Nicer handling of spatial args

[jcheng5]

• Most layer functions take lng, lat instead of lat, lng now
• Layers that can take formulas now also take a data override param
• Lots of ways to assign lng/lat values now: NULL to infer from data, formula evaluated against data, or direct. (You can now do leaflet(data) %>% addCircles())
• Don't serialize entire data object to client

[hadley]

Do you need a fallback error message here? (If no lng/lat cars found)

[hadley]

Looks good!

[jcheng5]

Thanks for the feedback.

[yihui]

Are there any other packages using the convention of "lng first lat second"? I thought a while on this issue again. Although lat is for the y-axis and lng for x-axis, by x first y second, the natural order for lat/lng seems to be lng, lat if we think on a Cartesian plane, but I guess the more natural order in English is lat, lng. Interestingly, the natural language order is lng, lat in Chinese :)

[hadley]

@yihui I'm pretty sure this is what sp does.

[yihui]

OK, I just looked at leafletjs.com and it uses the convention of lat, lng when the user provides a coordinate [x, y]. Personally I'm more comfortable with this.

[hadley]

I think it would be bad to adopt a convention (y, x) that's different to all other graphics packages in R

[yihui]

The question is, when you look at a map, do you think in a Cartesian coordinate system as in other R plots, or you think about latitudes/longitudes? I bet most people think about the latter, so they would not even notice they are actually using [y, x]. In fact, it was not until this morning that I realized latitudes were actually for the y-axis. I always thought latitudes were along the x-axis because I had been thinking of the latitude lines on the earth...

I looked at sp again, and it considers the data as [x, y] coordinates instead of latitudes/longitudes specifically. Based on the abstraction on the x-y plane, it makes sense to put x before y, but when we consider lat/lng specifically, I feel the order of function arguments is naturally lat before lng. When you click on a Google map, it also shows you the location [x, y] where x is lat and y is lng. So my slight preference would be lat before lng.

[jcheng5]

I keep going back and forth on this as well. Geojson does lng lat. I think
sp, ogr, etc can theoretically work with either depending on the
proj4string but lng lat is more natural.

That said I keep typing lat lng over and over by mistake, to the point that
I was flirting with the idea of linting using deparse(substitute()).
On Dec 10, 2014 8:33 AM, "Yihui Xie" notifications@github.com wrote:

The question is, when you look at a map, do you think in a Cartesian
coordinate system as in other R plots, or you think about
latitudes/longitudes? I bet most people think about the latter, so they
would not even notice they are actually using [y, x]. In fact, it was not
until this morning that I realized latitudes were actually for the y-axis.
I always thought latitudes were along the x-axis because I had been
thinking of the latitude lines on the earth...

I looked at sp again, and it considers the data as [x, y] coordinates
instead of latitudes/longitudes specifically. Based on the abstraction
on the x-y plane, it makes sense to put x before y, but when we consider
lat/lng specifically, I feel the order of function arguments is naturally
lat before lng. When you click on a Google map, it also shows you the
location [x, y] where x is lat and y is lng. So my slight preference
would be lat before lng.

—
Reply to this email directly or view it on GitHub
#12 (comment).

[yihui]

I just asked on Twitter, and apparently there is no consensus 😭 Let's just keep the order lng/lat then. Linting sounds helpful, which is also what Barry mentioned: https://twitter.com/geospacedman/status/542722245572165633