blame.py
python
sha256:f9828efc523c2f8ccaee12ae76086a09a9a6a10d6dd20e53f0b13793f0fdcf50
docs: add symlog (#53) and reflog (#54) follow-up issue files
Sonnet 4.6
19 days ago
| 1 | """Core VCS blame — attribute each line of a text file to the commit that last changed it. |
| 2 | |
| 3 | This is the domain-agnostic layer. The MIDI domain has ``note-blame`` |
| 4 | (per-bar attribution); the code domain has ``muse code blame`` (per-symbol |
| 5 | attribution). This module provides line-level blame for *any text file* |
| 6 | tracked in ``state/``, making it useful for any domain that stores text — |
| 7 | configuration files, lyrics, scripts, prose. |
| 8 | |
| 9 | Algorithm |
| 10 | --------- |
| 11 | Walk the commit graph from the requested ref backwards to the root: |
| 12 | |
| 13 | 1. At the starting commit, every line is "owned" by that commit. |
| 14 | 2. At each parent commit, compute the *unified diff* between the parent's |
| 15 | version and the child's version of the file. |
| 16 | 3. Lines that appear in both versions (context/unchanged) are attributed to |
| 17 | the *earliest* commit that produced them; we update the attribution when |
| 18 | we encounter a commit where those lines are *unchanged from the parent* — |
| 19 | i.e. they existed before this commit. |
| 20 | 4. Lines that are *added* by a commit stay attributed to that commit. |
| 21 | |
| 22 | This is equivalent to the ``git blame`` algorithm for single-parent chains. |
| 23 | For merge commits (two parents), we take the parent whose file content most |
| 24 | closely matches the merge result to avoid over-attributing lines to merges. |
| 25 | |
| 26 | Output |
| 27 | ------ |
| 28 | A list of ``BlameLine`` objects, one per line of the file at the requested |
| 29 | ref. |
| 30 | """ |
| 31 | |
| 32 | import difflib |
| 33 | import logging |
| 34 | import pathlib |
| 35 | from dataclasses import dataclass |
| 36 | |
| 37 | from muse.core.graph import iter_ancestors |
| 38 | from muse.core.object_store import read_object |
| 39 | from muse.core.commits import ( |
| 40 | read_commit, |
| 41 | resolve_commit_ref, |
| 42 | ) |
| 43 | from muse.core.snapshots import read_snapshot |
| 44 | |
| 45 | type _CommitCache = dict[str, tuple[str, str, str]] |
| 46 | logger = logging.getLogger(__name__) |
| 47 | |
| 48 | # --------------------------------------------------------------------------- |
| 49 | # Types |
| 50 | # --------------------------------------------------------------------------- |
| 51 | |
| 52 | @dataclass(frozen=True) |
| 53 | class BlameLine: |
| 54 | """Attribution for one line of text.""" |
| 55 | |
| 56 | lineno: int |
| 57 | """1-based line number in the final version of the file.""" |
| 58 | |
| 59 | commit_id: str |
| 60 | """Commit that last changed this line.""" |
| 61 | |
| 62 | author: str |
| 63 | """Author field from the commit record.""" |
| 64 | |
| 65 | committed_at: str |
| 66 | """ISO timestamp of the commit.""" |
| 67 | |
| 68 | message: str |
| 69 | """First line of the commit message.""" |
| 70 | |
| 71 | content: str |
| 72 | """The line content (without trailing newline).""" |
| 73 | |
| 74 | # --------------------------------------------------------------------------- |
| 75 | # File reading helper |
| 76 | # --------------------------------------------------------------------------- |
| 77 | |
| 78 | def _read_file_at_commit( |
| 79 | repo_root: pathlib.Path, |
| 80 | commit_id: str, |
| 81 | rel_path: str, |
| 82 | ) -> list[str] | None: |
| 83 | """Return lines of *rel_path* as it existed at *commit_id*, or None.""" |
| 84 | commit = read_commit(repo_root, commit_id) |
| 85 | if commit is None: |
| 86 | return None |
| 87 | snap = read_snapshot(repo_root, commit.snapshot_id) |
| 88 | if snap is None: |
| 89 | return None |
| 90 | obj_id = snap.manifest.get(rel_path) |
| 91 | if obj_id is None: |
| 92 | return None |
| 93 | try: |
| 94 | data = read_object(repo_root, obj_id) |
| 95 | except OSError: |
| 96 | return None |
| 97 | if data is None: |
| 98 | return None |
| 99 | return data.decode("utf-8", errors="replace").splitlines() |
| 100 | |
| 101 | # --------------------------------------------------------------------------- |
| 102 | # Commit graph walker |
| 103 | # --------------------------------------------------------------------------- |
| 104 | |
| 105 | def _walk_ancestry( |
| 106 | repo_root: pathlib.Path, |
| 107 | start_id: str, |
| 108 | ) -> list[str]: |
| 109 | """Return commit IDs from *start_id* to the root, newest-first.""" |
| 110 | return [c.commit_id for c in iter_ancestors(repo_root, start_id)] |
| 111 | |
| 112 | # --------------------------------------------------------------------------- |
| 113 | # Public API |
| 114 | # --------------------------------------------------------------------------- |
| 115 | |
| 116 | def blame_file( |
| 117 | repo_root: pathlib.Path, |
| 118 | rel_path: str, |
| 119 | commit_id: str, |
| 120 | ) -> list[BlameLine] | None: |
| 121 | """Attribute each line of *rel_path* to the commit that last modified it. |
| 122 | |
| 123 | Args: |
| 124 | repo_root: Repository root. |
| 125 | rel_path: Path relative to ``state/``, e.g. ``"README.md"``. |
| 126 | commit_id: The commit to start the blame from (usually HEAD). |
| 127 | |
| 128 | Returns: |
| 129 | A list of :class:`BlameLine` objects (1-indexed), or ``None`` if the |
| 130 | file does not exist at *commit_id*. |
| 131 | """ |
| 132 | current_lines = _read_file_at_commit(repo_root, commit_id, rel_path) |
| 133 | if current_lines is None: |
| 134 | return None |
| 135 | |
| 136 | n = len(current_lines) |
| 137 | # attribution[i] = commit_id that last changed line i (0-indexed) |
| 138 | attribution: list[str] = [commit_id] * n |
| 139 | |
| 140 | ancestry = _walk_ancestry(repo_root, commit_id) |
| 141 | |
| 142 | def _obj_id_at(cid: str) -> str | None: |
| 143 | """Return the object_id of rel_path at cid, without reading file bytes.""" |
| 144 | c = read_commit(repo_root, cid) |
| 145 | if c is None: |
| 146 | return None |
| 147 | s = read_snapshot(repo_root, c.snapshot_id) |
| 148 | if s is None: |
| 149 | return None |
| 150 | return s.manifest.get(rel_path) |
| 151 | |
| 152 | # Walk from the commit towards the root. When a parent has the same line |
| 153 | # content the attribution moves back to the parent (older is better). |
| 154 | child_lines = current_lines[:] |
| 155 | child_id = commit_id |
| 156 | child_obj_id = _obj_id_at(commit_id) |
| 157 | |
| 158 | for parent_id in ancestry[1:]: # skip the starting commit itself |
| 159 | parent_obj_id = _obj_id_at(parent_id) |
| 160 | |
| 161 | if parent_obj_id is None: |
| 162 | # File didn't exist in this ancestor — stop the walk. |
| 163 | break |
| 164 | |
| 165 | if parent_obj_id == child_obj_id: |
| 166 | # File content identical — all lines move attribution to this parent. |
| 167 | for k in range(n): |
| 168 | if attribution[k] == child_id: |
| 169 | attribution[k] = parent_id |
| 170 | child_id = parent_id |
| 171 | # child_obj_id stays the same (parent_obj_id == child_obj_id) |
| 172 | continue |
| 173 | |
| 174 | parent_lines = _read_file_at_commit(repo_root, parent_id, rel_path) |
| 175 | if parent_lines is None: |
| 176 | break |
| 177 | |
| 178 | # Use SequenceMatcher to align lines. |
| 179 | sm = difflib.SequenceMatcher(None, parent_lines, child_lines, autojunk=False) |
| 180 | for tag, i1, i2, j1, j2 in sm.get_opcodes(): |
| 181 | if tag == "equal": |
| 182 | # These lines are unchanged from parent → child. |
| 183 | # They may be attributed to an even older commit; update those |
| 184 | # that are currently attributed to child_id. |
| 185 | # j indices reference child_lines which may be larger than the |
| 186 | # original current_lines — skip out-of-bounds positions. |
| 187 | for k in range(j2 - j1): |
| 188 | idx = j1 + k |
| 189 | if idx < len(attribution) and attribution[idx] == child_id: |
| 190 | attribution[idx] = parent_id |
| 191 | |
| 192 | child_lines = parent_lines |
| 193 | child_id = parent_id |
| 194 | child_obj_id = parent_obj_id |
| 195 | |
| 196 | # Build the final BlameLine list. |
| 197 | result: list[BlameLine] = [] |
| 198 | commit_cache: _CommitCache = {} |
| 199 | |
| 200 | def _commit_meta(cid: str) -> tuple[str, str, str]: |
| 201 | if cid in commit_cache: |
| 202 | return commit_cache[cid] |
| 203 | c = read_commit(repo_root, cid) |
| 204 | if c is None: |
| 205 | meta = ("unknown", "", "") |
| 206 | else: |
| 207 | first_line = c.message.split("\n", 1)[0] if c.message else "" |
| 208 | meta = (c.author or "unknown", c.committed_at.isoformat(), first_line) |
| 209 | commit_cache[cid] = meta |
| 210 | return meta |
| 211 | |
| 212 | for idx, (line, cid) in enumerate(zip(current_lines, attribution)): |
| 213 | author, committed_at, message = _commit_meta(cid) |
| 214 | result.append(BlameLine( |
| 215 | lineno=idx + 1, |
| 216 | commit_id=cid, |
| 217 | author=author, |
| 218 | committed_at=committed_at, |
| 219 | message=message, |
| 220 | content=line, |
| 221 | )) |
| 222 | |
| 223 | return result |
File History
1 commit
sha256:f9828efc523c2f8ccaee12ae76086a09a9a6a10d6dd20e53f0b13793f0fdcf50
docs: add symlog (#53) and reflog (#54) follow-up issue files
Sonnet 4.6
19 days ago