From f30fff45a752cb0781067ad48c283e49345a5813 Mon Sep 17 00:00:00 2001 From: Russell Belfer Date: Wed, 19 Jun 2013 15:27:25 -0700 Subject: [PATCH] Add index pathspec-based operations This adds three new public APIs for manipulating the index: 1. `git_index_add_all` is similar to `git add -A` and will add files in the working directory that match a pathspec to the index while honoring ignores, etc. 2. `git_index_remove_all` removes files from the index that match a pathspec. 3. `git_index_update_all` updates entries in the index based on the current contents of the working directory, either added the new information or removing the entry from the index. --- include/git2/index.h | 115 +++++++++++++++++ src/diff.c | 6 +- src/index.c | 215 +++++++++++++++++++++++++++++++- tests-clar/index/addall.c | 254 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 586 insertions(+), 4 deletions(-) create mode 100644 tests-clar/index/addall.c diff --git a/include/git2/index.h b/include/git2/index.h index 58b0243e0..399d7c9a8 100644 --- a/include/git2/index.h +++ b/include/git2/index.h @@ -11,6 +11,7 @@ #include "indexer.h" #include "types.h" #include "oid.h" +#include "strarray.h" /** * @file git2/index.h @@ -125,6 +126,18 @@ typedef enum { GIT_INDEXCAP_FROM_OWNER = ~0u } git_indexcap_t; +/** Callback for APIs that add/remove/update files matching pathspec */ +typedef int (*git_index_matched_path_cb)( + const char *path, const char *matched_pathspec, void *payload); + +/** Flags for APIs that add files matching pathspec */ +typedef enum { + GIT_INDEX_ADD_DEFAULT = 0, + GIT_INDEX_ADD_FORCE = (1u << 0), + GIT_INDEX_ADD_DISABLE_PATHSPEC_MATCH = (1u << 1), + GIT_INDEX_ADD_CHECK_PATHSPEC = (1u << 2), +} git_index_add_option_t; + /** @name Index File Functions * * These functions work on the index file itself. @@ -420,6 +433,108 @@ GIT_EXTERN(int) git_index_add_bypath(git_index *index, const char *path); */ GIT_EXTERN(int) git_index_remove_bypath(git_index *index, const char *path); +/** + * Add or update index entries matching files in the working directory. + * + * This method will fail in bare index instances. + * + * The `pathspec` is a list of file names or shell glob patterns that will + * matched against files in the repository's working directory. Each file + * that matches will be added to the index (either updating an existing + * entry or adding a new entry). You can disable glob expansion and force + * exact matching with the `GIT_INDEX_ADD_DISABLE_PATHSPEC_MATCH` flag. + * + * Files that are ignored will be skipped (unlike `git_index_add_bypath`). + * If a file is already tracked in the index, then it *will* be updated + * even if it is ignored. Pass the `GIT_INDEX_ADD_FORCE` flag to + * skip the checking of ignore rules. + * + * To emulate `git add -A` and generate an error if the pathspec contains + * the exact path of an ignored file (when not using FORCE), add the + * `GIT_INDEX_ADD_CHECK_PATHSPEC` flag. This checks that each entry + * in the `pathspec` that is an exact match to a filename on disk is + * either not ignored or already in the index. If this check fails, the + * function will return GIT_EINVALIDSPEC. + * + * To emulate `git add -A` with the "dry-run" option, just use a callback + * function that always returns a positive value. See below for details. + * + * If any files are currently the result of a merge conflict, those files + * will no longer be marked as conflicting. The data about the conflicts + * will be moved to the "resolve undo" (REUC) section. + * + * If you provide a callback function, it will be invoked on each matching + * item in the working directory immediately *before* it is added to / + * updated in the index. Returning zero will add the item to the index, + * greater than zero will skip the item, and less than zero will abort the + * scan and cause GIT_EUSER to be returned. + * + * @param index an existing index object + * @param pathspec array of path patterns + * @param flags combination of git_index_add_option_t flags + * @param callback notification callback for each added/updated path (also + * gets index of matching pathspec entry); can be NULL; + * return 0 to add, >0 to skip, <0 to abort scan. + * @param payload payload passed through to callback function + * @return 0 or an error code + */ +GIT_EXTERN(int) git_index_add_all( + git_index *index, + const git_strarray *pathspec, + unsigned int flags, + git_index_matched_path_cb callback, + void *payload); + +/** + * Remove all matching index entries. + * + * If you provide a callback function, it will be invoked on each matching + * item in the index immediately *before* it is removed. Return 0 to + * remove the item, > 0 to skip the item, and < 0 to abort the scan. + * + * @param index An existing index object + * @param pathspec array of path patterns + * @param callback notification callback for each removed path (also + * gets index of matching pathspec entry); can be NULL; + * return 0 to add, >0 to skip, <0 to abort scan. + * @param payload payload passed through to callback function + * @return 0 or an error code + */ +GIT_EXTERN(int) git_index_remove_all( + git_index *index, + const git_strarray *pathspec, + git_index_matched_path_cb callback, + void *payload); + +/** + * Update all index entries to match the working directory + * + * This method will fail in bare index instances. + * + * This scans the existing index entries and synchronizes them with the + * working directory, deleting them if the corresponding working directory + * file no longer exists otherwise updating the information (including + * adding the latest version of file to the ODB if needed). + * + * If you provide a callback function, it will be invoked on each matching + * item in the index immediately *before* it is updated (either refreshed + * or removed depending on working directory state). Return 0 to proceed + * with updating the item, > 0 to skip the item, and < 0 to abort the scan. + * + * @param index An existing index object + * @param pathspec array of path patterns + * @param callback notification callback for each updated path (also + * gets index of matching pathspec entry); can be NULL; + * return 0 to add, >0 to skip, <0 to abort scan. + * @param payload payload passed through to callback function + * @return 0 or an error code + */ +GIT_EXTERN(int) git_index_update_all( + git_index *index, + const git_strarray *pathspec, + git_index_matched_path_cb callback, + void *payload); + /** * Find the first position of any entries which point to given * path in the Git index. diff --git a/src/diff.c b/src/diff.c index fa2c5c71d..633601699 100644 --- a/src/diff.c +++ b/src/diff.c @@ -675,8 +675,10 @@ static int maybe_modified( } } - /* if oids and modes match, then file is unmodified */ - else if (git_oid_equal(&oitem->oid, &nitem->oid) && omode == nmode) + /* if oids and modes match (and are valid), then file is unmodified */ + else if (git_oid_equal(&oitem->oid, &nitem->oid) && + omode == nmode && + !git_oid_iszero(&oitem->oid)) status = GIT_DELTA_UNMODIFIED; /* if we have an unknown OID and a workdir iterator, then check some diff --git a/src/index.c b/src/index.c index 560a257e7..e65dc052c 100644 --- a/src/index.c +++ b/src/index.c @@ -15,6 +15,8 @@ #include "hash.h" #include "iterator.h" #include "pathspec.h" +#include "ignore.h" + #include "git2/odb.h" #include "git2/oid.h" #include "git2/blob.h" @@ -997,7 +999,7 @@ static int index_conflict__get_byindex( int stage, len = 0; assert(ancestor_out && our_out && their_out && index); - + *ancestor_out = NULL; *our_out = NULL; *their_out = NULL; @@ -1010,7 +1012,7 @@ static int index_conflict__get_byindex( stage = GIT_IDXENTRY_STAGE(conflict_entry); path = conflict_entry->path; - + switch (stage) { case 3: *their_out = conflict_entry; @@ -2044,3 +2046,212 @@ git_repository *git_index_owner(const git_index *index) { return INDEX_OWNER(index); } + +int git_index_add_all( + git_index *index, + const git_strarray *paths, + unsigned int flags, + git_index_matched_path_cb cb, + void *payload) +{ + int error; + git_repository *repo; + git_iterator *wditer = NULL; + const git_index_entry *wd = NULL; + git_index_entry *entry; + git_pathspec_context ps; + const char *match; + size_t existing; + bool no_fnmatch = (flags & GIT_INDEX_ADD_DISABLE_PATHSPEC_MATCH) != 0; + int ignorecase; + git_oid blobid; + + assert(index); + + if (INDEX_OWNER(index) == NULL) + return create_index_error(-1, + "Could not add paths to index. " + "Index is not backed up by an existing repository."); + + repo = INDEX_OWNER(index); + if ((error = git_repository__ensure_not_bare(repo, "index add all")) < 0) + return error; + + if (git_repository__cvar(&ignorecase, repo, GIT_CVAR_IGNORECASE) < 0) + return -1; + + if ((error = git_pathspec_context_init(&ps, paths)) < 0) + return error; + + /* optionally check that pathspec doesn't mention any ignored files */ + if ((flags & GIT_INDEX_ADD_CHECK_PATHSPEC) != 0 && + (flags & GIT_INDEX_ADD_FORCE) == 0 && + (error = git_ignore__check_pathspec_for_exact_ignores( + repo, &ps.pathspec, no_fnmatch)) < 0) + goto cleanup; + + if ((error = git_iterator_for_workdir( + &wditer, repo, 0, ps.prefix, ps.prefix)) < 0) + goto cleanup; + + while (!(error = git_iterator_advance(&wd, wditer))) { + + /* check if path actually matches */ + if (!git_pathspec_match_path( + &ps.pathspec, wd->path, no_fnmatch, ignorecase, &match)) + continue; + + /* skip ignored items that are not already in the index */ + if ((flags & GIT_INDEX_ADD_FORCE) == 0 && + git_iterator_current_is_ignored(wditer) && + index_find(&existing, index, wd->path, 0) < 0) + continue; + + /* issue notification callback if requested */ + if (cb && (error = cb(wd->path, match, payload)) != 0) { + if (error > 0) /* return > 0 means skip this one */ + continue; + if (error < 0) { /* return < 0 means abort */ + giterr_clear(); + error = GIT_EUSER; + break; + } + } + + /* TODO: Should we check if the file on disk is already an exact + * match to the file in the index and skip this work if it is? + */ + + /* write the blob to disk and get the oid */ + if ((error = git_blob_create_fromworkdir(&blobid, repo, wd->path)) < 0) + break; + + /* make the new entry to insert */ + if ((entry = index_entry_dup(wd)) == NULL) { + error = -1; + break; + } + entry->oid = blobid; + + /* add working directory item to index */ + if ((error = index_insert(index, entry, 1)) < 0) { + index_entry_free(entry); + break; + } + + git_tree_cache_invalidate_path(index->tree, wd->path); + + /* add implies conflict resolved, move conflict entries to REUC */ + if ((error = index_conflict_to_reuc(index, wd->path)) < 0) { + if (error != GIT_ENOTFOUND) + break; + giterr_clear(); + } + } + + if (error == GIT_ITEROVER) + error = 0; + +cleanup: + git_iterator_free(wditer); + git_pathspec_context_free(&ps); + + return error; +} + +enum { + INDEX_ACTION_NONE = 0, + INDEX_ACTION_UPDATE = 1, + INDEX_ACTION_REMOVE = 2, +}; + +static int index_apply_to_all( + git_index *index, + int action, + const git_strarray *paths, + git_index_matched_path_cb cb, + void *payload) +{ + int error = 0; + size_t i; + git_pathspec_context ps; + const char *match; + + assert(index); + + if ((error = git_pathspec_context_init(&ps, paths)) < 0) + return error; + + git_vector_sort(&index->entries); + + for (i = 0; !error && i < index->entries.length; ++i) { + git_index_entry *entry = git_vector_get(&index->entries, i); + + /* check if path actually matches */ + if (!git_pathspec_match_path( + &ps.pathspec, entry->path, false, index->ignore_case, &match)) + continue; + + /* issue notification callback if requested */ + if (cb && (error = cb(entry->path, match, payload)) != 0) { + if (error > 0) { /* return > 0 means skip this one */ + error = 0; + continue; + } + if (error < 0) { /* return < 0 means abort */ + giterr_clear(); + error = GIT_EUSER; + break; + } + } + + switch (action) { + case INDEX_ACTION_NONE: + break; + case INDEX_ACTION_UPDATE: + error = git_index_add_bypath(index, entry->path); + + if (error == GIT_ENOTFOUND) { + giterr_clear(); + + error = git_index_remove_bypath(index, entry->path); + + if (!error) /* back up foreach if we removed this */ + i--; + } + break; + case INDEX_ACTION_REMOVE: + if (!(error = git_index_remove_bypath(index, entry->path))) + i--; /* back up foreach if we removed this */ + break; + default: + giterr_set(GITERR_INVALID, "Unknown index action %d", action); + error = -1; + break; + } + } + + git_pathspec_context_free(&ps); + + return error; +} + +int git_index_remove_all( + git_index *index, + const git_strarray *pathspec, + git_index_matched_path_cb cb, + void *payload) +{ + return index_apply_to_all( + index, INDEX_ACTION_REMOVE, pathspec, cb, payload); +} + +int git_index_update_all( + git_index *index, + const git_strarray *pathspec, + git_index_matched_path_cb cb, + void *payload) +{ + return index_apply_to_all( + index, INDEX_ACTION_UPDATE, pathspec, cb, payload); +} diff --git a/tests-clar/index/addall.c b/tests-clar/index/addall.c new file mode 100644 index 000000000..33873cb7a --- /dev/null +++ b/tests-clar/index/addall.c @@ -0,0 +1,254 @@ +#include "clar_libgit2.h" +#include "../status/status_helpers.h" +#include "posix.h" + +git_repository *g_repo = NULL; + +void test_index_addall__initialize(void) +{ +} + +void test_index_addall__cleanup(void) +{ + git_repository_free(g_repo); + g_repo = NULL; +} + +#define STATUS_INDEX_FLAGS \ + (GIT_STATUS_INDEX_NEW | GIT_STATUS_INDEX_MODIFIED | \ + GIT_STATUS_INDEX_DELETED | GIT_STATUS_INDEX_RENAMED | \ + GIT_STATUS_INDEX_TYPECHANGE) + +#define STATUS_WT_FLAGS \ + (GIT_STATUS_WT_NEW | GIT_STATUS_WT_MODIFIED | \ + GIT_STATUS_WT_DELETED | GIT_STATUS_WT_TYPECHANGE | \ + GIT_STATUS_WT_RENAMED) + +typedef struct { + size_t index_adds; + size_t index_dels; + size_t index_mods; + size_t wt_adds; + size_t wt_dels; + size_t wt_mods; + size_t ignores; +} index_status_counts; + +static int index_status_cb( + const char *path, unsigned int status_flags, void *payload) +{ + index_status_counts *vals = payload; + + /* cb_status__print(path, status_flags, NULL); */ + + GIT_UNUSED(path); + + if (status_flags & GIT_STATUS_INDEX_NEW) + vals->index_adds++; + if (status_flags & GIT_STATUS_INDEX_MODIFIED) + vals->index_mods++; + if (status_flags & GIT_STATUS_INDEX_DELETED) + vals->index_dels++; + if (status_flags & GIT_STATUS_INDEX_TYPECHANGE) + vals->index_mods++; + + if (status_flags & GIT_STATUS_WT_NEW) + vals->wt_adds++; + if (status_flags & GIT_STATUS_WT_MODIFIED) + vals->wt_mods++; + if (status_flags & GIT_STATUS_WT_DELETED) + vals->wt_dels++; + if (status_flags & GIT_STATUS_WT_TYPECHANGE) + vals->wt_mods++; + + if (status_flags & GIT_STATUS_IGNORED) + vals->ignores++; + + return 0; +} + +static void check_status( + git_repository *repo, + size_t index_adds, size_t index_dels, size_t index_mods, + size_t wt_adds, size_t wt_dels, size_t wt_mods, size_t ignores) +{ + index_status_counts vals; + + memset(&vals, 0, sizeof(vals)); + + cl_git_pass(git_status_foreach(repo, index_status_cb, &vals)); + + cl_assert_equal_sz(index_adds, vals.index_adds); + cl_assert_equal_sz(index_dels, vals.index_dels); + cl_assert_equal_sz(index_mods, vals.index_mods); + cl_assert_equal_sz(wt_adds, vals.wt_adds); + cl_assert_equal_sz(wt_dels, vals.wt_dels); + cl_assert_equal_sz(wt_mods, vals.wt_mods); + cl_assert_equal_sz(ignores, vals.ignores); +} + +static void check_stat_data(git_index *index, const char *path, bool match) +{ + const git_index_entry *entry; + struct stat st; + + cl_must_pass(p_lstat(path, &st)); + + /* skip repo base dir name */ + while (*path != '/') + ++path; + ++path; + + entry = git_index_get_bypath(index, path, 0); + cl_assert(entry); + + if (match) { + cl_assert(st.st_ctime == entry->ctime.seconds); + cl_assert(st.st_mtime == entry->mtime.seconds); + cl_assert(st.st_size == entry->file_size); + cl_assert(st.st_uid == entry->uid); + cl_assert(st.st_gid == entry->gid); + cl_assert_equal_b(st.st_mode & ~0777, entry->mode & ~0777); + cl_assert_equal_b(st.st_mode & 0111, entry->mode & 0111); + } else { + /* most things will still match */ + cl_assert(st.st_size != entry->file_size); + /* would check mtime, but with second resolution it won't work :( */ + } +} + +static void commit_index_to_head( + git_repository *repo, + const char *commit_message) +{ + git_index *index; + git_oid tree_id, commit_id; + git_tree *tree; + git_signature *sig; + git_commit *parent = NULL; + + git_revparse_single((git_object **)&parent, repo, "HEAD"); + /* it is okay if looking up the HEAD fails */ + + cl_git_pass(git_repository_index(&index, repo)); + cl_git_pass(git_index_write_tree(&tree_id, index)); + git_index_free(index); + + cl_git_pass(git_tree_lookup(&tree, repo, &tree_id)); + + cl_git_pass(git_signature_now(&sig, "Testy McTester", "tt@tester.test")); + + cl_git_pass(git_commit_create_v( + &commit_id, repo, "HEAD", sig, sig, + NULL, commit_message, tree, parent ? 1 : 0, parent)); + + git_commit_free(parent); + git_tree_free(tree); + git_signature_free(sig); +} + +void test_index_addall__repo_lifecycle(void) +{ + int error; + git_index *index; + git_strarray paths = { NULL, 0 }; + char *strs[1]; + + cl_git_pass(git_repository_init(&g_repo, "addall", false)); + check_status(g_repo, 0, 0, 0, 0, 0, 0, 0); + + cl_git_pass(git_repository_index(&index, g_repo)); + + cl_git_mkfile("addall/file.foo", "a file"); + check_status(g_repo, 0, 0, 0, 1, 0, 0, 0); + + cl_git_mkfile("addall/.gitignore", "*.foo\n"); + check_status(g_repo, 0, 0, 0, 1, 0, 0, 1); + + cl_git_mkfile("addall/file.bar", "another file"); + check_status(g_repo, 0, 0, 0, 2, 0, 0, 1); + + strs[0] = "file.*"; + paths.strings = strs; + paths.count = 1; + + cl_git_pass(git_index_add_all(index, &paths, 0, NULL, NULL)); + check_stat_data(index, "addall/file.bar", true); + check_status(g_repo, 1, 0, 0, 1, 0, 0, 1); + + cl_git_rewritefile("addall/file.bar", "new content for file"); + check_stat_data(index, "addall/file.bar", false); + check_status(g_repo, 1, 0, 0, 1, 0, 1, 1); + + cl_git_mkfile("addall/file.zzz", "yet another one"); + cl_git_mkfile("addall/other.zzz", "yet another one"); + cl_git_mkfile("addall/more.zzz", "yet another one"); + check_status(g_repo, 1, 0, 0, 4, 0, 1, 1); + + cl_git_pass(git_index_update_all(index, NULL, NULL, NULL)); + check_stat_data(index, "addall/file.bar", true); + check_status(g_repo, 1, 0, 0, 4, 0, 0, 1); + + cl_git_pass(git_index_add_all(index, &paths, 0, NULL, NULL)); + check_stat_data(index, "addall/file.zzz", true); + check_status(g_repo, 2, 0, 0, 3, 0, 0, 1); + + commit_index_to_head(g_repo, "first commit"); + check_status(g_repo, 0, 0, 0, 3, 0, 0, 1); + + /* attempt to add an ignored file - does nothing */ + strs[0] = "file.foo"; + cl_git_pass(git_index_add_all(index, &paths, 0, NULL, NULL)); + check_status(g_repo, 0, 0, 0, 3, 0, 0, 1); + + /* add with check - should generate error */ + error = git_index_add_all( + index, &paths, GIT_INDEX_ADD_CHECK_PATHSPEC, NULL, NULL); + cl_assert_equal_i(GIT_EINVALIDSPEC, error); + check_status(g_repo, 0, 0, 0, 3, 0, 0, 1); + + /* add with force - should allow */ + cl_git_pass(git_index_add_all( + index, &paths, GIT_INDEX_ADD_FORCE, NULL, NULL)); + check_stat_data(index, "addall/file.foo", true); + check_status(g_repo, 1, 0, 0, 3, 0, 0, 0); + + /* now it's in the index, so regular add should work */ + cl_git_rewritefile("addall/file.foo", "new content for file"); + check_stat_data(index, "addall/file.foo", false); + check_status(g_repo, 1, 0, 0, 3, 0, 1, 0); + + cl_git_pass(git_index_add_all(index, &paths, 0, NULL, NULL)); + check_stat_data(index, "addall/file.foo", true); + check_status(g_repo, 1, 0, 0, 3, 0, 0, 0); + + cl_git_pass(git_index_add_bypath(index, "more.zzz")); + check_stat_data(index, "addall/more.zzz", true); + check_status(g_repo, 2, 0, 0, 2, 0, 0, 0); + + cl_git_rewritefile("addall/file.zzz", "new content for file"); + check_status(g_repo, 2, 0, 0, 2, 0, 1, 0); + + cl_git_pass(git_index_add_bypath(index, "file.zzz")); + check_stat_data(index, "addall/file.zzz", true); + check_status(g_repo, 2, 0, 1, 2, 0, 0, 0); + + strs[0] = "*.zzz"; + cl_git_pass(git_index_remove_all(index, &paths, NULL, NULL)); + check_status(g_repo, 1, 1, 0, 4, 0, 0, 0); + + cl_git_pass(git_index_add_bypath(index, "file.zzz")); + check_status(g_repo, 1, 0, 1, 3, 0, 0, 0); + + commit_index_to_head(g_repo, "second commit"); + check_status(g_repo, 0, 0, 0, 3, 0, 0, 0); + + cl_must_pass(p_unlink("addall/file.zzz")); + check_status(g_repo, 0, 0, 0, 3, 1, 0, 0); + + /* update_all should be able to remove entries */ + cl_git_pass(git_index_update_all(index, NULL, NULL, NULL)); + check_status(g_repo, 0, 1, 0, 3, 0, 0, 0); + + git_index_free(index); +}