This is the first "payoff" series in the sparse-index work. It makes 'git status' very fast when a sparse-index is enabled on a repository with cone-mode sparse-checkout (and a small populated set). This is based on ds/sparse-index-protections AND mt/add-rm-sparse-checkout. The latter branch is needed because it changes the behavior of 'git add' around sparse entries, which changes the expectations of a test added in patch 1. The approach here is to audit the places where ensure_full_index() pops up while doing normal commands with pathspecs within the sparse-checkout definition. Each of these are checked and tested. In the end, the sparse-index is integrated with these features: * git status * FS Monitor index extension. The performance tests in p2000-sparse-operations.sh improve by 95% or more, even when compared with the full-index cases, not just the sparse-index cases that previously had extra overhead. Hopefully this is the first example of how ds/sparse-index-protections has done the basic work to do these conversions safely, making them look easier than they seemed when starting this adventure. Thanks, -Stolee Updates in V5 ============= I replaced one patch with a few that are more complicated. The reason is that I started integrating with git checkout and realized that some of the changes I was making in unpack-trees.c were incorrect for that situation, so I might as well do them right here. The tests can't demonstrate the bugs with the previous case until we integrate with git checkout, which will follow in another series after this one is submitted. For testing, I've integrated this series along with extensions that work for git commit and git checkout into the Scalar functional tests, which test many scenarios with cone mode sparse-checkout and hence provides good evidence that this is working correctly. Updates in V4 ============= * The previous patch "unpack-trees: stop recursing into sparse directories" was confusing, and actually a bit sloppy. * It has been replaced with "unpack-trees: be careful around sparse directory entries" which takes the sparse-directory checks and raises them higher up into unpack_trees.c instead of in diff-lib.c. Updates in V3 ============= Sorry that this was a long time coming. I got a little side-tracked on other projects, but I also worked to get the sparse-index feature working against the Scalar functional tests, which contain many special cases around the sparse-checkout feature as they were inherited from special cases that arose in the virtualized environment of VFS for Git. This version contains my fixes based on that investigation. Most of these were easy to identify and fix, but I was blocked for a long time struggling with a bug when combining the sparse-index with the builtin FS Monitor feature, but I've reported my findings already [1]. [1] https://lore.kernel.org/git/0b9e54ba-ac27-e537-7bef-1b4448f92352@xxxxxxxxx/ * Updated comments and tests based on the v2 feedback. * Expanded the test repository data shape based on the special cases found during my investigation. * Added several commits that either fix errors in the status code, or fix errors in the previous sparse-index series, specifically: * When in a conflict state, the cache-tree fails to update. For now, skip writing a sparse-index until this can be resolved more carefully. * When expanding a sparse-directory entry, we set the CE_SKIP_WORKTREE bit but forgot the CE_EXTENDED bit. * git status had failures if there was a sparse-directory entry as the first entry within a directory. * When expanding a directory to report its status, such as when a sparse-directory is staged but doesn't exist at HEAD (such as in an orphaned commit) we did not previously recurse correctly into subdirectories. * Be extra careful with the FS Monitor data when expanding or contracting an index. This version now abandons all FS Monitor data at these conversion points with the expectation that in the future these conversions will be rare so the FS Monitor feature can work efficiently. Updates in V2 ---------------------------------------------------------------------------- * Based on the feedback, it is clear that 'git add' will require much more careful testing and thought. I'm splitting it out of this series and it will return with a follow-up. * Test cases are improved, both in coverage and organization. * The previous "unpack-trees: make sparse aware" patch is split into three now. * Stale messages based on an old implementation of the "protections" topic are now fixed. * Performance tests were re-run. Derrick Stolee (14): sparse-index: skip indexes with unmerged entries sparse-index: include EXTENDED flag when expanding t1092: replace incorrect 'echo' with 'cat' t1092: expand repository data shape t1092: add tests for status/add and sparse files unpack-trees: preserve cache_bottom unpack-trees: compare sparse directories correctly unpack-trees: unpack sparse directory entries dir.c: accept a directory as part of cone-mode patterns diff-lib: handle index diffs with sparse dirs status: skip sparse-checkout percentage with sparse-index status: use sparse-index throughout wt-status: expand added sparse directory entries fsmonitor: integrate with sparse index builtin/commit.c | 3 + diff-lib.c | 188 +++++++++++++++++++++++ dir.c | 11 ++ read-cache.c | 10 +- sparse-index.c | 27 +++- t/t1092-sparse-checkout-compatibility.sh | 158 ++++++++++++++++++- t/t7519-status-fsmonitor.sh | 48 ++++++ unpack-trees.c | 121 +++++++++++++-- wt-status.c | 64 +++++++- wt-status.h | 1 + 10 files changed, 607 insertions(+), 24 deletions(-) base-commit: f723f370c89ad61f4f40aabfd3540b1ce19c00e5 Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-932%2Fderrickstolee%2Fsparse-index%2Fstatus-and-add-v5 Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-932/derrickstolee/sparse-index/status-and-add-v5 Pull-Request: https://github.com/gitgitgadget/git/pull/932 Range-diff vs v4: 1: 5a2ed3d1d701 = 1: 5a2ed3d1d701 sparse-index: skip indexes with unmerged entries 2: 8aa41e749471 = 2: 8aa41e749471 sparse-index: include EXTENDED flag when expanding -: ------------ > 3: b99371c7dd61 t1092: replace incorrect 'echo' with 'cat' 3: 70971b1f9261 ! 4: f4dddac1859e t1092: expand repository data shape @@ t/t1092-sparse-checkout-compatibility.sh: test_expect_success 'setup' ' + mkdir deep/deeper1/0/0 && + touch deep/deeper1/0/1 && + touch deep/deeper1/0/0/0 && ++ cp -r deep/deeper1/0 folder1 && ++ cp -r deep/deeper1/0 folder2 && ++ echo >>folder1/0/0/0 && ++ echo >>folder2/0/1 && git add . && git commit -m "initial commit" && git checkout -b base && +@@ t/t1092-sparse-checkout-compatibility.sh: test_expect_success 'setup' ' + mv folder1/a folder2/b && + mv folder1/larger-content folder2/edited-content && + echo >>folder2/edited-content && ++ echo >>folder2/0/1 && ++ echo stuff >>deep/deeper1/a && + git add . && + git commit -m "rename folder1/... to folder2/..." && + + git checkout -b rename-out-to-in rename-base && + mv folder1/a deep/deeper1/b && ++ echo more stuff >>deep/deeper1/a && ++ rm folder2/0/1 && ++ mkdir folder2/0/1 && ++ echo >>folder2/0/1/1 && + mv folder1/larger-content deep/deeper1/edited-content && + echo >>deep/deeper1/edited-content && + git add . && +@@ t/t1092-sparse-checkout-compatibility.sh: test_expect_success 'setup' ' + + git checkout -b rename-in-to-out rename-base && + mv deep/deeper1/a folder1/b && ++ echo >>folder2/0/1 && ++ rm -rf folder1/0/0 && ++ echo >>folder1/0/0 && + mv deep/deeper1/larger-content folder1/edited-content && + echo >>folder1/edited-content && + git add . && +@@ t/t1092-sparse-checkout-compatibility.sh: test_expect_success 'diff --staged' ' + test_all_match git diff --staged + ' + +-test_expect_success 'diff with renames' ' ++test_expect_success 'diff with renames and conflicts' ' + init_repos && + + for branch in rename-out-to-out rename-out-to-in rename-in-to-out + do + test_all_match git checkout rename-base && + test_all_match git checkout $branch -- .&& ++ test_all_match git status --porcelain=v2 && ++ test_all_match git diff --staged --no-renames && ++ test_all_match git diff --staged --find-renames || return 1 ++ done ++' ++ ++test_expect_success 'diff with directory/file conflicts' ' ++ init_repos && ++ ++ for branch in rename-out-to-out rename-out-to-in rename-in-to-out ++ do ++ git -C full-checkout reset --hard && ++ test_sparse_match git reset --hard && ++ test_all_match git checkout $branch && ++ test_all_match git checkout rename-base -- . && ++ test_all_match git status --porcelain=v2 && + test_all_match git diff --staged --no-renames && + test_all_match git diff --staged --find-renames || return 1 + done 4: a80b5a41153f = 5: 856346b72f79 t1092: add tests for status/add and sparse files 5: 07a45b661c4a = 6: f3f6223e955f unpack-trees: preserve cache_bottom 6: cc4a526e7947 = 7: 45ae96adf285 unpack-trees: compare sparse directories correctly 7: e28df7f9395d ! 8: 724194eef9f6 unpack-trees: be careful around sparse directory entries @@ Metadata Author: Derrick Stolee <dstolee@xxxxxxxxxxxxx> ## Commit message ## - unpack-trees: be careful around sparse directory entries + unpack-trees: unpack sparse directory entries - The methods traverse_by_cache_tree() and unpack_nondirectories() have - similar behavior in trying to demonstrate the difference between and - index and a tree, with some differences about how they walk the index. + During unpack_callback(), index entries are compared against tree + entries. These are matched according to names and types. One goal is to + decide if we should recurse into subtrees or simply operate on one index + entry. - Each of these is expecting every cache entry to correspond to a file - path. We need to skip over the sparse directory entries in the case of a - sparse-index. Those entries are discovered in the portion that looks for - subtrees among the cache entries by scanning the paths for slashes. + In the case of a sparse-directory entry, we do not want to recurse into + that subtree and instead simply compare the trees. In some cases, we + might want to perform a merge operation on the entry, such as during + 'git checkout <commit>' which wants to replace a sparse tree entry with + the tree for that path at the target commit. We extend the logic within + unpack_nondirectories() to create a sparse-directory entry in this case, + and then that is sent to call_unpack_fn(). - Skipping these sparse directory entries will have a measurable effect - when we relax 'git status' to work with sparse-indexes: without this - change these methods would call call_unpack_fn() which in turn calls - oneway_diff() and then shows these sparse directory entries as added or - modified files. + There are some subtleties in this process. For instance, we need to + update find_cache_entry() to allow finding a sparse-directory entry that + exactly matches a given path. Signed-off-by: Derrick Stolee <dstolee@xxxxxxxxxxxxx> ## unpack-trees.c ## -@@ unpack-trees.c: static int traverse_by_cache_tree(int pos, int nr_entries, int nr_names, +@@ unpack-trees.c: static struct cache_entry *create_ce_entry(const struct traverse_info *info, + const struct name_entry *n, + int stage, + struct index_state *istate, +- int is_transient) ++ int is_transient, ++ int is_sparse_directory) + { + size_t len = traverse_path_len(info, tree_entry_len(n)); ++ size_t alloc_len = is_sparse_directory ? len + 1 : len; + struct cache_entry *ce = + is_transient ? +- make_empty_transient_cache_entry(len) : +- make_empty_cache_entry(istate, len); ++ make_empty_transient_cache_entry(alloc_len) : ++ make_empty_cache_entry(istate, alloc_len); - src[0] = o->src_index->cache[pos + i]; + ce->ce_mode = create_ce_mode(n->mode); + ce->ce_flags = create_ce_flags(stage); +@@ unpack-trees.c: static struct cache_entry *create_ce_entry(const struct traverse_info *info, + /* len+1 because the cache_entry allocates space for NUL */ + make_traverse_path(ce->name, len + 1, info, n->path, n->pathlen); -+ if (S_ISSPARSEDIR(src[0]->ce_mode)) -+ continue; ++ if (is_sparse_directory) { ++ ce->name[len] = '/'; ++ ce->name[len + 1] = 0; ++ ce->ce_namelen++; ++ ce->ce_flags |= CE_SKIP_WORKTREE; ++ } + - len = ce_namelen(src[0]); - new_ce_len = cache_entry_size(len); + return ce; + } @@ unpack-trees.c: static int unpack_nondirectories(int n, unsigned long mask, + unsigned long dirmask, + struct cache_entry **src, + const struct name_entry *names, +- const struct traverse_info *info) ++ const struct traverse_info *info, ++ int sparse_directory) + { + int i; + struct unpack_trees_options *o = info->data; + unsigned long conflicts = info->df_conflicts | dirmask; + +- /* Do we have *only* directories? Nothing to do */ if (mask == dirmask && !src[0]) return 0; -+ if (src[0] && S_ISSPARSEDIR(src[0]->ce_mode)) ++ /* no-op if our cache entry doesn't match the expectations. */ ++ if (sparse_directory) { ++ if (src[0] && !S_ISSPARSEDIR(src[0]->ce_mode)) ++ BUG("expected sparse directory entry"); ++ } else if (src[0] && S_ISSPARSEDIR(src[0]->ce_mode)) { + return 0; ++ } + /* * Ok, we've filled in up to any potential index entry in src[0], * now do the rest. +@@ unpack-trees.c: static int unpack_nondirectories(int n, unsigned long mask, + * not stored in the index. otherwise construct the + * cache entry from the index aware logic. + */ +- src[i + o->merge] = create_ce_entry(info, names + i, stage, &o->result, o->merge); ++ src[i + o->merge] = create_ce_entry(info, names + i, stage, ++ &o->result, o->merge, ++ sparse_directory); + } + + if (o->merge) { +@@ unpack-trees.c: static int find_cache_pos(struct traverse_info *info, + static struct cache_entry *find_cache_entry(struct traverse_info *info, + const struct name_entry *p) + { ++ struct cache_entry *ce; + int pos = find_cache_pos(info, p->path, p->pathlen); + struct unpack_trees_options *o = info->data; + + if (0 <= pos) + return o->src_index->cache[pos]; +- else ++ ++ /* ++ * Check for a sparse-directory entry named "path/". ++ * Due to the input p->path not having a trailing ++ * slash, the negative 'pos' value overshoots the ++ * expected position by one, hence "-2" here. ++ */ ++ pos = -pos - 2; ++ ++ if (pos < 0 || pos >= o->src_index->cache_nr) ++ return NULL; ++ ++ ce = o->src_index->cache[pos]; ++ ++ if (!S_ISSPARSEDIR(ce->ce_mode)) + return NULL; ++ ++ /* ++ * Compare ce->name to info->name + '/' + p->path + '/' ++ * if info->name is non-empty. Compare ce->name to ++ * p-.path + '/' otherwise. ++ */ ++ if (info->namelen) { ++ if (ce->ce_namelen == info->namelen + p->pathlen + 2 && ++ ce->name[info->namelen] == '/' && ++ !strncmp(ce->name, info->name, info->namelen) && ++ !strncmp(ce->name + info->namelen + 1, p->path, p->pathlen)) ++ return ce; ++ } else if (ce->ce_namelen == p->pathlen + 1 && ++ !strncmp(ce->name, p->path, p->pathlen)) ++ return ce; ++ return NULL; + } + + static void debug_path(struct traverse_info *info) +@@ unpack-trees.c: static void debug_unpack_callback(int n, + debug_name_entry(i, names + i); + } + ++/* ++ * Returns true if and only if the given cache_entry is a ++ * sparse-directory entry that matches the given name_entry ++ * from the tree walk at the given traverse_info. ++ */ ++static int is_sparse_directory_entry(struct cache_entry *ce, struct name_entry *name, struct traverse_info *info) ++{ ++ size_t expected_len, name_start; ++ ++ if (!ce || !name || !S_ISSPARSEDIR(ce->ce_mode)) ++ return 0; ++ ++ if (info->namelen) ++ name_start = info->namelen + 1; ++ else ++ name_start = 0; ++ expected_len = name->pathlen + 1 + name_start; ++ ++ if (ce->ce_namelen != expected_len || ++ strncmp(ce->name, info->name, info->namelen) || ++ strncmp(ce->name + name_start, name->path, name->pathlen)) ++ return 0; ++ ++ return 1; ++} ++ + /* + * Note that traverse_by_cache_tree() duplicates some logic in this function + * without actually calling it. If you change the logic here you may need to +@@ unpack-trees.c: static int unpack_callback(int n, unsigned long mask, unsigned long dirmask, str + } + } + +- if (unpack_nondirectories(n, mask, dirmask, src, names, info) < 0) ++ if (unpack_nondirectories(n, mask, dirmask, src, names, info, 0) < 0) + return -1; + + if (o->merge && src[0]) { +@@ unpack-trees.c: static int unpack_callback(int n, unsigned long mask, unsigned long dirmask, str + } + } + +- if (traverse_trees_recursive(n, dirmask, mask & ~dirmask, +- names, info) < 0) ++ if (is_sparse_directory_entry(src[0], names, info)) { ++ if (unpack_nondirectories(n, dirmask, mask & ~dirmask, src, names, info, 1) < 0) ++ return -1; ++ } else if (traverse_trees_recursive(n, dirmask, mask & ~dirmask, ++ names, info) < 0) { + return -1; ++ } ++ + return mask; + } + 8: 2cc3a93d4434 = 9: b8ff179f43e3 dir.c: accept a directory as part of cone-mode patterns -: ------------ > 10: b9b97e011293 diff-lib: handle index diffs with sparse dirs 9: 5011feb1aa04 = 11: 611b9f61fb2c status: skip sparse-checkout percentage with sparse-index 10: 9f2ce5301dc9 = 12: 0c0a765dde80 status: use sparse-index throughout 11: 24417e095243 ! 13: 02f2c7b63982 wt-status: expand added sparse directory entries @@ t/t1092-sparse-checkout-compatibility.sh: test_expect_success 'sparse-index is n + init_repos && + + test_all_match git checkout rename-out-to-in && -+ test_all_match git reset --mixed HEAD~1 && ++ ++ # Sparse checkouts do not agree with full checkouts about ++ # how to report a directory/file conflict during a reset. ++ # This command would fail with test_all_match because the ++ # full checkout reports "T folder1/0/1" while a sparse ++ # checkout reports "D folder1/0/1". This matches because ++ # the sparse checkouts skip "adding" the other side of ++ # the conflict. ++ test_sparse_match git reset --mixed HEAD~1 && + test_sparse_match test-tool read-cache --table --expand && -+ test_all_match git status --porcelain=v2 && -+ test_all_match git status --porcelain=v2 && ++ test_sparse_match git status --porcelain=v2 && ++ test_sparse_match git status --porcelain=v2 && + + # At this point, sparse-checkouts behave differently + # from the full-checkout. 12: 584d4b559a91 = 14: 46ca150c3548 fsmonitor: integrate with sparse index -- gitgitgadget
