The continuous problem scanner detects turns in two gray zone ranges where Best Continuation (BC) matching fails, making routing instruction classifications unreliable. When BC matching is ambiguous (bc_count ≠ 1) within these zones, angles are flagged as PROBLEM. All flagged turns display with the "Angle to avoid" color.
Updated: Bidirectional pair evaluation now ensures all segment combinations are checked (both A→B and B→A), fixing the issue where PROBLEM angles were missed due to segment ordering.
Angle Range: 44° - 47° (TURN_ANGLE ± GRAY_ZONE = 45.5° ± 1.5°)
Routing Types: KEEP_LEFT, KEEP_RIGHT, TURN_LEFT, TURN_RIGHT (ambiguous boundary)
Issue: Best Continuation matching is ambiguous — the algorithm cannot determine a clear "straight" path, making KEEP vs TURN classification unpredictable
- Test: Find junction with 44.5° angle
- Verify: Marker displays as orange/red (PROBLEM type)
- Verify: Angle label shows as 44.5° (or your decimal setting)
- Verify: Continuous scan AND on-demand mode both detect it
- Verify: Marker placed at correct bearing from node
Angle Range: 166° - 170° (U_TURN_ANGLE ± GRAY_ZONE = 168.24° ± 1.5°)
Routing Types: PROBLEM (explicitly flagged by ja_classify_turn_angle)
Issue: Angle near 180° but unclear if Waze will route it as U-turn or regular turn; affected by median length and lane guidance
- Test: Find junction with 169° angle
- Verify: Marker displays as orange/red (PROBLEM type)
- Verify: Angle label shows as 169° (or your decimal setting)
- Verify: Continuous scan AND on-demand mode both detect it
- Verify: Marker placed at correct bearing from node
- Test: Pan away and back to same area - markers should appear instantly (cache hit)
- Verify: Debug log shows "PHASE 1 complete" with cache hits
- Test: Edit a segment at a cached node - verify cache is invalidated
- Verify: Debug log shows cache recalculation after edit with "Cleared layer" message
- Test: Zoom in/out at same location - markers should persist (cache still valid)
- Verify: Bidirectional pair evaluation finds all PROBLEM angles regardless of segment order
- Test: Pan across area with 20+ problem nodes - no lag or stuttering
- Test: High zoom (22) - markers visible and well-positioned
- Test: Medium zoom (17-20) - markers visible with appropriate offset
- Test: Low zoom (13-16) - markers still visible (zoom filtering by viewport)
- Test: Toggle "Highlight problem angles" checkbox - layer appears/disappears
- Test: 4-way junction with problem angle - only ONE marker drawn (not 2)
- Test: 6-way junction with multiple problems - correct count (not doubled)
- Verify: On-demand JAI mode doesn't duplicate either (independent check)
- Test: Problem markers use user-configured "Angle to avoid" color
- Test: Marker size consistent across viewport (no size variation from collision)
- Test: Marker z-index correct (appears above base map, below selection layer)
- Test: Layer toggle in Layer Switcher works
- Verify: Layer name is "problem_angles_continuous"
- Test: Node with only 2 segments - no markers (skipped correctly)
- Test: Node with 10+ segments - handles correctly without performance drop
- Test: Roundabout nodes - markers appear correctly
- Test: Junction Box nodes - markers appear correctly
- Test: Private Road junctions - markers appear if road type included in settings
- [ ] USER-SETTINGS.md updated with "Highlight problem angles" description
- [ ] Explain what each gray zone means
- [ ] Explain why these angles are problematic
- [ ] Provide examples of how to fix each type
- All 2 gray zones have been located and verified in actual map data (44-47° and 166-170°)
- Continuous scanner correctly identifies same problems as on-demand JAI
- Performance is acceptable at all zoom levels
- Markers display correctly with user's color preferences
- Documentation is complete and accurate