git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH] fixup! diff: batch fetching of missing blobs

From
Duy Nguyen <pclouds@gmail.com>
Date
Apr 8, 2019, 09:58 UTC
Message-ID
<CACsJy8BHqaqOHVbwtONU5=RiG7Q8WNNAN5EGV_nm7NyNWeyuiQ@mail.gmail.com>
In-Reply-To
<xmqqmul1b0pg.fsf@gitster-ct.c.googlers.com>
On Mon, Apr 8, 2019 at 11:06 AM Junio C Hamano <gitster@pobox.com> wrote:
Show 21 quoted lines
>
> Duy Nguyen <pclouds@gmail.com> writes:
>
> >> Avoid a usage of the_repository by propagating the configured repository
> >> to add_if_missing(). Also, prefetch only if the repository being diffed
> >> is the_repository (because we do not support lazy fetching for any other
> >> repository anyway).
>
> If we are willing to stay limited to the default repository anyway,
> allowing add_if_missing() to take an arbitrary repository does not
> really matter, but before the caller of add_if_missing() befcomes
> ready to work on an arbitrary repository, this change has to happen.
>
> To update the caller, it seems to me that fetch_objects() must learn
> to take an arbitrary repository, but is that the only thing needed?
> After that, the function that the caller resides in and callchain
> upwards can learn to take a repository instance if we want to be
> able to diff inside an arbitrary repository.
>
> But.  Such a change still would not allow us to compare a tree in
> one repository against a tree in another repository.

I feel lost (and the answer "go read partial clone code!" is perfectly acceptable) but why would we need to diff trees of two different repositories?

Show 8 quoted lines
> It is likely
> that a caller with such a need would simply make sure that objects
> in both repositories are available by using the in-core alternate
> object store mechanism, making it a more-or-less moot point to be
> able to pass a repository instance through the callchain X-<.  We
> probably should make it, and spell it out somewhere in a long term
> vision shared among the developers, an explicit goal to get rid of
> the internal (ab)use of the alternate object store mechanism.

I think submodule code so far is doing this way. Though I don't see any reason we need it for submodule code. Objects are not supposed to be shared between the super- and the sub-repo.

Show 190 quoted lines
>
> With squashing the fix-up commit in, the 2/2 patch has become like
> so.
>
> Thanks, both.
>
> -- >8 --
> From: Jonathan Tan <jonathantanmy@google.com>
> Date: Fri, 5 Apr 2019 10:09:34 -0700
> Subject: [PATCH] diff: batch fetching of missing blobs
>
> When running a command like "git show" or "git diff" in a partial clone,
> batch all missing blobs to be fetched as one request.
>
> This is similar to c0c578b33c ("unpack-trees: batch fetching of missing
> blobs", 2017-12-08), but for another command.
>
> Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
> ---
>  diff.c                        |  33 +++++++++++
>  t/t4067-diff-partial-clone.sh | 103 ++++++++++++++++++++++++++++++++++
>  2 files changed, 136 insertions(+)
>  create mode 100755 t/t4067-diff-partial-clone.sh
>
> diff --git a/diff.c b/diff.c
> index ec5c095199..811afbdfb1 100644
> --- a/diff.c
> +++ b/diff.c
> @@ -25,6 +25,7 @@
>  #include "packfile.h"
>  #include "parse-options.h"
>  #include "help.h"
> +#include "fetch-object.h"
>
>  #ifdef NO_FAST_WORKING_DIRECTORY
>  #define FAST_WORKING_DIRECTORY 0
> @@ -6366,8 +6367,40 @@ void diffcore_fix_diff_index(void)
>         QSORT(q->queue, q->nr, diffnamecmp);
>  }
>
> +static void add_if_missing(struct oid_array *to_fetch, struct repository *r,
> +                          const struct diff_filespec *filespec)
> +{
> +       if (filespec && filespec->oid_valid &&
> +           oid_object_info_extended(r, &filespec->oid, NULL,
> +                                    OBJECT_INFO_FOR_PREFETCH))
> +               oid_array_append(to_fetch, &filespec->oid);
> +}
> +
>  void diffcore_std(struct diff_options *options)
>  {
> +       if (options->repo == the_repository &&
> +           repository_format_partial_clone) {
> +               /*
> +                * Prefetch the diff pairs that are about to be flushed.
> +                */
> +               int i;
> +               struct diff_queue_struct *q = &diff_queued_diff;
> +               struct oid_array to_fetch = OID_ARRAY_INIT;
> +
> +               for (i = 0; i < q->nr; i++) {
> +                       struct diff_filepair *p = q->queue[i];
> +                       add_if_missing(&to_fetch, options->repo, p->one);
> +                       add_if_missing(&to_fetch, options->repo, p->two);
> +               }
> +               if (to_fetch.nr)
> +                       /*
> +                        * NEEDSWORK: Consider deduplicating the OIDs sent.
> +                        */
> +                       fetch_objects(repository_format_partial_clone,
> +                                     to_fetch.oid, to_fetch.nr);
> +               oid_array_clear(&to_fetch);
> +       }
> +
>         /* NOTE please keep the following in sync with diff_tree_combined() */
>         if (options->skip_stat_unmatch)
>                 diffcore_skip_stat_unmatch(options);
> diff --git a/t/t4067-diff-partial-clone.sh b/t/t4067-diff-partial-clone.sh
> new file mode 100755
> index 0000000000..90c8fb2901
> --- /dev/null
> +++ b/t/t4067-diff-partial-clone.sh
> @@ -0,0 +1,103 @@
> +#!/bin/sh
> +
> +test_description='behavior of diff when reading objects in a partial clone'
> +
> +. ./test-lib.sh
> +
> +test_expect_success 'git show batches blobs' '
> +       test_when_finished "rm -rf server client trace" &&
> +
> +       test_create_repo server &&
> +       echo a >server/a &&
> +       echo b >server/b &&
> +       git -C server add a b &&
> +       git -C server commit -m x &&
> +
> +       test_config -C server uploadpack.allowfilter 1 &&
> +       test_config -C server uploadpack.allowanysha1inwant 1 &&
> +       git clone --bare --filter=blob:limit=0 "file://$(pwd)/server" client &&
> +
> +       # Ensure that there is exactly 1 negotiation by checking that there is
> +       # only 1 "done" line sent. ("done" marks the end of negotiation.)
> +       GIT_TRACE_PACKET="$(pwd)/trace" git -C client show HEAD &&
> +       grep "git> done" trace >done_lines &&
> +       test_line_count = 1 done_lines
> +'
> +
> +test_expect_success 'diff batches blobs' '
> +       test_when_finished "rm -rf server client trace" &&
> +
> +       test_create_repo server &&
> +       echo a >server/a &&
> +       echo b >server/b &&
> +       git -C server add a b &&
> +       git -C server commit -m x &&
> +       echo c >server/c &&
> +       echo d >server/d &&
> +       git -C server add c d &&
> +       git -C server commit -m x &&
> +
> +       test_config -C server uploadpack.allowfilter 1 &&
> +       test_config -C server uploadpack.allowanysha1inwant 1 &&
> +       git clone --bare --filter=blob:limit=0 "file://$(pwd)/server" client &&
> +
> +       # Ensure that there is exactly 1 negotiation by checking that there is
> +       # only 1 "done" line sent. ("done" marks the end of negotiation.)
> +       GIT_TRACE_PACKET="$(pwd)/trace" git -C client diff HEAD^ HEAD &&
> +       grep "git> done" trace >done_lines &&
> +       test_line_count = 1 done_lines
> +'
> +
> +test_expect_success 'diff skips same-OID blobs' '
> +       test_when_finished "rm -rf server client trace" &&
> +
> +       test_create_repo server &&
> +       echo a >server/a &&
> +       echo b >server/b &&
> +       git -C server add a b &&
> +       git -C server commit -m x &&
> +       echo another-a >server/a &&
> +       git -C server add a &&
> +       git -C server commit -m x &&
> +
> +       test_config -C server uploadpack.allowfilter 1 &&
> +       test_config -C server uploadpack.allowanysha1inwant 1 &&
> +       git clone --bare --filter=blob:limit=0 "file://$(pwd)/server" client &&
> +
> +       echo a | git hash-object --stdin >hash-old-a &&
> +       echo another-a | git hash-object --stdin >hash-new-a &&
> +       echo b | git hash-object --stdin >hash-b &&
> +
> +       # Ensure that only a and another-a are fetched.
> +       GIT_TRACE_PACKET="$(pwd)/trace" git -C client diff HEAD^ HEAD &&
> +       grep "want $(cat hash-old-a)" trace &&
> +       grep "want $(cat hash-new-a)" trace &&
> +       ! grep "want $(cat hash-b)" trace
> +'
> +
> +test_expect_success 'diff with rename detection batches blobs' '
> +       test_when_finished "rm -rf server client trace" &&
> +
> +       test_create_repo server &&
> +       echo a >server/a &&
> +       printf "b\nb\nb\nb\nb\n" >server/b &&
> +       git -C server add a b &&
> +       git -C server commit -m x &&
> +       rm server/b &&
> +       printf "b\nb\nb\nb\nbX\n" >server/c &&
> +       git -C server add c &&
> +       git -C server commit -a -m x &&
> +
> +       test_config -C server uploadpack.allowfilter 1 &&
> +       test_config -C server uploadpack.allowanysha1inwant 1 &&
> +       git clone --bare --filter=blob:limit=0 "file://$(pwd)/server" client &&
> +
> +       # Ensure that there is exactly 1 negotiation by checking that there is
> +       # only 1 "done" line sent. ("done" marks the end of negotiation.)
> +       GIT_TRACE_PACKET="$(pwd)/trace" git -C client diff -M HEAD^ HEAD >out &&
> +       grep "similarity index" out &&
> +       grep "git> done" trace >done_lines &&
> +       test_line_count = 1 done_lines
> +'
> +
> +test_done
> --
> 2.21.0-196-g041f5ea1cf
>
-- 
Duy
Previous: Junio C HamanoNext: Junio C Hamano
Message 26 of 29 in “diff: batch fetching of missing blobs”
  1. diff: batch fetching of missing blobsJonathan Tan, Mar 26, 2019
  2. SZEDER GáborMar 27, 2019
  3. Johannes SchindelinMar 27, 2019
  4. Jeff KingMar 28, 2019
  5. 0/2 Batch fetching of missing blobs in diff and showJonathan Tan, Mar 29, 2019
  6. 1/2 sha1-file: support OBJECT_INFO_FOR_PREFETCHJonathan Tan, Mar 29, 2019
  7. Johannes SchindelinApr 5, 2019
  8. Jeff KingApr 5, 2019
  9. 2/2 diff: batch fetching of missing blobsJonathan Tan, Mar 29, 2019
  10. SZEDER GáborApr 4, 2019
  11. Johannes SchindelinApr 5, 2019
  12. Christian CouderApr 7, 2019
  13. Junio C HamanoApr 8, 2019
  14. Junio C HamanoApr 8, 2019
  15. Junio C HamanoApr 8, 2019
  16. Christian CouderApr 8, 2019
  17. Christian CouderApr 8, 2019
  18. Junio C HamanoApr 8, 2019
  19. Christian CouderApr 8, 2019
  20. Duy NguyenApr 5, 2019
  21. fixup! diff: batch fetching of missing blobsJonathan Tan, Apr 5, 2019
  22. Johannes SchindelinApr 5, 2019
  23. Duy NguyenApr 6, 2019
  24. Junio C HamanoApr 8, 2019
  25. Junio C HamanoApr 8, 2019
  26. Duy NguyenApr 8, 2019
  27. Junio C HamanoApr 9, 2019
  28. Johannes SchindelinApr 5, 2019
  29. Jeff KingApr 5, 2019

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.