[PATCH v3 0/3] Integrate changed-path Bloom filters with 'git blame'

["Derrick Stolee via GitGitGadget" <gitgitgadget@gmail.com>]

If the changed-path Bloom filters are relatively stable, then I propose
trying to build upon them as a way to discover any deficiencies. Also, it's
good to use them when we can.

The goal I set out to achieve was to use Bloom filters as much as possible
in git blame. I think I have achieved most of that, except I did not
integrate it with the -C mode. In that case, the blob-diff computation takes
a majority of the computation time, so short-circuiting the tree diff using
Bloom filters. Also, it's just really complicated. If someone else thinks
there is an easy win there, then please go ahead and give it a try with the
extra logic presented here in PATCH 3.

While I was playing around with Bloom filters and git blame, I purposefully
got it working with some scenarios but not all. Then I tried to trigger a
failing build in the blame tests using GIT_TEST_COMMIT_GRAPH=1 and 
GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS=1. But the tests all succeeded!

Examining the data, I see that the commit-graph didn't have the Bloom filter
chunks at all. This is because we are not setting the flag to write them in
the right spot. The GIT_TEST_COMMIT_GRAPH=1 variable triggers a commit-graph
write during git commit, so we need to update the code there instead of just
inspecting the variable in git commit-graph write. (This is PATCH 2.)

By updating this variable, I saw some test failures in other tests regarding
non-exact pathspecs. I fixed these in PATCH 1 so we keep clean builds.

I based this change on [1] but it would apply cleanly (and logically) on
gs/commit-graph-path-filter

Updates in v2:

 * Added PATCH 3 to write commit-graph files during 'git merge' if
   GIT_TEST_COMMIT_GRAPH is enabled.
   
   
 * Updated PATCH 1 with the simplification recommended by Taylor.
   
   
 * Fixed the lower-case "bloom" in the commit message.
   
   

Thanks, -Stolee

[1] 
https://lore.kernel.org/git/pull.601.v2.git.1586437211842.gitgitgadget@gmail.com/

Derrick Stolee (3):
  revision: complicated pathspecs disable filters
  tests: write commit-graph with Bloom filters
  blame: use changed-path Bloom filters

 blame.c          | 139 ++++++++++++++++++++++++++++++++++++++++++++---
 blame.h          |   6 ++
 builtin/blame.c  |  10 ++++
 builtin/commit.c |   4 +-
 builtin/merge.c  |   7 ++-
 commit-graph.c   |  14 +++++
 commit-graph.h   |   9 +++
 revision.c       |  19 ++++++-
 8 files changed, 193 insertions(+), 15 deletions(-)


base-commit: f4df00a0dd448edce0e854a97f63598fefe27d27
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-609%2Fderrickstolee%2Fbloom-blame-v3
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-609/derrickstolee/bloom-blame-v3
Pull-Request: https://github.com/gitgitgadget/git/pull/609

Range-diff vs v2:

 1:  adc03eee4ac = 1:  adc03eee4ac revision: complicated pathspecs disable filters
 2:  7e8f1aed113 < -:  ----------- commit: write commit-graph with Bloom filters
 3:  824f8ad067b ! 2:  4073c8fe42f commit-graph: write commit-graph in more tests
     @@ Metadata
      Author: Derrick Stolee <dstolee@microsoft.com>
      
       ## Commit message ##
     -    commit-graph: write commit-graph in more tests
     +    tests: write commit-graph with Bloom filters
      
     -    The GIT_TEST_COMMIT_GRAPH test environment variable triggers
     -    commit-graph writes during each "git commit" process. To expand
     -    the number of tests that have commits in the commit-graph file,
     -    add a helper method that computes the commit-graph and place
     +    The GIT_TEST_COMMIT_GRAPH environment variable updates the commit-
     +    graph file whenever "git commit" is run, ensuring that we always
     +    have an updated commit-graph throughout the test suite. The
     +    GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS environment variable was
     +    introduced to write the changed-path Bloom filters whenever "git
     +    commit-graph write" is run. However, the GIT_TEST_COMMIT_GRAPH
     +    trick doesn't launch a separate process and instead writes it
     +    directly.
     +
     +    To expand the number of tests that have commits in the commit-graph
     +    file, add a helper method that computes the commit-graph and place
          that helper inside "git commit" and "git merge".
      
     +    In the helper method, check GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS
     +    to ensure we are writing changed-path Bloom filters whenever
     +    possible.
     +
          Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
      
       ## builtin/commit.c ##
     @@ commit-graph.c
       
      +void git_test_write_commit_graph_or_die(void)
      +{
     -+	if (git_env_bool(GIT_TEST_COMMIT_GRAPH, 0) &&
     -+	    write_commit_graph_reachable(the_repository->objects->odb, 0, NULL))
     ++	int flags = 0;
     ++	if (!git_env_bool(GIT_TEST_COMMIT_GRAPH, 0))
     ++		return;
     ++
     ++	if (git_env_bool(GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS, 0))
     ++		flags = COMMIT_GRAPH_WRITE_BLOOM_FILTERS;
     ++	
     ++	if (write_commit_graph_reachable(the_repository->objects->odb,
     ++					 flags, NULL))
      +		die("failed to write commit-graph under GIT_TEST_COMMIT_GRAPH");
      +}
      +
     @@ commit-graph.h
       #define GIT_TEST_COMMIT_GRAPH_DIE_ON_LOAD "GIT_TEST_COMMIT_GRAPH_DIE_ON_LOAD"
       #define GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS "GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS"
       
     ++/*
     ++ * This method is only used to enhance coverage of the commit-graph
     ++ * feature in the test suite with the GIT_TEST_COMMIT_GRAPH and
     ++ * GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS environment variables. Do not
     ++ * call this method oustide of a builtin, and only if you know what
     ++ * you are doing!
     ++ */
      +void git_test_write_commit_graph_or_die(void);
      +
       struct commit;
 4:  4ae196d6355 = 3:  463d6bf5033 blame: use changed-path Bloom filters

-- 
gitgitgadget