reading from the
fast-import stream and shouldn't have insisted on 40-hex.

Will replace the body of the change with only these two hunks and
requeue.  Thanks.

[PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

We don't care about blobs, or any object other than commits, but in
order to find the type of object, we are parsing the whole thing, which
is slow, specially in big repositories with lots of big files.

There's no need for that, we can query the object information with
sha1_object_info();

Before this, loading the objects of a fresh emacs import, with 260598
blobs took 14 minutes, after this patch, it takes 3 seconds.

This is the way fast-import does it. Also die if the object is not
found (like fast-import).

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
---
 builtin/fast-export.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/builtin/fast-export.c b/builtin/fast-export.c
index a4dee14..a5b8da8 100644
--- a/builtin/fast-export.c
+++ b/builtin/fast-export.c
@@ -613,6 +613,7 @@ static void import_marks(char *input_file)
 		char *line_end, *mark_end;
 		unsigned char sha1[20];
 		struct object *object;
+		enum object_type type;
 
 		line_end = strchr(line, '\n');
 		if (line[0] != ':' || !line_end)
@@ -627,17 +628,19 @@ static void import_marks(char *input_file)
 		if (last_idnum < mark)
 			last_idnum = mark;
 
-		object = parse_object(sha1);
-		if (!object)
+		type = sha1_object_info(sha1, NULL);
+		if (type < 0)
+			die("object not found: %s", sha1_to_hex(sha1));
+
+		if (type != OBJ_COMMIT)
+			/* only commits */
 			continue;
 
+		object = parse_object(sha1);
+
 		if (object->flags & SHOWN)
 			error("Object %s already has a mark", sha1_to_hex(sha1));
 
-		if (object->type != OBJ_COMMIT)
-			/* only commits */
-			continue;
-
 		mark_object(object, mark);
 
 		object->flags |= SHOWN;
-- 
1.8.3.rc0.401.g45bba44

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Jeff King]

On Sun, May 05, 2013 at 05:38:53PM -0500, Felipe Contreras wrote:

> We don't care about blobs, or any object other than commits, but in
> order to find the type of object, we are parsing the whole thing, which
> is slow, specially in big repositories with lots of big files.

I did a double-take on reading this subject line and first paragraph,
thinking "surely fast-export needs to actually output blobs?".

Reading the patch, I see that this is only about not bothering to load
blob marks from --import-marks. It might be nice to mention that in the
commit message, which is otherwise quite confusing.

I'm also not sure why your claim "we don't care about blobs" is true,
because naively we would want future runs of fast-export to avoid having
to write out the whole blob content when mentioning the blob again. I
think one argument could be "if we write a mark for blob X, we will also
have written a mark for commit Y which contains it; on subsequent runs,
we will just show the mark for Y in the first place, and not even care
about showing X (as a part of Y) either way. We would only refer to the
mark for X if it appears as part of a different commit, but that is a
rare case not worth worrying about."

Does that match your reasoning?

> Before this, loading the objects of a fresh emacs import, with 260598
> blobs took 14 minutes, after this patch, it takes 3 seconds.

Presumably most of that speed improvement comes from not parsing the
blob objects. I wonder if you could get similar speedups by applying the
"do not bother parsing" rule from your patch 3. You would still incur
some cost to create a "struct blob", but it may or may not be
measurable.  That would mean we get the "case not worth worrying about"
from above for free. I doubt it would make that big a difference,
though, given the rarity of it. So I am OK with it either way.

-Peff

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Junio C Hamano]

Jeff King <peff@peff.net> writes:

> On Sun, May 05, 2013 at 05:38:53PM -0500, Felipe Contreras wrote:
>
>> We don't care about blobs, or any object other than commits, but in
>> order to find the type of object, we are parsing the whole thing, which
>> is slow, specially in big repositories with lots of big files.
>
> I did a double-take on reading this subject line and first paragraph,
> thinking "surely fast-export needs to actually output blobs?".
>
> Reading the patch, I see that this is only about not bothering to load
> blob marks from --import-marks. It might be nice to mention that in the
> commit message, which is otherwise quite confusing.

I had the same reaction first, but not writing the blob _objects_
out to the output stream would not make any sense, so it was fairly
easy to guess what the author wanted to say ;-).

> I'm also not sure why your claim "we don't care about blobs" is true,
> because naively we would want future runs of fast-export to avoid having
> to write out the whole blob content when mentioning the blob again.

The existing documentation is fairly clear that marks for objects
other than commits are not exported, and the import-marks codepath
discards anything but commits, so there is no mechanism for the
existing fast-export users to leave blob marks in the marks file for
later runs of fast-export to take advantage of.  The second
invocation cannot refer to such a blob in the first place.

The story is different on the fast-import side, where we do say we
dump the full table and a later run can depend on these marks.

By discarding marks on blobs, we may be robbing some optimization
possibilities, and by discarding marks on tags, we may be robbing
some features, from users of fast-export; we might want to add an
option "--use-object-marks={blob,commit,tag}" or something to both
fast-export and fast-import, so that the former can optionally write
marks for non-commits out, and the latter can omit non commit marks
if the user do not need them. But that is a separate issue.

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Junio C Hamano]

Junio C Hamano <gitster@pobox.com> writes:

> By discarding marks on blobs, we may be robbing some optimization
> possibilities, and by discarding marks on tags, we may be robbing
> some features, from users of fast-export; we might want to add an
> option "--use-object-marks={blob,commit,tag}" or something to both
> fast-export and fast-import, so that the former can optionally write
> marks for non-commits out, and the latter can omit non commit marks
> if the user do not need them. But that is a separate issue.

s/--use-object-marks/--persistent-object-marks/; I think that would
express the issue better.

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Jeff King]

On Mon, May 06, 2013 at 08:08:45AM -0700, Junio C Hamano wrote:

> > I'm also not sure why your claim "we don't care about blobs" is true,
> > because naively we would want future runs of fast-export to avoid having
> > to write out the whole blob content when mentioning the blob again.
> 
> The existing documentation is fairly clear that marks for objects
> other than commits are not exported, and the import-marks codepath
> discards anything but commits, so there is no mechanism for the
> existing fast-export users to leave blob marks in the marks file for
> later runs of fast-export to take advantage of.  The second
> invocation cannot refer to such a blob in the first place.

OK. If the argument is "we do not write them, so do not bother reading
them back in", I think that is reasonable. It could hurt anybody trying
to run "fast-export" against a marks file created by somebody else, but
that is also the same case that is being helped here (since otherwise,
we would not be seeing blob entries at all).

I do not offhand know enough about the internals of import/export-style
remote-helpers to say whether the "hurt" case even exists, let alone how
common it is.

> By discarding marks on blobs, we may be robbing some optimization
> possibilities, and by discarding marks on tags, we may be robbing
> some features, from users of fast-export; we might want to add an
> option "--use-object-marks={blob,commit,tag}" or something to both
> fast-export and fast-import, so that the former can optionally write
> marks for non-commits out, and the latter can omit non commit marks
> if the user do not need them. But that is a separate issue.

Yeah, that would allow the old behavior (and more) if anybody is hurt by
this. It is nice if the order of implementation is "more features, then
flip the default" because it provides an immediate escape hatch for
anybody who is hurt by the change in default. But again, I do not know
enough to say whether such hurt cases even exist.

-Peff

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Junio C Hamano]

Jeff King <peff@peff.net> writes:

> On Mon, May 06, 2013 at 08:08:45AM -0700, Junio C Hamano wrote:
>
>> > I'm also not sure why your claim "we don't care about blobs" is true,
>> > because naively we would want future runs of fast-export to avoid having
>> > to write out the whole blob content when mentioning the blob again.
>> 
>> The existing documentation is fairly clear that marks for objects
>> other than commits are not exported, and the import-marks codepath
>> discards anything but commits, so there is no mechanism for the
>> existing fast-export users to leave blob marks in the marks file for
>> later runs of fast-export to take advantage of.  The second
>> invocation cannot refer to such a blob in the first place.
>
> OK. If the argument is "we do not write them, so do not bother reading
> them back in", I think that is reasonable.

The way I read builtin/fast-export.c::import_marks() is that it is
more like "we do not write them, and we do not read them back in
either IN THE CURRENT CODE".

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Jeff King]

On Mon, May 06, 2013 at 09:32:41AM -0700, Junio C Hamano wrote:

> > OK. If the argument is "we do not write them, so do not bother reading
> > them back in", I think that is reasonable.
> 
> The way I read builtin/fast-export.c::import_marks() is that it is
> more like "we do not write them, and we do not read them back in
> either IN THE CURRENT CODE".

Ahh...I see now. It is not about skipping the blobs as a new behavior,
but rather about skipping them _earlier_, before we have loaded the
object contents from disk.

I took the "we don't care about" as "the general use of fast-export does
not care about", but it is "we will literally just drop them a few lines
later".

So yes, I think this is an obviously correct optimization. Thanks for
clarifying, and sorry to be so slow.

-Peff

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Junio C Hamano]

Jeff King <peff@peff.net> writes:

> So yes, I think this is an obviously correct optimization. Thanks for
> clarifying, and sorry to be so slow.

No need to be sorry.  It just shows that the log message could have
been more helpful.

Here is what I tentatively queued.

commit 83582e91d22c66413b291d4d6d45bbeafddc2af9
Author: Felipe Contreras <felipe.contreras@gmail.com>
Date:   Sun May 5 17:38:53 2013 -0500

    fast-export: do not parse non-commit objects while reading marks file
    
    We read from the marks file and keep only marked commits, but in
    order to find the type of object, we are parsing the whole thing,
    which is slow, specially in big repositories with lots of big files.
    
    There's no need for that, we can query the object information with
    sha1_object_info().
    
    Before this, loading the objects of a fresh emacs import, with 260598
    blobs took 14 minutes, after this patch, it takes 3 seconds.
    
    This is the way fast-import does it. Also die if the object is not
    found (like fast-import).
    
    Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
    Signed-off-by: Junio C Hamano <gitster@pobox.com>

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Jeff King]

On Mon, May 06, 2013 at 10:17:41AM -0700, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> > So yes, I think this is an obviously correct optimization. Thanks for
> > clarifying, and sorry to be so slow.
> 
> No need to be sorry.  It just shows that the log message could have
> been more helpful.
> 
> Here is what I tentatively queued.
> [...]

Yeah, that is much for to understand (to me, at least).

Thanks.

-Peff

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Jeff King]

On Mon, May 06, 2013 at 01:19:35PM -0400, Jeff King wrote:

> > Here is what I tentatively queued.
> > [...]
> 
> Yeah, that is much for to understand (to me, at least).

Ugh. That was supposed to be "much easier to understand". Perhaps I will
learn to type one day.

-Peff

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

On Mon, May 6, 2013 at 11:20 AM, Jeff King <peff@peff.net> wrote:
> On Mon, May 06, 2013 at 08:08:45AM -0700, Junio C Hamano wrote:
>
>> > I'm also not sure why your claim "we don't care about blobs" is true,
>> > because naively we would want future runs of fast-export to avoid having
>> > to write out the whole blob content when mentioning the blob again.
>>
>> The existing documentation is fairly clear that marks for objects
>> other than commits are not exported, and the import-marks codepath
>> discards anything but commits, so there is no mechanism for the
>> existing fast-export users to leave blob marks in the marks file for
>> later runs of fast-export to take advantage of.  The second
>> invocation cannot refer to such a blob in the first place.
>
> OK. If the argument is "we do not write them, so do not bother reading
> them back in", I think that is reasonable.

We already do that:

5d3698f fast-export: avoid importing blob marks

> It could hurt anybody trying
> to run "fast-export" against a marks file created by somebody else, but
> that is also the same case that is being helped here (since otherwise,
> we would not be seeing blob entries at all).
>
> I do not offhand know enough about the internals of import/export-style
> remote-helpers to say whether the "hurt" case even exists, let alone how
> common it is.
>
>> By discarding marks on blobs, we may be robbing some optimization
>> possibilities, and by discarding marks on tags, we may be robbing
>> some features, from users of fast-export; we might want to add an
>> option "--use-object-marks={blob,commit,tag}" or something to both
>> fast-export and fast-import, so that the former can optionally write
>> marks for non-commits out, and the latter can omit non commit marks
>> if the user do not need them. But that is a separate issue.
>
> Yeah, that would allow the old behavior (and more) if anybody is hurt by
> this.

There is no behavior change in this patch. We do *exactly* the same as before.

-- 
Felipe Contreras

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

On Mon, May 6, 2013 at 10:08 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Jeff King <peff@peff.net> writes:
>
>> On Sun, May 05, 2013 at 05:38:53PM -0500, Felipe Contreras wrote:
>>
>>> We don't care about blobs, or any object other than commits, but in
>>> order to find the type of object, we are parsing the whole thing, which
>>> is slow, specially in big repositories with lots of big files.
>>
>> I did a double-take on reading this subject line and first paragraph,
>> thinking "surely fast-export needs to actually output blobs?".
>>
>> Reading the patch, I see that this is only about not bothering to load
>> blob marks from --import-marks. It might be nice to mention that in the
>> commit message, which is otherwise quite confusing.
>
> I had the same reaction first, but not writing the blob _objects_
> out to the output stream would not make any sense, so it was fairly
> easy to guess what the author wanted to say ;-).

That's how fast-export has worked since --export-marks was introduced.

>> I'm also not sure why your claim "we don't care about blobs" is true,
>> because naively we would want future runs of fast-export to avoid having
>> to write out the whole blob content when mentioning the blob again.
>
> The existing documentation is fairly clear that marks for objects
> other than commits are not exported, and the import-marks codepath
> discards anything but commits, so there is no mechanism for the
> existing fast-export users to leave blob marks in the marks file for
> later runs of fast-export to take advantage of.  The second
> invocation cannot refer to such a blob in the first place.
>
> The story is different on the fast-import side, where we do say we
> dump the full table and a later run can depend on these marks.

Yes, and gaining nothing but increased disk-space.

> By discarding marks on blobs, we may be robbing some optimization
> possibilities, and by discarding marks on tags, we may be robbing
> some features, from users of fast-export; we might want to add an
> option "--use-object-marks={blob,commit,tag}" or something to both
> fast-export and fast-import, so that the former can optionally write
> marks for non-commits out, and the latter can omit non commit marks
> if the user do not need them. But that is a separate issue.

How? The only way we might rob optimizations is if there's an obscene
amount files, otherwise the number of blob marks that we are
*actually* going to use ever again is extremely tiny.

-- 
Felipe Contreras

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Junio C Hamano]

Felipe Contreras <felipe.contreras@gmail.com> writes:

>> The story is different on the fast-import side, where we do say we
>> dump the full table and a later run can depend on these marks.
>
> Yes, and gaining nothing but increased disk-space.

I thought that the "gaining nothing" has already been refuted by the
discussion several hours ago...

cf. http://thread.gmane.org/gmane.comp.version-control.git/223275/focus=223440

Puzzled...

>> By discarding marks on blobs, we may be robbing some optimization
>> possibilities, and by discarding marks on tags, we may be robbing
>> some features, from users of fast-export; we might want to add an
>> option "--use-object-marks={blob,commit,tag}" or something to both
>> fast-export and fast-import, so that the former can optionally write
>> marks for non-commits out, and the latter can omit non commit marks
>> if the user do not need them. But that is a separate issue.
>
> How?

 * if we teach fast-import to optionally not write marks for blobs
   and trees out, your remote-bzr can take advantage of it, because
   it does not reuse marks for non-commits in later runs, right?
   Existing users like cvs2git that do not ask to skip marks for
   non-commits will not be hurt and keep referring to blobs an
   earlier run wrote out.

 * if we teach fast-export to optionally write marks for blobs and
   trees out, the users of fast-export could reuse marks for blobs
   and trees in later runs (perhaps they can drive fast-export from
   the output of "git log --raw", noticing blob object names they
   already saw).  Existing users that do not ask for such a feature
   will not be hurt.

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

On Mon, May 6, 2013 at 3:58 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Felipe Contreras <felipe.contreras@gmail.com> writes:
>
>>> The story is different on the fast-import side, where we do say we
>>> dump the full table and a later run can depend on these marks.
>>
>> Yes, and gaining nothing but increased disk-space.
>
> I thought that the "gaining nothing" has already been refuted by the
> discussion several hours ago...
>
> cf. http://thread.gmane.org/gmane.comp.version-control.git/223275/focus=223440
>
> Puzzled...

What is being gained there? Nothing.

>>> By discarding marks on blobs, we may be robbing some optimization
>>> possibilities, and by discarding marks on tags, we may be robbing
>>> some features, from users of fast-export; we might want to add an
>>> option "--use-object-marks={blob,commit,tag}" or something to both
>>> fast-export and fast-import, so that the former can optionally write
>>> marks for non-commits out, and the latter can omit non commit marks
>>> if the user do not need them. But that is a separate issue.
>>
>> How?
>
>  * if we teach fast-import to optionally not write marks for blobs
>    and trees out, your remote-bzr can take advantage of it,

I already said remote-bzr is irrelevant. *Everybody* benefits.

>    Existing users like cvs2git that do not ask to skip marks for
>    non-commits will not be hurt and keep referring to blobs an
>    earlier run wrote out.

cvs2git does *not* store marks. It doesn't get any benefit or get hurt *at all*.

>  * if we teach fast-export to optionally write marks for blobs and
>    trees out, the users of fast-export could reuse marks for blobs
>    and trees in later runs (perhaps they can drive fast-export from
>    the output of "git log --raw", noticing blob object names they
>    already saw).  Existing users that do not ask for such a feature
>    will not be hurt.

There's absolutely no benefit from fast-export being able to load and
store blobs, the only effect is that the mark files will have tons of
entries nobody will ever use.

If you want to add options for features that only hurt, go ahead, but
the only sane default is to only store commit marks, both for
fast-export, and fast-import. Period.

Cheers.

-- 
Felipe Contreras

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Junio C Hamano]

Felipe Contreras <felipe.contreras@gmail.com> writes:

>>> How?
>>
>>  * if we teach fast-import to optionally not write marks for blobs
>>    and trees out, your remote-bzr can take advantage of it,
>
> I already said remote-bzr is irrelevant. *Everybody* benefits.

Everybody who does _not_ need to look at marks for non-commits from
previous run does.  What about the others who do?

Surely, some lucky ones may get the benefit of a new optimization
for free if you turn it on uncondtionally without an escape hatch,
but that goes against our goal of not to knowingly introduce any
regression.  Michael's cvs2git might have a way to work the breakage
around (I will let him comment on your suggested workaround), but as
long as he has been coding it using the documented feature, why
should he even change anything for no gain at all in the first
place?  Even if you have a workaround, that does not change the fact
that a removal of a feature that somebody has been depending on is a
regression.

What's so difficult to understand that by default the responsibility
of making sure an optimization applies safely to a program that uses
a new optmization lies on that program, in other words, a new
feature is by default an opt-in?

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

On Mon, May 6, 2013 at 8:59 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Felipe Contreras <felipe.contreras@gmail.com> writes:
>
>>>> How?
>>>
>>>  * if we teach fast-import to optionally not write marks for blobs
>>>    and trees out, your remote-bzr can take advantage of it,
>>
>> I already said remote-bzr is irrelevant. *Everybody* benefits.
>
> Everybody who does _not_ need to look at marks for non-commits from
> previous run does.

IOW; everyone.

> What about the others who do?

Like who?

> Surely, some lucky ones may get the benefit of a new optimization
> for free if you turn it on uncondtionally without an escape hatch,
> but that goes against our goal of not to knowingly introduce any
> regression.

That's different. One thing is to turn it on unconditionally, and
another thing is to turn it on by *default*.

> Michael's cvs2git might have a way to work the breakage
> around (I will let him comment on your suggested workaround), but as
> long as he has been coding it using the documented feature, why
> should he even change anything for no gain at all in the first
> place?  Even if you have a workaround, that does not change the fact
> that a removal of a feature that somebody has been depending on is a
> regression.

Who is depending on it? Michael didn't say that he used that feature,
merely that it was documented in cvs2git, because Windows doesn't have
'cat'. He claimed that other people *might* be using that "feature",
but we don't *know*.

Is a couple of commands somebody wrote in some documentation which can
be easily fixed, reason enough to punish everyone else?

> What's so difficult to understand that by default the responsibility
> of making sure an optimization applies safely to a program that uses
> a new optmization lies on that program, in other words, a new
> feature is by default an opt-in?

Is that written in some git bible descended from some god that I
missed? If not, everything are guidelines, and guidelines are there
for a reason, and those reasons can be challenged, and so can the
guidelines.

Sometimes it makes sense to make a new feature opt-in, sometimes it
doesn't, there are no absolutes, there should be no dogmas.

-- 
Felipe Contreras

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

On Mon, May 6, 2013 at 7:31 AM, Jeff King <peff@peff.net> wrote:
> On Sun, May 05, 2013 at 05:38:53PM -0500, Felipe Contreras wrote:
>
>> We don't care about blobs, or any object other than commits, but in
>> order to find the type of object, we are parsing the whole thing, which
>> is slow, specially in big repositories with lots of big files.
>
> I did a double-take on reading this subject line and first paragraph,
> thinking "surely fast-export needs to actually output blobs?".

If you think that, then you are not familiar with the code.

--export-marks=<file>::
	Dumps the internal marks table to <file> when complete.
	Marks are written one per line as `:markid SHA-1`. Only marks
	for revisions are dumped; marks for blobs are ignored.

		if (deco->base && deco->base->type == 1) {
			mark = ptr_to_mark(deco->decoration);
			if (fprintf(f, ":%"PRIu32" %s\n", mark,
				sha1_to_hex(deco->base->sha1)) < 0) {
			    e = 1;
			    break;
			}
		}

> Reading the patch, I see that this is only about not bothering to load
> blob marks from --import-marks. It might be nice to mention that in the
> commit message, which is otherwise quite confusing.

The commit message says it exactly like it is: we don't care about blobs.

If an object is not a commit, we *already* skip it. But as the commit
message already says, we do so by parsing the whole thing.

> I'm also not sure why your claim "we don't care about blobs" is true,
> because naively we would want future runs of fast-export to avoid having
> to write out the whole blob content when mentioning the blob again.

Because it's pointless to have hundreds and thousands of blob marks
that are *never* going to be used, only for an extremely tiny minority
that would.

> Does that match your reasoning?

It doesn't matter, it has been that way since --export-marks was introduced.

>> Before this, loading the objects of a fresh emacs import, with 260598
>> blobs took 14 minutes, after this patch, it takes 3 seconds.
>
> Presumably most of that speed improvement comes from not parsing the
> blob objects. I wonder if you could get similar speedups by applying the
> "do not bother parsing" rule from your patch 3. You would still incur
> some cost to create a "struct blob", but it may or may not be
> measurable.  That would mean we get the "case not worth worrying about"
> from above for free. I doubt it would make that big a difference,
> though, given the rarity of it. So I am OK with it either way.

How would I know if it's a blob or a commit, if not by the code this
patch introduces?

-- 
Felipe Contreras

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Jeff King]

On Mon, May 06, 2013 at 02:02:13PM -0500, Felipe Contreras wrote:

> > I did a double-take on reading this subject line and first paragraph,
> > thinking "surely fast-export needs to actually output blobs?".
> 
> If you think that, then you are not familiar with the code.
> 
> --export-marks=<file>::
> [...]

My point was that nothing in the subject line nor that first paragraph
(nor, for that matter, the entire commit message) says that we are
talking about marks here.

> > Reading the patch, I see that this is only about not bothering to load
> > blob marks from --import-marks. It might be nice to mention that in the
> > commit message, which is otherwise quite confusing.
> 
> The commit message says it exactly like it is: we don't care about blobs.

If you guess that "we" means the marks code and not all of fast-export,
then yes. But I do not have any desire to get into another debate trying
to convince you that there is value to having a clear commit message.
Junio has already proposed a much more readable one.

-Peff

Re: [PATCH v2 2/3] fast-export: improve speed by skipping blobs

[Felipe Contreras]

On Mon, May 6, 2013 at 2:11 PM, Jeff King <peff@peff.net> wrote:
> On Mon, May 06, 2013 at 02:02:13PM -0500, Felipe Contreras wrote:
>
>> > I did a double-take on reading this subject line and first paragraph,
>> > thinking "surely fast-export needs to actually output blobs?".
>>
>> If you think that, then you are not familiar with the code.
>>
>> --export-marks=<file>::
>> [...]
>
> My point was that nothing in the subject line nor that first paragraph
> (nor, for that matter, the entire commit message) says that we are
> talking about marks here.

s/$/ while loading marks/. Fixed.

-- 
Felipe Contreras

[PATCH v2 3/3] fast-export: don't parse all the commits

[Felipe Contreras]

We don't need the parsed objects at this point, merely the information
that they have marks.

Seems to be three times faster in my setup with lots of objects.

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
---
 builtin/fast-export.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/builtin/fast-export.c b/builtin/fast-export.c
index a5b8da8..d1d68e9 100644
--- a/builtin/fast-export.c
+++ b/builtin/fast-export.c
@@ -613,6 +613,7 @@ static void import_marks(char *input_file)
 		char *line_end, *mark_end;
 		unsigned char sha1[20];
 		struct object *object;
+		struct commit *commit;
 		enum object_type type;
 
 		line_end = strchr(line, '\n');
@@ -636,7 +637,11 @@ static void import_marks(char *input_file)
 			/* only commits */
 			continue;
 
-		object = parse_object(sha1);
+		commit = lookup_commit(sha1);
+		if (!commit)
+			die("not a commit? can't happen: %s", sha1_to_hex(sha1));
+
+		object = &commit->object;
 
 		if (object->flags & SHOWN)
 			error("Object %s already has a mark", sha1_to_hex(sha1));
-- 
1.8.3.rc0.401.g45bba44