Interactive maps in articles

[mtennekes]

Interactive maps don't work yet... https://r-tmap.github.io/tmap/articles/32_versus_mapview.html

They do work in mapview: https://r-spatial.github.io/mapview/articles/mapview_01-basics.html

@tim-salabim I checked your rmd code, but couldn't find out what did the trick.
(Also check out the content, and let me know your thoughts)

[tim-salabim]

I wouldn't trust what you see in the mapview pages right now. They're probably ancient, as the daily runs have been failing for quite some time now. The only thing I could imagine right from the top of my head is that mapview's knit_print is simply using the leaflet part (so relying on Posit to make sure things work).

Regarding content, there's only one thing. I'd prefer if you used zcol instead of z here

[mtennekes]

@olivroy do you have any idea how to fix this? Not only relevant for this vignette, but also for a tmap-basics vignette about switching modes.

FYI: this is the last mapview commit that passed the automatic tests: https://github.com/r-spatial/mapview/tree/4202de2257214bd3c22b68fa8790e9596736fe83

[olivroy]

Sorry for being unresponsive.. It seems very strange... I did some housekeeping in mapview to revive GitHub pages deployment r-spatial/mapview#497. And it seems that the maps are correctly deployed on their site :/
https://r-spatial.github.io/mapview/ in the vignettes look fine.

Maybe tmap is doing too much with its knit_print() method somehow? Maybe we could try if using dev leafem for deployment helps.. I am sending a PR shortly.

[olivroy]

Maybe there is some sort of token that is required ?

I remember this from leaflet a couple months ago rstudio/leaflet#913 (comment) rstudio/leaflet#912

Maybe this is related?

[olivroy]

dev leafem didn't work.. It is probably an issue with tmap knit_print method.. leaflet doesn't provide a knit_print() method. And it seems that tmap is resorting to knitr.. Perhaps it should delegate to `htmlwidget instead? htmlwidgets provides htmlwidgets:::knit_print.htmlwidget.

Maybe that if active. tmap should set the class to "htmlwidget" to help with S3 dispatch..

A leaflet map has class = c("leaflet", "htmlwidget")

[mtennekes]

Very strange indeed. Also strange that leaflet doesn't provide a knit_print() method, and still works (at least, I assume)

[tim-salabim]

Very strange indeed. Also strange that leaflet doesn't provide a knit_print() method, and still works (at least, I assume)

I guess that's because a leaflet map is an htmkwidget

[mtennekes]

I guess that's because a leaflet map is an htmkwidget

Yes that's true. So what happens if we remove knit_print.tmap? The tmap small multiples are returned as objects of class "shiny.tag.list" (when tmap_leaflet is used), and there is a htmltools::knit_print.shiny.tag.list. So that should also be covered.

[tim-salabim]

That would be great!

[mtennekes]

Not successful yet. After removing the knit_print.tmap, no maps are included.

It boils down to this problem: tmap_leaflet returns a leaflet object, but it is not printed (also not in the console). In the console, I have to print it:

print(tmap_leaflet(tm_shape(World) + tm_polygons()))

to make it appear.

However, in an Rmd file, this chunk:

tmap_mode("view")
print(tmap_leaflet(tm_shape(World) + tm_polygons()))

doesn't produce a map.

This does work:

library(leaflet)
leaflet() |> addTiles()

[mtennekes]

(in addition)

Does not work:

tmap_mode("view")
tmap_leaflet(tm_shape(World) + tm_polygons())

Works:

tmap_mode("view")
lf = tmap_leaflet(tm_shape(World) + tm_polygons())
lf

[mtennekes]

Even printing lf doesn't work with pkgdown. I did that in the tmap-mapview vignette:

Source: https://github.com/r-tmap/tmap/blob/master/vignettes/32_versus_mapview.Rmd
Rendered: https://r-tmap.github.io/tmap/articles/32_versus_mapview.html

as an attempt to workaround this issue.

@olivroy @tim-salabim any other tricks up your sleeves?

[tim-salabim]

Not really, just noticed that the rmd headers between tmap and mapview are quite different...

[olivroy]

Not really, just noticed that the rmd headers between tmap and mapview are quite different...

Okay, this was the issue. #1021 fixed that. pkgdown: as-is: TRUE was probably the problem. Also height / width / retina was incorrect for interactive maps...

I will see if I didn't create new issues with the PR..