support MULTILINESTRING

[timelyportfolio]

promoted from #48

Before I implement the logic into mapedit, I wanted to post the results from my initial exploration.

library(mapview)
library(mapedit)
library(sf)

# @tim-salabim reports, this fails badly
#   and while leaflet draws the multilinestring correctly
#   Leaflet.draw seems to not like it
editFeature(trails[4,])

# we will need to cast to LINESTRING first
#  but be careful to retain some mechanism
#  to preserve the group for proper conversion back
#  to MULTILINESTRING
editFeatures(sf::st_cast(trails[4,], "LINESTRING"))


# one option will be to add a second column identifier
#  after we cast to LINESTRING
#  here is a minimal example before attempting to integrate
#  into the existing module
tr <- trails[c(1,4),]
tr$edit_id <- seq_len(nrow(tr))
# now will need to carefully cast
#  for some reason casting creates sfc
#  so also need to st_sf to get back to sf
tr2 <- st_sf(st_cast(tr, "LINESTRING"))
tr2


# ignore the editMap part for now
#  and see if we can convert back
#  I might be making this way harder than it should be
#  but I could not find a combination of aggregate.sf
#  or dplyr::group_by that would accomplish this
tr3 <- tr
st_geometry(tr3) <- split(tr2, tr2$edit_id) %>%
  lapply(function(x){st_multilinestring(st_geometry(x))}) %>%
  st_sfc
plot(tr3)
tr3

@edzer, is there a way to group_by and then make st_multilinestring on each grouped edit_id? I also tried to do this with aggregate.sf with no luck since it seems the aggregate function could only apply to non-geometry columns.

[tim-salabim]

@timelyportfolio for re-casting to MULTILINESTRING you can use do.call(c, lapply(split(tr2, tr2$edit_id), st_combine)).

[timelyportfolio]

@tim-salabim, thanks for that bit of code. c will work well in this scenario. In a case with mixed types, won't st_combine convert everything to MULTI*? I will test to make sure that this will not interfere with our cast back to original (which I think is a very nice feature).

[timelyportfolio]

Here is my test sf (challenge) with mixed types to get working with edit.

# since we have discovered a way to go from MULTILINESTRING
#   to a group of LINESTRING and back, now the trick will be
#   to perform this magic in a sf with mixed types.
#   let's make a mixed type sf
mixsf <- st_sf(geometry = c(
  st_geometry(franconia[1:2,]),
  st_geometry(trails[1:4,]),
  st_geometry(breweries[1:5,])
))

[tim-salabim]

As long as it's not in a GEOMETRYCOLLECTION, st_dimension will show us the different types.

st_dimension(mixsf)
[1] 2 2 1 1 1 1 0 0 0 0 0

When encapsulated within a GEOMETRYCOLLECTION we need to st_cast(gc) first in order to get the individual components.

[timelyportfolio]

@tim-salabim, I probably need to clear my head, and I'm guessing the answer is obvious but why does this not work? Never mind, as I thought, was very dumb on my part. I need to st_transform.

library(sf)
library(mapview)
library(mapedit)

mixsf <- st_sf(geometry = c(
  st_geometry(franconia[1:2,]),
  st_transform(st_geometry(trails[1:4,]), crs=4326),
  st_geometry(breweries[1:5,])
))

mapview(mixsf)

The code above will not render the lines.

[tim-salabim]

Yeah, that one has caught me several times too. I left it delibarately in non-longlat as a gentle reminder that not all data comes in longlat...

Though this brings up a good point for mapedit *Features. We need to decide what we do with data supplied in non-longlat. mapview::checkAdjustProjection will ensure that data is projected correctly for rendering and drawn/edited features are currently returned in longlat. Should we transform back to initial supplied CRS? Or simply return longlat with a warning?

[timelyportfolio]

@tim-salabim looks like the easy solution is to use editor="leafpm", since Leaflet.pm handles MULTILINESTRING.

library(sf)
library(mapview)
library(mapedit)

editFeatures(trails[4,], editor="leafpm")

However, I'm not sure this is an acceptable solution for the project.

[timelyportfolio]

0.5.0 will merge in support for leaflet.pm using the new package leafpm. This handles MULTILINESTRING well, so I will consider this issue closed.

Please reopen if anyone would like to discuss further.