Remove no-argument LatLngBounds constructor

[tmcw]

https://github.com/mapbox/llmr/blob/master/js/geometry/latlngbounds.js#L8

This seems like a trap: if you call new LatLngBounds() no error is thrown but calling any method off of the object will fail.

[mourner]

It's often useful, e.g. when you create empty bounds and then run some algorithm that calls bounds.extend(point) many times to form the bounds. Leaflet uses this pattern in some places.

[tmcw]

Hm, okay - can't believe I'm saying this, but this might be a good place for a Builder that would facilitate that usecase but would manage the risk of unitialized LatLngBounds objects

[mourner]

What do you mean by Builder?

[tmcw]

var builder = latLngBoundsBuilder();

someListOfLatLngs.forEach(function(latlng) {
  builder.expand(latlng);
});

builder.getValue();

I know, it's a little silly, and it's a pattern, but we do this in the android sdk and it does address the problem of ephemeral invalid data pretty well

[mourner]

address the problem of ephemeral invalid data

I assume you meant to say "address the ephemeral problem of invalid data", right?

Seriously though, I don't think it's a problem. No one ever reported about it in Leaflet. Creating new API entities to protect against rare cases of invalid API use just isn't worth it. Otherwise we could for example bloat the API by asserting arguments in each method as well.

[mourner]

Also, Leaflet has bounds.isValid() check which can be used if necessary.

[tmcw]

¯\_(ツ)_/¯ okay, let's leave it as is.