Fix: HexagonDrawable 4-arg constructor renders invisible hexagons

[jhotadhari]

Closes #1300

The bug

HexagonDrawable has two constructors that build geometry in different coordinate spaces, causing the 4-argument version to render invisible hexagons.

The 2-arg constructor correctly uses geographic lat/lng when building the JTS polygon via GeomBuilder:

// 2-arg constructor (works correctly)
GeoPoint point = findGeoPointWithGivenDistance(center, i * Math.PI / 3, radiusKm);
gb.points(point.getLongitude(), point.getLatitude());  // geographic degrees

The 4-arg constructor was projecting through Mercator first and feeding screen-space x/y into GeomBuilder as if they were lat/lng:

// 4-arg constructor (broken — before the fix)
GeoPoint point = findGeoPointWithGivenDistance(center, rotationRad + i * Math.PI / 3, radiusKm);
MercatorProjection.project(point, tmp);
gb.points(tmp.x, tmp.y);  // Mercator pixels, NOT geographic degrees!

Since VectorLayer later interprets whatever it receives as geographic coordinates, the Mercator-projected hexagon lands at a completely wrong map position and is effectively invisible.

This was the only usage of MercatorProjection in the entire geometries package — other style-accepting drawables (CircleDrawable, RectangleDrawable, PolygonDrawable) all pass raw lon/lat directly.

The fix

Remove the MercatorProjection.project() call and Point usage from the 4-arg constructor. Instead, pass geographic coordinates directly to GeomBuilder, matching the 2-arg constructor behavior:

// 4-arg constructor (fixed)
GeoPoint point = findGeoPointWithGivenDistance(center, rotationRad + i * Math.PI / 3, radiusKm);
gb.points(point.getLongitude(), point.getLatitude());  // geographic degrees

Also removed the now-unused MercatorProjection and Point imports.

Proof

HexagonDrawableTest renders 4 hexagons at distinct European locations and logs every vertex coordinate. All vertices are valid geographic degrees, confirming the 4-arg constructor now correctly places hexagons at the intended map positions:

--- 2-arg constructor (BLUE) — center: [lat=52.0,lon=5.0] ---
  Hexagon is built from 6 geographic coordinates:
    vertex 0: lon=5.000000 lat=56.496600 — OK: YES
    vertex 1: lon=11.644154 lat=54.068501 — OK: YES
    vertex 2: lon=11.012326 lat=49.592311 — OK: YES
    vertex 3: lon=5.000000 lat=47.503399 — OK: YES
    vertex 4: lon=-1.012326 lat=49.592311 — OK: YES
    vertex 5: lon=-1.644154 lat=54.068501 — OK: YES
--- 4-arg constructor 0° rotation (GREEN) — center: [lat=52.0,lon=12.0] ---
  Hexagon is built from 6 geographic coordinates:
    vertex 0: lon=12.000000 lat=56.496600 — OK: YES
    vertex 1: lon=18.644154 lat=54.068501 — OK: YES
    vertex 2: lon=18.012326 lat=49.592311 — OK: YES
    vertex 3: lon=12.000000 lat=47.503399 — OK: YES
    vertex 4: lon=5.987673 lat=49.592311 — OK: YES
    vertex 5: lon=5.355845 lat=54.068501 — OK: YES
--- 4-arg constructor 30° rotation (RED) — center: [lat=46.0,lon=8.0] ---
  Hexagon is built from 6 geographic coordinates:
    vertex 0: lon=11.484998 lat=49.843906 — OK: YES
    vertex 1: lon=14.458934 lat=45.817676 — OK: YES
    vertex 2: lon=11.026736 lat=42.064046 — OK: YES
    vertex 3: lon=4.973263 lat=42.064046 — OK: YES
    vertex 4: lon=1.541065 lat=45.817676 — OK: YES
    vertex 5: lon=4.515001 lat=49.843906 — OK: YES
--- 4-arg constructor 60° rotation, 250km (MAGENTA) — center: [lat=42.0,lon=12.0] ---
  Hexagon is built from 6 geographic coordinates:
    vertex 0: lon=14.666647 lat=43.093611 — OK: YES
    vertex 1: lon=14.574145 lat=40.846796 — OK: YES
    vertex 2: lon=12.000000 lat=39.751699 — OK: YES
    vertex 3: lon=9.425854 lat=40.846796 — OK: YES
    vertex 4: lon=9.333352 lat=43.093611 — OK: YES
    vertex 5: lon=12.000000 lat=44.248300 — OK: YES

Before the fix, the 4-arg constructor would have produced Mercator-projected x/y values (on the order of millions) fed into GeomBuilder as if they were geographic degrees — making those same hexagons land nowhere near the map viewport (effectively invisible).

Side note: running the playground test

The playground build.gradle needed two temporary tweaks to run the test on Ubuntu 22.04 — neither is part of this PR:

1. LWJGL version: Swapped implementation project(':vtm-desktop-lwjgl') → ':vtm-desktop-lwjgl3'. LWJGL 2's native .so loading hits a glibc ld.so assertion on Ubuntu 22.04+. LWJGL 3 has no such issue.
2. Main class wiring: Added mainClass = project.findProperty('mainClassName') so the application plugin picks up the main class from -PmainClassName= on the command line.

Run with:

./gradlew :vtm-playground:run -PmainClassName=org.oscim.test.HexagonDrawableTest

🤖 Generated with Claude Code

[devemux86]

LWJGL version: Swapped implementation project(':vtm-desktop-lwjgl') → ':vtm-desktop-lwjgl3'. LWJGL 2's native .so loading hits a glibc ld.so assertion on Ubuntu 22.04+. LWJGL 3 has no such issue.

Testing on Debian with Intel GPU, it works correctly.

Perhaps something with your graphics drivers?
Anyway developers / users can modify the project to work better on their desktops.

[jhotadhari]

Perhaps something with your graphics drivers?

could be. changing to vtm-desktop-lwjgl3 locally for testing for me is alright.
and o android (where I want to use the library) I dont run into this issue.

Keep in mind that the the issue generation, the PR, the fix, all description and report ... is all generated by claude and I have no idea about java.

This described fix described above appears to me that it might break some apps, that already implemented a workaround for the hexagon. Im not sure if it is good to roll out this fix

[devemux86]

@jhotadhari Thanks for the fix!