Extents.extents ?

[alex-s-gardner]

What are peoples thoughts about adding

function extents(;X , Y) ex = Extents(X=extrema(X), Y = extrema(Y)) return ex end

[rafaqz]

What are X and Y? Arrays/ranges?

One problem is the extent is often not the start/end of a range, because the last step needs to be added for intervals.

[alex-s-gardner]

Can this be handled with multiple dispatch or are there too many edge cases?

[rafaqz]

How would you handle it with multiple dispatch? A Vector{Float64} can represent intervals or points, but have a different extent.

I dont really understand what you are proposing.

[alex-s-gardner]

OK, to me it seemed strait forward. If you have a dataset and you want to construct an Extent you could simply pass Extents.extents the coordinates or a range. I do this every time I create a new Tyler plot or want to check if two datasets have overlapping bounding boxes.

[rafaqz]

But what is the range

[alex-s-gardner]

For Extents.extents is would simply give the Extent of whatever is input.. it wouldn't attempt guess as to what the user intends by the input data. If a range was input then it would give the Extent of that range. If the user wanted something else then they would need to define their Extent manually. Extents.extents would only work if the inputs X and Y actually contain the extents. Does that make sense?

[rafaqz]

But what is your particular range from?

[rafaqz]

Im trying to tease out the problem that a range can represent two (even four) different extents, not one. And we dont know which.

We can use the start and end points as a point to point extent, of course. So (first(r), last(r)). But ranges have a step size, so we can also calculate the interval extent (first(r), last(r)+step(r)) if the values are for the starts of each pixel.

Its very common in raster or image data that the range of pixel point locations represent the start or center of intervals. When you pass a range of tick values to a heatmap in Makie, theyre the centers of intervals. So the extent is half a pizel larger on each side. So `(first(r) - step(r)/2, last(r)+step(r)/2).

And if you load a tif with Rasters.jl chances are thats what the lookup ranges mean. When you call extent on them this math is done for you because Rasters.jl tracks these things in the wrapper.

[alex-s-gardner]

OK, I follow but in my eyes this has already been defined extrema(Range)... if the user is using Range in a way in which the Extents are not defined by Range alone then they would need to explicitly define their Extent ... the function could be defined as:

    extents(; X, Y)
Returns the Extent based on the minimum and maximum values of inputs X and Y

[alex-s-gardner]

Pixel centers vs. pixel edges is an issue that has plagued raster data for decades... I wish we had only opted for a single convention

[rafaqz]

I think people can just call X=extrema(xs) themselves tho if they mean that?

I prefer that to the swathes of inevetably wrong Makie plots for the next decade that people will bug me about.

[alex-s-gardner]

Fair point.