C.E.L_Slide_test2/.orchestrator/tmp/u3_comment.md

Executing unit: u3

[Claude #5] Stage 3 Round #3 — unit u3 implemented.

unit_executed: u3

files_changed:

• src/phase_z2_pipeline.py (+99 / -0) — adds _override_to_grid_tracks immediately above the # Layout preset → zone position 순서 … comment block (i.e., after _build_cols_dynamic and before build_layout_css).

diff_summary:

• New override-path builder _override_to_grid_tracks(preset, zones_data, override_zone_geometries, gap=GRID_GAP) -> dict.
• Body:
  1. Parse preset["css_areas"] via _parse_css_areas to get R × C rows_grid (length R rows × C cols).
  2. Build occupancy list of (zone, occupied_rows, occupied_cols) triples (mirrors the u1 helper layout but operates over override_zone_geometries[pos]["h"|"w"] rather than content_weight).
  3. Per-row aggregation: for each r ∈ [0, R) collect single-row zones (occupied_rows == {r}); fallback to all-span zones (occupied_rows == set(range(R))); take max of their override h. Missing track → 0.0.
  4. Per-col aggregation: same with w on columns.
  5. Degenerate guard: if total_h == 0 or total_w == 0, return _build_grid_dynamic_2d(preset, zones_data, gap=gap) (u2 default fallback per Stage 2 plan).
  6. Normalize per axis (v / total), multiply by avail_h = SLIDE_BODY_HEIGHT - gap*(R-1) / avail_w = SLIDE_BODY_WIDTH - gap*(C-1), then last-element diff absorb (heights_px[-1] += diff_h, widths_px[-1] += diff_w).
  7. Assemble px-string rows / cols and return the same dict shape as _build_grid_dynamic_2d: areas, cols, rows, heights_px, widths_px, ratios, width_ratios, computation, dynamic_rows, dynamic_cols, raw_zone_layout.
• Output flags exactly per Stage 2 contract: computation="user_override_geometry", dynamic_rows=True, dynamic_cols=True.
• raw_zone_layout = {"override_applied": True, "source": override_zone_geometries} — same shape the legacy horizontal-2 / vertical-2 override branches already emit (consistent for step08 trace + test loader excludes the key).
• Dispatcher (build_layout_css) NOT wired — the existing [override-warning] stderr + _build_fr_default fallback for the 5 target presets still fires until u4 lands. The 10 RED fixtures remain RED by design.

evidence (smoke run, not committed):

• git diff --stat src/phase_z2_pipeline.py → 1 file changed, 204 insertions(+) (50 from u1 + 55 from u2 + 99 from u3 — matches expected cumulative delta).
• git diff --name-only src/phase_z2_pipeline.py tests/phase_z2/test_build_layout_css_pr1.py tests/phase_z2/fixtures/build_layout_css → only src/phase_z2_pipeline.py.
• python -c "import ast, pathlib; ast.parse(pathlib.Path('src/phase_z2_pipeline.py').read_text(encoding='utf-8'))" → syntax OK.
• git diff --check src/phase_z2_pipeline.py → no whitespace errors.
• Hand-feed via python -c "from src.phase_z2_pipeline import _override_to_grid_tracks, LAYOUT_PRESETS; ..." across all 5 target presets — byte-for-byte match to tests/phase_z2/fixtures/build_layout_css/*_override.yaml (excluding raw_zone_layout, which the loader strips):
  • top-1-bottom-2: heights=[171, 400] widths=[583, 583] ratios=[0.3, 0.7] width_ratios=[0.5, 0.5] cols='583px 583px' rows='171px 400px'. ✓
  • top-2-bottom-1: heights=[228, 343] widths=[700, 466] ratios=[0.4, 0.6] width_ratios=[0.6, 0.4]. ✓
  • left-1-right-2: heights=[286, 285] widths=[466, 700] ratios=[0.5, 0.5] width_ratios=[0.4, 0.6] (last-element diff absorb on heights row: 286+286 → 286+285 because 2×285.5 rounds to 286+286, diff = -1). ✓
  • left-2-right-1: heights=[343, 228] widths=[408, 758] ratios=[0.6, 0.4] width_ratios=[0.35, 0.65]. ✓
  • grid-2x2: heights=[228, 343] widths=[641, 525] ratios=[0.4, 0.6] width_ratios=[0.55, 0.45]. ✓

scope-lock check:

• Single file: src/phase_z2_pipeline.py. No edits to build_layout_css dispatcher, _build_fr_default, the existing horizontal-2 / vertical-2 override inline branches, _build_rows_dynamic, _build_cols_dynamic, _build_grid_dynamic_2d, _aggregate_zone_signals_per_track, _compute_per_zone_geometry, compute_zone_layout, compute_zone_layout_cols, or PR 1 lock test.
• No fixture added / removed.
• Kei archive path (build_containers_type_b, space_allocator.py) untouched.
• Step08 trace untouched (that lands in u4).
• horizontal-2 / vertical-2 / single regression surface untouched.
• [override-warning] stderr + PR 2 territory comment still in place — u4 removes them as part of dispatcher wiring (Stage 2 plan).

remaining_units:

• u4 — wire dispatchers in build_layout_css (dynamic branch routes 5 presets to _build_grid_dynamic_2d; override branch routes 5 presets to _override_to_grid_tracks; drop [override-warning] stderr; add 2-D-aware step08 trace print).
• u5 — retire PR 1 lock test test_top_1_bottom_2_fr_default_populates_geometry (rename + flip assertions to PR 2 reality).

follow_up_issue_candidates: none (no scope-creep axis discovered in this round; helper sits cleanly in the IMP-09 builders cluster).