Support for CRS.Simple

[AB-Kent]

Would it be possible to support maps which are not in the standard projection? Specifically I would like to use addGlPoints with a map created using L.CRS.Simple. My base map is a microscope image of a tissue section; I would like to draw points representing hundreds of thousands of cells as a layer on top. It works (slowly) to use leaflet::addCircles with coordinates in the range 0-255; this fails with leafgl::addGlPoints because it tries to convert the points to EPSG:4326.

Example with no base map:

library(leaflet)
library(leafgl)
library(magrittr)

n=1000
pts = data.frame(id = 1:n,
                 x = rnorm(n, 128, 30),
                 y = rnorm(n, -128, 18))

crs <- leafletCRS(crsClass = "L.CRS.Simple")
map = leaflet(options = leafletOptions(crs = crs)) %>% 
  setView(128, -128, 1) 

# Works
map %>% addCircles(data=pts, lat=~y, lng=~x)

# Doesn't work
map %>% addGlPoints(st_as_sf(pts, coords=c('x', 'y')))

#Error in st_transform.sfc(st_geometry(x), crs, ...) : 
#  sfc object should have crs set
#Calls: %>% ... st_transform.sf -> st_transform -> st_transform.sfc

[AB-Kent]

The changes needed are small. There is one change needed in robertleeplummerjr/Leaflet.glify#29.

In this package, calls to sf::st_transform(data, 4326) need to be conditional on the CRS of the map:

if (map$x$options$crs$crsClass != 'L.CRS.Simple')
    data = sf::st_transform(data, 4326)

I have tried this in addGlPoints and (with the Leaflet.glify change) it fixes the example in my original post.

[AB-Kent]

The change proposed above only works if map is a leaflet object; if it is a leaflet_proxy then a different method is needed.

[tim-salabim]

@AB-Kent I've seen you've proposed a PR in the upstream repo. We should wait for it to be implemented there. I'll have a look at the leaflet_proxy issue once the upstream lib supports L.CRS.Simple

[AB-Kent]

Thanks! For the proxy, I think some way to query the browser is needed.

[AB-Kent]

The upstream pull request has been merged: robertleeplummerjr/Leaflet.glify#33

Do you have any thoughts about how to determine the CRS of a leaflet_proxy?

Why I care: I work with microscope images stored in pyramidal TIFF format. Using leaflet and a tile server, the images can be displayed in a zoomable, pannable image. I'm using Shiny and leafgl to display hundreds of thousands of points, representing different cell types, on top of the microscope image. I'm planning to release this viewer as open source once the tile server is available.

Here is a short screencast showcasing the technology. leafgl comes into use just after 2:00.
https://akoya.app.box.com/s/99ori44qxuz9r5j4sd520a7vj23lsg1l/file/538913743990

[tim-salabim]

Wow, this is an impressive app that you're showcasing there! Not that I have any clue what it does, but it is awesome to see how far a mapping engine can be stretched for different purposes!

In b92a8d8 I've simply removed the on-the-fly-reprojection to 4326 (which should not happen anyways). This should be the easiest approach to achieve what you want as we don't need to infer the crs from the map from anywhere (assuming you control the map setup in the first place).

library(leaflet)
library(leafgl)
library(mapview)
library(leafem)
library(sf)

m = leaflet::leaflet(
  options = leaflet::leafletOptions(
    minZoom = -1000,
    crs = leafletCRS(crsClass = "L.CRS.Simple")
  )
)

brew = st_transform(breweries, 3035)
fran = st_transform(st_cast(franconia, "POLYGON"), 3035)
trls = st_transform(st_cast(trails, "LINESTRING"), 3035)

m %>%
  addGlPoints(brew, popup = "founded") %>%
  addGlPolygons(fran) %>%
  addGlPolylines(trls) %>%
  setView(lng = 4389376, lat = 2964480, zoom = -9) %>%
  addMouseCoordinates()

now gives

The only thing that isn't working with L.CRS.Simple are lines. No idea why though... Maybe something in https://github.com/robertleeplummerjr/Leaflet.glify @robertleeplummerjr ?

[AB-Kent]

Yes that works brilliantly, thank you! I'm not using lines with leafgl at this time. The breakage there seems to be independent of this change.

[tim-salabim]

Good to hear that things work for you now.

[tim-salabim]

UPDATE: L.CRS.Simple now also working with lines.

The above code snippet now gives:

I think we can close this here. Please re-open if you disagree.

[AB-Kent]

Nice, thank you!