fix(example): updated marker stress test to use more stressful markers

[ReinisSprogis]

This better reflects real life usage and looks better than rectangles. Updated to match single marker demo.

[ReinisSprogis]

As a note. In this edited demo I noticed that when all markers are visible, preformance is ok, but as soon as I pan and some markers are outside the screen performance degrades significantly. Should be opposit, that performance improves. as less markers are on screen.

2025-07-08.22-00-37.mp4

[josxha]

many_markers.dart seems to have a formatting issue. Try to run dart format .

[JaffaKetchup]

If this demo has a performance issue like @ReinisSprogis mentioned, then that needs to be fixed before merge. I'll test it out.

[ReinisSprogis]

@josxha Ah. Yes. I have a custom formatting rule that allows for longer lines before brake it. dart fomat . did original formatting. And thanks for seed tip on Random, thats useful to know.

[ReinisSprogis]

@JaffaKetchupI don't think it's related to the demo itself. I tested the original demo by simply swapping ColoredBox with Icon and observed the same result. ColoredBox might have been too primitive to reveal the issue. Replacing it with any slightly more complex widget causes noticeable lag.

I'm just speculating here, but maybe when all markers use the same widget, same color, size, and structure. Flutter applies some optimization under the hood, recognizing they're identical. As you can see, there's a clear difference between using colored markers and markers with a single color.

Still, it's unusual that removing markers would degrade performance, unless there's inefficient filtering happening under the hood.

 Marker(
     point: LatLng(doubleInRange(r, 37, 55), doubleInRange(r, -9, 30)),
     height: 12,
     width: 12,
     child: Icon(
         Icons.location_on,
         size: 12,
         color: Colors.primaries[r.nextInt(Colors.primaries.length)]
     ),
 ),

2025-07-09.14-00-08.mp4

2025-07-09.14-12-43.mp4

[JaffaKetchup]

From what I can see, the lag occurs when markers are being added or pruned. Need to look into that. But the good news is, the culling still seems to work - when fewer markers are on screen (ignoring the initial spike caused by the culling itself somehow), it's more performant.

[JaffaKetchup]

LGTM, thanks!