1. 18 Apr, 2020 4 commits
    • Volker Krause's avatar
      Look for stations slightly outside of terminal bounds as well · 412cfbea
      Volker Krause authored
      This fixes HAM, and should fix GDN/FRA/LIS too eventually, once the issues
      exposed there now get addressed.
    • Volker Krause's avatar
      Also consider stations inside terminal buildings outside the airport bounds · b4911486
      Volker Krause authored
      This can happen when terminal buildings overlap the airport bounds. Fixes
      CPH for example.
    • Volker Krause's avatar
      Check the precise boundary for including terminals, not just the bbox · 91b4705e
      Volker Krause authored
      This removes terminal buildings from adjacent facilities from the
      consideration, such as the Airbus site next to TLS, or the government
      terminal next to TXL.
    • Volker Krause's avatar
      Initial experiments to improve airport coordinates based on OSM data · 4b5489d0
      Volker Krause authored
      So far the coordinates we use for navigating to/from airports come from
      Wikidata, and are typically somewhere on the center of the whole airport
      area, typically a runway. That's far from ideal for navigation, as that's
      neither a location you want to or event can get to, and in extreme cases
      this even leads to navigation "snapping" to the opposite side of the
      airport entirely.
      Instead we want a coordinate around somewhere around the entrance. So far
      there's three types of information from OSM we consider here:
      * terminal buildings (or rather the center point of their bounding box)
      * entrance nodes on terminal building polygons (unfortunately not reliably
      available in the input data).
      * railway stations on the premise of the airport.
      This improves the result for many smaller or mid-sized airports
      considerably already. What this cannot improve is the situation
      at large airports with widely spread terminals (LHR, CDG, MXP, etc).
      Those however cannot meaningfully represented by a single coordinate
      Code isn't hooked up yet, this is just for local experiments at this