A city that owns its state and its scene, and a loading overlay that waits for the whole world

api/models/manifest.py

@@ -194,17 +194,6 @@ class CommitLeader(BaseModel):
     date: str
 
 
-# Both fields required-nullable: the scanner always emits them (null for a repo
-# with no commits), so they're present-but-nullable on the wire, not optional.
-class CommitDateRange(BaseModel):
-    oldest: Optional[str] = Field(
-        description="Oldest commit date (YYYY-MM-DD), or null when the repo has no commits"
-    )
-    newest: Optional[str] = Field(
-        description="Newest commit date (YYYY-MM-DD), or null when the repo has no commits"
-    )
-
-
 class DayLeader(BaseModel):
     date: str
     count: int
@@ -252,7 +241,6 @@ class RepoStats(BaseModel):
     commitCount: int = Field(
         description="Commits in the full history, however many `commits` carries"
     )
-    commitDates: CommitDateRange
     maxCommitsPerDay: Optional[DayLeader]
     maxCommitStreakDays: int
     authors: list[AuthorStat]
@@ -306,13 +294,6 @@ class SignatureResponse(BaseModel):
     content_signature: str
 
 
-class DateRangeMs(BaseModel):
-    minCreated: int
-    maxCreated: int
-    minModified: int
-    maxModified: int
-
-
 class TimelineChange(BaseModel):
     path: str
     sha: Optional[str] = Field(description="New blob sha, or null when deleted")
@@ -340,7 +321,6 @@ class TimelineBundle(BaseModel):
     blobLines: dict[str, Optional[int]]
     blobSizes: dict[str, Optional[int]]
     commitLineRanges: list[RangeStat]
-    commitDateRanges: list[DateRangeMs]
     notes: list[str]
 
 

api/scan/stats.py

@@ -11,7 +11,6 @@
 from api.models.manifest import (
     AuthorStat,
     BusynessThresholds,
-    CommitDateRange,
     CommitEntry,
     CommitLeader,
     DayLeader,
@@ -242,7 +241,6 @@ def compute_repo_stats(tree: DirNode, commits: list[CommitEntry]) -> RepoStats:
                 newest_dir = d
 
     grandest = sparsest = None  # commits, by files changed
-    oldest_commit = newest_commit = None  # commit-date range
     # The walk yields oldest first, so the ends ARE the leaders. Comparing dates
     # instead would tie on a busy day and keep the first, which is not newest.
     first = commits[0] if commits else None
@@ -259,10 +257,6 @@ def compute_repo_stats(tree: DirNode, commits: list[CommitEntry]) -> RepoStats:
         # multi-counting. 0 until manifest-wrap bakes it; read that as 1.
         d = day_of(c.date)
         day_totals[d] = max(day_totals.get(d, 0), c.same_day_total or 1)
-        if oldest_commit is None or d < oldest_commit:
-            oldest_commit = d
-        if newest_commit is None or d > newest_commit:
-            newest_commit = d
 
     busiest_day: DayLeader | None = None
     if day_totals:
@@ -320,7 +314,6 @@ def _commit_leader(c: Optional[CommitEntry]) -> CommitLeader | None:
         newestCommit=_commit_leader(last),
         commitCount=len(commits),
         # YYYY-MM-DD sorts lexically in chronological order, so min/max are exact.
-        commitDates=CommitDateRange(oldest=oldest_commit, newest=newest_commit),
         maxCommitsPerDay=busiest_day,
         maxCommitStreakDays=_longest_streak(list(day_totals.keys())),
         authors=authors,

api/scan/timeline.py

@@ -20,7 +20,6 @@
 from api.core.constants import TimelineStage
 from api.models.manifest import (
     CommitEntry,
-    DateRangeMs,
     FileNode,
     Manifest,
     RangeStat,
@@ -31,7 +30,6 @@
 from api.scan.blobmeta import MISSING_BLOB, blob_file_node, resolve_blob_stats
 from api.scan.manifest import wrap_manifest
 from api.core.progress import Throttle, log
-from api.utils.dates import iso_to_ms
 from api.core.exceptions import NotAGitRepoError
 from api.scan.signatures import derive_tree_signals, new_signature
 from api.scan.skiprules import SkipRules
@@ -45,9 +43,9 @@
 
 _HISTORY_HEARTBEAT_EVERY = 2000  # commits between progress ticks
 
-# Four steps, a quarter of the percent each. The last is the router's:
+# Three steps, a third of the percent each. The last is the router's:
 # serialising the bundle is the step nothing else can report.
-ASSEMBLE_STEPS = 4
+ASSEMBLE_STEPS = 3
 # Commits between ticks inside a step that reports its own progress.
 _ASSEMBLE_HEARTBEAT_EVERY = 200
 
@@ -290,81 +288,6 @@ def compute_commit_line_ranges(
     return ranges
 
 
-def compute_commit_date_ranges(
-    deltas: list[CommitDelta],
-    commits: list[CommitEntry],
-    git_created: dict[str, str],
-    git_modified: dict[str, str],
-    *,
-    on_commit: Callable[[int, int], None] | None = None,
-) -> list[DateRangeMs]:
-    """Per-commit created/modified ms ranges over the files present at each
-    commit; range[HEAD] equals the live manifest's dateRanges (weathering
-    normalizes against these). replay.ts walks the same deltas for a different
-    output (per-frame scrub index) — neither is a copy of the other."""
-    commit_ms = [iso_to_ms(c.date) or 0 for c in commits]
-    # Parsed once per path, not per (commit, path): the loop below runs ~98M
-    # times on a big repo, and re-parsing there is what made this the slow step.
-    created_ms = {p: iso_to_ms(v) for p, v in git_created.items()}
-    modified_ms = {p: iso_to_ms(v) for p, v in git_modified.items()}
-
-    final_idx: dict[str, int] = {}
-    genesis_idx: dict[str, int] = {}
-    for i, delta in enumerate(deltas):
-        for path, sha in delta.changes:
-            final_idx[path] = i
-            if sha is not None and path not in genesis_idx:
-                genesis_idx[path] = i
-
-    present: set[str] = set()
-    last_change: dict[str, int] = {}
-    ranges: list[DateRangeMs] = []
-    for i, delta in enumerate(deltas):
-        for path, sha in delta.changes:
-            last_change[path] = i
-            if sha is None:
-                present.discard(path)
-            else:
-                present.add(path)
-
-        min_created = min_modified = None
-        max_created = max_modified = None
-        for path in present:
-            lm = last_change.get(path, 0)
-            modified = None
-            if lm >= final_idx.get(path, 0):
-                modified = modified_ms.get(path)
-            if modified is None:
-                modified = commit_ms[lm] if lm < len(commit_ms) else 0
-            created = created_ms.get(path)
-            if created is None:
-                gi = genesis_idx.get(path, 0)
-                created = commit_ms[gi] if gi < len(commit_ms) else 0
-
-            if min_created is None or created < min_created:
-                min_created = created
-            if max_created is None or created > max_created:
-                max_created = created
-            if min_modified is None or modified < min_modified:
-                min_modified = modified
-            if max_modified is None or modified > max_modified:
-                max_modified = modified
-
-        # An empty present set collapses to a zero span, which the client already
-        # treats as "no spread" (freshest / newest).
-        ranges.append(
-            DateRangeMs(
-                minCreated=min_created or 0,
-                maxCreated=max_created or 0,
-                minModified=min_modified or 0,
-                maxModified=max_modified or 0,
-            )
-        )
-        if on_commit is not None and i % _ASSEMBLE_HEARTBEAT_EVERY == 0:
-            on_commit(i, len(deltas))
-    return ranges
-
-
 def build_timeline_bundle(
     root: str,
     source: SourceRef,
@@ -450,16 +373,6 @@ def build_timeline_bundle(
     )
     assemble_tick(on_progress, 2)  # per-commit line ranges
     commit_line_ranges = compute_commit_line_ranges(deltas, blob_lines)
-    # The long one: it walks every file present at every commit, so it reports
-    # from inside rather than sitting on its own step for minutes.
-    assemble_tick(on_progress, 3)
-    commit_date_ranges = compute_commit_date_ranges(
-        deltas,
-        commits,
-        history.created,
-        history.modified,
-        on_commit=lambda i, total: assemble_tick(on_progress, 3, i / max(total, 1)),
-    )
     log("timeline bundle complete")
 
     return TimelineBundle(
@@ -475,6 +388,5 @@ def build_timeline_bundle(
         blobLines=blob_lines,
         blobSizes=blob_sizes,
         commitLineRanges=commit_line_ranges,
-        commitDateRanges=commit_date_ranges,
         notes=notes,
     )

api/tests/conftest.py

@@ -545,7 +545,6 @@ def make_manifest(src: str = "/repo", **overrides: Any) -> Manifest:
             "dirtyFileCount": 0,
             "codeBytes": 0,
             "commitCount": 0,
-            "commitDates": {"oldest": None, "newest": None},
             "maxCommitStreakDays": 0,
             "authors": [],
         },
@@ -565,7 +564,6 @@ def make_timeline_bundle(**overrides: Any) -> TimelineBundle:
         "blobLines": {},
         "blobSizes": {},
         "commitLineRanges": [],
-        "commitDateRanges": [],
         "notes": [],
     }
     return TimelineBundle.model_validate({**fields, **overrides})

api/tests/scan/test_stats.py

@@ -3,7 +3,6 @@
 from api.models.manifest import (
     AuthorStat,
     BusynessThresholds,
-    CommitDateRange,
     CommitLeader,
     RangeStat,
 )
@@ -178,7 +177,6 @@ def test_commit_leaders_authors_and_streak():
     assert s.newestCommit.sha == "ddd"
     assert s.newestCommit.date == "2022-02-10"
     assert s.minFilesPerCommit == CommitLeader(sha="ccc", files=1, date="2022-01-03")
-    assert s.commitDates == CommitDateRange(oldest="2022-01-01", newest="2022-02-10")
     assert s.maxCommitsPerDay.count == 2
     assert s.maxCommitStreakDays == 3
     assert s.authors[0] == AuthorStat(name="Ada", commits=3, hue=_author_hue("Ada"))
@@ -203,8 +201,6 @@ def test_newest_commit_is_the_last_one_that_day():
 
     assert s.newestCommit.sha == "ccc"
     assert s.oldestCommit.sha == "aaa"
-    # The day range is still a day, whatever the timestamps carry.
-    assert s.commitDates == CommitDateRange(oldest="2026-08-15", newest="2026-08-15")
 
 
 def test_empty_tree_and_no_commits():
@@ -213,7 +209,6 @@ def test_empty_tree_and_no_commits():
     assert s.maxFilesPerCommit is None
     assert s.maxCommitStreakDays == 0
     assert s.authors == []
-    assert s.commitDates == CommitDateRange(oldest=None, newest=None)
     assert s.lineCountRange == RangeStat(min=0, max=0)
     assert s.mediaCount == 0
 

api/tests/scan/test_timeline.py

@@ -4,7 +4,7 @@
 import os
 import subprocess
 from pathlib import Path
-from api.models.manifest import DateRangeMs, RangeStat
+from api.models.manifest import RangeStat
 
 from api.scan.timeline import walk_deltas
 from api.git import SourceRef
@@ -350,69 +350,6 @@ def test_head_line_range_matches_live_scan(tmp_path: Path) -> None:
     assert ranges[-1] == live_range  # Timeline HEAD == Live
 
 
-def test_head_date_range_matches_live_scan(tmp_path: Path) -> None:
-    """The date analog of the line-range contract: commitDateRanges[-1] (HEAD)
-    equals the live scan's dateRanges for the same repo, so weathering at HEAD
-    normalizes identically. Compared as epoch ms, which is what the client
-    consumes."""
-    from api.scan.timeline import compute_commit_date_ranges, walk_deltas
-    from api.utils.dates import iso_to_ms
-    from api.git.meta import collect_git_history
-    from api.scan.scanner import scan_tree
-
-    _init(tmp_path)
-    (tmp_path / "a.txt").write_text("one\n")
-    _commit(tmp_path, "c1")
-    (tmp_path / "b.txt").write_text("two\n")
-    _commit(tmp_path, "c2")
-    (tmp_path / "a.txt").write_text("one\nmore\n")
-    _commit(tmp_path, "c3")
-
-    live = None
-    for ev in scan_tree(str(tmp_path), SourceRef(str(tmp_path)), use_cache=False):
-        if ev.phase == "manifest-complete":
-            live = ev.manifest
-    assert live is not None
-    live_dates = live.dateRanges
-
-    deltas = walk_deltas(tmp_path)
-    git_created, git_modified, commits = collect_git_history(tmp_path, use_cache=False)
-    ranges = compute_commit_date_ranges(deltas, commits, git_created, git_modified)
-
-    assert ranges[-1] == DateRangeMs(
-        minCreated=iso_to_ms(live_dates.minCreated),
-        maxCreated=iso_to_ms(live_dates.maxCreated),
-        minModified=iso_to_ms(live_dates.minModified),
-        maxModified=iso_to_ms(live_dates.maxModified),
-    )
-    # One range per commit, and the created span never runs backwards.
-    assert len(ranges) == len(deltas)
-    assert all(r.minCreated <= r.maxCreated for r in ranges)
-
-
-def test_commit_date_ranges_track_the_present_set(tmp_path: Path) -> None:
-    """A deleted file drops out of the range at the commit that removed it."""
-    from api.scan.timeline import compute_commit_date_ranges, walk_deltas
-    from api.git.meta import collect_git_history
-
-    _init(tmp_path)
-    (tmp_path / "old.txt").write_text("x\n")
-    _commit(tmp_path, "c1")
-    (tmp_path / "new.txt").write_text("y\n")
-    _commit(tmp_path, "c2")
-    (tmp_path / "old.txt").unlink()
-    _commit(tmp_path, "c3")
-
-    deltas = walk_deltas(tmp_path)
-    git_created, git_modified, commits = collect_git_history(tmp_path, use_cache=False)
-    ranges = compute_commit_date_ranges(deltas, commits, git_created, git_modified)
-
-    # At c1 only old.txt exists, so the span is a single instant.
-    assert ranges[0].minCreated == ranges[0].maxCreated
-    # At HEAD only new.txt survives, so its creation is both ends again.
-    assert ranges[-1].minCreated == ranges[-1].maxCreated
-
-
 def test_bundle_replay_matches_reconstruct(tmp_path: Path) -> None:
     """Replaying deltas[0..i] reproduces reconstruct_manifest's file set + lines
     at that ref — ties the bundle to the proven phase-1 reconstruction. The

api/tests/test_models.py

@@ -148,7 +148,6 @@ def test_manifest_serializes(self) -> None:
                 "oldestCommit": None,
                 "newestCommit": None,
                 "commitCount": 0,
-                "commitDates": {"oldest": None, "newest": None},
                 "maxCommitsPerDay": None,
                 "maxCommitStreakDays": 0,
                 "authors": [],