Module diff

Expand description

Tree and content diffs, rename detection, merge diffs, and related helpers.

Library guide: mod

use grit_lib::repo::{InitOptions, Repository};
let dir = tempfile::tempdir()?;
let _repo = Repository::init(dir.path(), &InitOptions::default())?;
let _ = _repo.diff();

Re-exports§

pub use handle::Diff;
pub use has_changes::text_sides_differ;
pub use hunk_id::hunk_ids;
pub use hunk_id::hunk_ids_with_duplicate_peers;
pub use hunk_id::peer_hunks_by_primary_path;
pub use hunk_id::peer_scope_includes_path;
pub use hunk_id::DiffSide;
pub use hunk_id::HunkId;
pub use hunk_id::HunkIdError;
pub use pickaxe::Pickaxe;
pub use pickaxe::PickaxeError;
pub use pickaxe::PickaxeKind;

Modules§

bloom: bloom API (re-exported at this path).
combined_diff_patch: combined_diff_patch API (re-exported at this path).
combined_tree_diff: combined_tree_diff API (re-exported at this path).
fmt_merge_msg: fmt_merge_msg API (re-exported at this path).
handle: Repository-scoped tree and index diffs via Diff.
hunk_id: Stable identifiers for structured diff hunks (patch selection in later steps).
merge_diff: merge_diff API (re-exported at this path).
merge_file: merge_file API (re-exported at this path).
merge_trees: merge_trees API (re-exported at this path).
patch_ids: patch_ids API (re-exported at this path).
pickaxe: Pickaxe filters (-S / -G) for revision walks and diffs.
rerere: rerere API (re-exported at this path).
structured: Structured line diffs for CLI and embedders (grit diff, grit show).
userdiff: userdiff API (re-exported at this path).
ws: ASCII whitespace helpers used by the diff engine.

Structs§

CommitFilesManyInput: One commit’s identity for Diff::commit_files_many_work, avoiding a second ODB read.
CommitFilesManyOptions: Options for Diff::commit_files_many.
CommitFilesResult: One commit’s diff output in Diff::commit_files_many (same order as input oids).
DetectRenamesOptions: Options for detect_renames_with_policy.
DiffArguments: Resolved diff mode and normalized pathspecs.
DiffEntriesResult: Tree or commit diff entries after rename detection, with an optional limit notice.
DiffEntry: A single diff entry representing one changed path.
DiffFilesOutput: Output of build_file_diffs.
DiffFilter: Typed include/exclude set for DiffStatus using Git’s letter syntax.
DiffIndexToWorktreeOptions: Additional inputs for diff_index_to_worktree_with_options.
DiffOptions: Line-diff options shared by structured output, log patches, and benchmarks.
DiffPostprocessOptions: Options for Git-style rename, copy, and break-rewrite post-processing on tree diffs.
DiffPostprocessWorktree: Index and git directory used to read canonical (clean) worktree bytes during break-rewrite post-processing on tree↔worktree diffs.
DiffRequest: Unparsed grit diff flags and operands, as the user typed them.
DiffStat: Per-file line counts plus totals for a whole diff.
DiffTotals: Summary across all files in a stat-style diff.
DiffTreeToWorktreeExtras: Optional worktree access policy for diff_tree_to_worktree_with_git_dir_and_rules’s slow path.
FileChange: One changed path with both sides’ bytes loaded for diffing.
FileDiff: One changed file with structured hunks and metadata.
FileStat: One file’s row in a DiffStat.
HunkDiffBundle: Staged and unstaged file diffs for hunk id assignment (one staged scan per call).
InexactRenameDetectionSkipped: Emitted when unmatched rename candidates exceed diff.renameLimit after exact and basename pairing.
LineStat: Line insertion/deletion counts for one file (--numstat).
ListedFileDiff: One changed file with optional patch body or line counts.
LogPatchOptions: Options for write_commit_log_patch.
PatchResolvedSide: One file for write_patch_resolved: diff metadata plus where each side’s content is.
PatchTextOptions: Options for write_patch (git diff --binary-compatible output for git apply).
RenameDetectionResult: Result of detect_renames_with_policy.
StatBar: Column counts for one diffstat row’s change bar (see LineStat::bar).
SubmodulePorcelainFlags: Submodule dirty bits aligned with Git’s DIRTY_SUBMODULE_* / porcelain v2 S??? token.
TreeDiffContent: One tree diff entry with both sides loaded for line-level diffing.
TreeDiffStructured: Tree diff entry with structured line hunks ready for CLI rendering.
UncommittedDiff: Staged, unstaged, and working-tree diffs over one loaded index.
UnknownDiffAlgorithm: Unknown value for DiffAlgorithm’s FromStr implementation.
WhitespaceMode: Whitespace handling for line diffs (super::DiffOptions::whitespace).
WorktreeAddRefreshParams: Inputs for classify_worktree_entry_for_add.

Enums§

DiffAlgorithm: Line diff algorithm names matching Git’s diff.algorithm and --diff-algorithm.
DiffDetail: How much content to compute for each changed file.
DiffMerges: How to diff a merge commit’s tree against its parents.
DiffStatus: The kind of change between two sides of a diff.
DiffTarget: What to compare when producing a structured file diff list.
InvalidDiffFilter: Unknown or unsupported --diff-filter letter.
LogPatchBinaryFormat: Binary file encoding inside a log/show patch (git log -p vs git diff --binary).
LogPatchCommitHeader: How much commit metadata precedes each patch in write_commit_log_patch.
NoIndexError: Failure modes specific to diff_paths.
PatchSide: Where write_patch_resolved takes the content of one side of a file.
RenameDetection: Rename detection policy for tree diffs.
ResolveDiffArgumentsError: Invalid grit diff argument combinations or revision parsing failures.
WorktreeAddRefresh: Outcome of comparing a tracked index entry to its worktree path for staging refresh.

Constants§

GIT_DIFF_DEFAULT_BREAK_SCORE: Default break threshold (DEFAULT_BREAK_SCORE in diffcore.h), internal 0–GIT_DIFF_MAX_SCORE scale.
GIT_DIFF_MAX_SCORE: Internal maximum diff score used by Git rename/break heuristics (MAX_SCORE in diffcore.h).
RENAME_DETECTION_DISABLED: u32 rename APIs: thresholds 0–100 run detection at that Git -M percent; RENAME_DETECTION_DISABLED (101) and above skip detection (legacy postprocess and --no-renames plumbing).
WORD_LEVEL_INLINE_MAX_CHANGED_FILES: Changed-file count above which word-level inline emphasis is skipped (human TTY diffs).

Functions§

anchored_unified_diff: Compute a unified diff with anchored lines.
annotate_unified_diff_no_newline_at_eof: Insert Git \ No newline at end of file after -/+ hunk lines when that side of the file lacks a trailing newline on the last line.
apply_diff_filter: Drop entries that do not match filter (Git --diff-filter, after rename detection).
apply_rename_detection_to_diff_entries: Apply Git -M rename detection to a flat diff (index↔tree, tree↔tree, or index↔worktree).
build_file_diffs: Fill per-file diff bodies for tree/index diff entries.
build_file_diffs_streaming: Build patch FileDiff rows one entry at a time without retaining every hunk list.
bytes_for_diff: Classify raw blob bytes for line diffs (NUL in the first 8 KiB ⇒ binary).
classify_worktree_entry_for_add: Decide whether staging must re-read and re-hash a tracked path.
combined_structured_patch_body: Format a CombinedStructuredTextDiff as Git combined patch body bytes.
commit_diff_files: Build per-file diffs for one commit, using combined merge diff when requested.
count_changes_git_histogram: Count insertions and deletions using Git’s histogram line-diff engine.
count_changes_with_algorithm: Count insertions and deletions using the given line-diff algorithm.
count_git_lines: Line count for diffstat/--numstat, matching Git’s count_lines() in diff.c.
detect_copies: Detect copies among diff entries.
detect_copies_with_policy: Like detect_copies, but honors parallel thread policy and rename limits.
detect_renames: Detect renames by pairing Deleted and Added entries with similar content.
detect_renames_with_policy: Like detect_renames, but honors index.threads / --threads via thread_policy.
diff_commits: Diff the trees recorded at two commits (base is the old side, tip the new side).
diff_commits_with_renames: Diff the trees at two commits and pair deleted/added paths as renames when similar enough.
diff_commits_with_renames_and_policy: Like diff_commits_with_renames, but honors index.threads / embedder thread policy.
diff_index_to_tree: Compare the index against a tree (usually HEAD’s tree).
diff_index_to_tree_with_renames: Diff the index against a tree and pair deleted/added paths as renames when similar enough.
diff_index_to_worktree: Compare the index against the working tree.
diff_index_to_worktree_for_staging: Like diff_index_to_worktree_with_options, but loads config from repository_git_dir (the main repository for linked worktrees) and persists each hashed worktree blob into odb so staging can update the index without re-reading file contents. Symlinks store their target bytes. With DiffIndexToWorktreeOptions::hash_blobs_without_write, returns the same blob IDs without writing objects, for staging previews.
diff_index_to_worktree_with_options: Compare tracked index entries with the worktree, using link-target bytes for symlinks.
diff_index_to_worktree_with_renames: Diff the index against the working tree and pair deleted/added paths as renames when similar.
diff_paths: Diff two filesystem paths (files, directories, or one of each).
diff_slice_ops_compacted: Line-diff ops for string slices after Git xdl_change_compact (and optional indent heuristic).
diff_tree_path_change_count: Count path-level differences between two trees without building a DiffEntry list.
diff_tree_to_worktree: Compare a tree against the working tree.
diff_tree_to_worktree_with_git_dir: Like diff_tree_to_worktree, but loads config from repository_git_dir (linked worktrees / gitfiles).
diff_tree_to_worktree_with_git_dir_and_rules: Like diff_tree_to_worktree_with_git_dir, with optional shared crate::worktree_rules::WorktreeRules.
diff_tree_to_worktree_with_git_dir_and_rules_extras: Like diff_tree_to_worktree_with_git_dir_and_rules, with worktree access policy for the slow path.
diff_trees: Compare two trees and return the list of changed entries.
diff_trees_show_tree_entries: Like diff_trees, but emit entries for tree objects themselves in addition to their recursive contents (Git’s diff-tree -t behavior).
diff_trees_with_renames: Diff two trees and pair deleted/added paths as renames when content is similar enough.
diff_trees_with_renames_and_policy: Like diff_trees_with_renames, but honors index.threads / embedder thread policy.
diff_trees_with_renames_and_policy_result: Like diff_trees_with_renames_and_policy, but also returns rename-limit skip metadata.
diffcore_count_changes: Approximate copied vs added material between two blobs (Git diffcore_count_changes).
empty_blob_oid: Return the ObjectId for the empty blob object.
entry_is_racy: Whether an index entry’s cached mtime is at or after the on-disk index file’s mtime.
entry_is_racy_with_policy: Like entry_is_racy, honoring core.usenanosec for sub-second racy detection.
flatten_tree_entries: Depth-first list of (path, mode, oid) blob entries in a tree (for copy detection).
for_each_tree_diff_structured: Visit structured tree diffs in path order without retaining every file’s hunks at once.
format_mode: Format a numeric mode as a zero-padded octal string.
format_raw: Format a diff entry in Git’s raw diff format.
format_raw_abbrev: Format a diff entry with abbreviated OIDs.
hash_worktree_file: Hash normalized worktree bytes for a path (does not write to the object database).
head_path_states: Paths present in HEAD’s tree with mode and blob/commit OID (for status porcelain v2).
hunk_diff_bundle: Load unstaged diffs and the staged peer hunks on the same paths.
indent_heuristic_from_config: diff.indentHeuristic from config (Git defaults to true when unset).
index_mtime_for_diff: Effective index-file mtime for racy-git checks.
line_change_counts: Count insertions and deletions between two text sides using DiffOptions’s algorithm.
line_diff_unified_hunk_body: Unified hunk bodies (@@ …) for two text sides using DiffOptions.
load_object_changes: Load both sides of each entry from the object database.
load_odb_blob_side_for_diff: Load ODB blob bytes for diffing, skipping inflation when over core.bigFileThreshold.
load_worktree_side_for_diff: Load worktree bytes for diffing, optionally skipping reads above core.bigFileThreshold.
materialize_worktree_blob: Read, normalize (CRLF/attributes), and store a worktree blob once.
mode_from_metadata: Derive a Git file mode from filesystem metadata.
name_status_line: Git --name-status letter for a diff entry (uppercase status + rename score suffix).
normalize_ignore_space_change: Normalise text like Git’s -b / --ignore-space-change: on each line, collapse runs of whitespace to a single ASCII space and trim trailing spaces.
normalize_ignore_space_change_line: Normalise one line like Git’s -b / --ignore-space-change (ASCII ws only).
parse_diff_rename_score_token: Parse a single Git parse_rename_score token (50, 50%, decimal forms) into internal 0–GIT_DIFF_MAX_SCORE units.
parse_diff_rename_threshold_percent: Parse -M / --find-renames threshold into Git’s 0–100 similarity percent.
postprocess_diff_entries: Apply break-rewrite, rename, and copy detection to a flat tree or index diff.
read_submodule_head_for_checkout: Read the HEAD commit OID from a submodule checkout directory.
read_submodule_head_oid: Read the HEAD commit OID from a submodule working tree directory.
read_worktree_bytes_for_mode: Read worktree content for diffing, honoring Git tree modes.
read_worktree_side_bytes_for_diff: Read worktree file bytes in Git (repository) form for line diffs against index or tree blobs.
refresh_index_stat_content_verified: Refresh cached stat data for stage-0 file/symlink entries whose worktree content still matches the recorded OID but whose on-disk stat went stale.
refresh_index_stat_content_verified_with_rules: Like refresh_index_stat_content_verified, resolving attributes and conversion through a shared crate::worktree_rules::WorktreeRules when one is supplied (its config snapshot then takes precedence over config).
rename_similarity_score: Compute rename/copy similarity percentage (0–100) between two byte slices.
renames_from_config: diff.renames: false disables; true or copy modes use default similarity.
resolve_commit_diff_pathspecs: Normalize pathspecs for a commit or range diff to repository-relative form.
resolve_diff_arguments: Parse grit diff flags and positionals into a DiffTarget and pathspecs.
resolve_diff_request: Decide what grit diff compares and which paths it is limited to.
resolve_indent_heuristic: Resolve indent heuristic: --no-indent-heuristic and --indent-heuristic override config.
rewrite_dissimilarity_index_percent: Percentage shown in dissimilarity index N% for a rewrite (similarity_index in Git’s diff.c).
rewrite_merge_score: Git merge_score from diffcore-break.c when a pair is considered broken: how much of the source blob was removed (0–DIFF_MAX_SCORE scale). Used for dissimilarity index metadata.
should_break_rewrite_pair: Whether an in-place blob edit should be split into delete+create for rename/copy (should_break in diffcore-break.c). break_score is on the internal 0–GIT_DIFF_MAX_SCORE scale (default DIFF_DEFAULT_BREAK_SCORE).
smudge_racily_clean_entries: Clear racily-clean cache lines on entries whose blob no longer matches the worktree.
sort_entries_by_path: Stable-sort diff entries by DiffEntry::path in raw byte order, Git’s output order for diff lists (non-UTF-8 paths compare by their decoded bytes).
stat_matches: Whether worktree metadata matches cached index stat data (Git ie_match_stat / match_stat_data).
stat_matches_with_policy: Like stat_matches, honoring core.trustctime, core.checkstat, and core.usenanosec.
status_apply_rename_copy_detection: Apply Git-style rename and optional copy detection for index↔worktree diffs.
submodule_commit_subject_line: First line of a commit’s message for git diff --submodule=log output.
submodule_embedded_git_dir: Resolve the embedded git directory for a submodule work tree (sub_dir/.git).
submodule_head_object_broken: True when a populated submodule checkout is broken: its HEAD resolves to a commit OID, but that commit object cannot be read from the submodule’s own object database.
submodule_porcelain_flags: Inspect a checked-out submodule at rel_path (relative to super_worktree) and return flags used for git status --porcelain=v2 submodule tokens.
text_sides_differ: Whether two text sides differ under whitespace options in opts.
tree_diff_contents_parallel: Load old/new blob bytes for each entry, using a worker pool when thresholds allow.
tree_diff_stat_parallel: Per-file line counts for stat modes (loads blobs, no hunk allocation).
tree_diff_structured_entry: Load sides and compute a structured line diff for one tree diff entry.
tree_diff_structured_parallel: Load sides and compute structured line diffs for each entry in parallel.
tree_oid_of_commit: Return the tree OID stored in a commit object.
tree_oid_of_commit_in_repo: Like tree_oid_of_commit, but uses the repository commit-graph cache when available.
unified_diff: Generate a unified diff patch for two blobs.
unified_diff_histogram_hunks_only: Unified diff hunks for Git’s histogram algorithm (no --- / +++ lines).
unified_diff_histogram_ignore_with_prefix_and_funcname: Full unified diff (--- / +++ / hunks) using Git’s histogram algorithm, applying --ignore-blank-lines / -I change-record suppression (histogram_unified_body_ignore_fc) and then attaching function-name hunk headers like unified_diff_histogram_with_prefix_and_funcname_indent.
unified_diff_with_prefix: Same as unified_diff but with configurable source/destination prefixes.
unified_diff_with_prefix_and_funcname: Same as unified_diff_with_prefix with optional custom hunk-header function-name matching.
unified_diff_with_prefix_and_funcname_and_algorithm: Same as unified_diff_with_prefix_and_funcname but allows callers to choose the line diff algorithm used for hunk generation.
worktree_differs_from_index_entry: Quick stat check: does the index entry’s cached stat data match the file? Returns true when the file at ie’s path differs from the index entry (mode or blob).
write_commit_header_and_message: Write commit header (author, date, indented message) after the commit line.
write_commit_log_patch: Write one commit in git log -p --no-color --format=%H shape (no patch for merges).
write_commits_log_patch: Like write_commit_log_patch for many commits (avoids re-reading parent commits).
write_diff_entry_log_patch: Write the diff --git header and unified body for one DiffEntry.
write_patch: Write git diff --binary-style patches for tree diff entries (for git apply).
write_patch_resolved: Write git diff --binary-compatible patch text for files whose sides may live in memory, the object database, or the work tree (index and work-tree diffs).
write_patch_structured: Write patch text from precomputed TreeDiffStructured rows.
write_paths_patch: Write git diff --no-index --binary-style patch bytes for two filesystem paths.
write_unified_diff_with_prefix_and_funcname_and_algorithm: Same as unified_diff_with_prefix_and_funcname_and_algorithm, streaming to out.
zero_oid: Return the zero ObjectId.

Type Aliases§

IndexPathSkipPredicate: Predicate over stored index path bytes for DiffIndexToWorktreeOptions::skip_index_path.

Module diff

Module diff Copy item path

Re-exports§

Modules§

Structs§

Enums§

Constants§

Functions§

Type Aliases§

Module diff