Tasks

• Benchmark heatmap algorithm with 10 / 100 / 1000 shadings
• Profile heatmap SQL function with plprofiler
• Add and benchmark variations
  • v1
  • v2
• add README.md

Implemented:

• v1: Fetch plant relations once
  • => Big improvement for plant relationship score calculation
• v2: Inline everything (because PL/pgSQL functions are "optimization fences")
  • little or negative effect for larger maps; some improvements for smaller maps
  • terrible maintainability

Other ideas:

• Hard cutoff of effect after some distance (needs discussion)
  • Effect of plant relationships decreases quadratically
• Return bad values early (needs discussion)
  • e.g. really bad shading cannot be improved with good companions and vice versa
  • may become even more important later on, when more layers are added to the heatmap