Label placement system: remaining work tracker

[madsCodeBuddy]

Status: On Hold

An automatic label collision resolver was prototyped but is not ready to merge.

What was built

• Occupancy grid — rasterizes ellipse/line/point glyphs into pixel-space grid
• Upper-left preferred placement — initial label positions biased upper-left of anchor
• Glyph-aware force placement — occupancy gradient descent pushes labels away from glyphs
• Leader lines — dotted lines connecting displaced labels to anchors
• resolve_label_overlaps() — force-directed solver with repulsion, attraction springs, occupancy forces

Remaining problems

1. Labels still overlap glyphs — solver reduces label-label overlap but labels land on ellipses
2. Solver doesn't converge — 6 overlapping pairs remain after 200 iterations on the 15-label Boyd dataset
3. Occupancy grid is slow — enabling glyph_data causes timeout on moderate datasets
4. Zero leader lines triggered — distance threshold too generous, displaced labels don't cross it
5. Integration not started — wiring into add_predefined_processes / add_processes blocked by Pandas .align() name clash and np.log10(int) TypeError

Alternative approaches to consider

• Simulated annealing (better at escaping local minima)
• Greedy placement with backtracking
• Constraint-based optimization
• Manual offset overrides in CSV (already partially supported via x_offset/y_offset columns)

Migrated from timeSpace-private #234