importance filter error with spatial tasks

[h-a-graham]

Hey, wasn't sure whether to ask this here or in {mlr3filters} but here goes...

I may be on the wrong track so please tell me if so.

Creating an importance filter using a TaskClassif task results in an error like so. Is this the expected behaviour? Didn't want to file an issue in case this was inentional.

Please see below for a reprex - filter works fine with non spatial task but errors with the spatial one:

library(mlr3)
library(mlr3spatiotempcv)
library(mlr3filters)
library(mlr3learners)
library(sf)
#> Linking to GEOS 3.10.2, GDAL 3.4.3, PROJ 8.2.1; sf_use_s2() is TRUE

# load sample points
data_sf = sf::st_as_sf(ecuador, coords = c("x", "y"), crs = 32717)

# create `Task` 
task_no_sp <- TaskClassif$new(id="not_spat", backend=st_drop_geometry(data_sf), target="slides")
# create `TaskClassifST` from `sf` object
task_sp = as_task_classif_st(data_sf, id = "ecuador_task", target = "slides", positive = "TRUE")

spcv = rsmp("spcv_coords", folds = 5)

lrn = lrn("classif.ranger", importance = "impurity")
filter_no_sp = flt("importance", learner = lrn)

filter_no_sp$calculate(task_no_sp)
print(filter_no_sp)
#> <FilterImportance:importance>: Importance Score
#> Task Types: classif
#> Properties: -
#> Task Properties: -
#> Packages: mlr3, mlr3learners, ranger
#> Feature types: logical, integer, numeric, character, factor, ordered
#>            feature     score
#>  1: distslidespast 48.028203
#>  2:            dem 45.147079
#>  3:         cslope 40.519218
#>  4:          slope 39.185849
#>  5:          hcurv 36.300626
#>  6:          carea 34.861884
#>  7:      log.carea 34.831498
#>  8:          vcurv 34.188082
#>  9:   distdeforest 13.428703
#> 10:       distroad  6.428149

filter_sp = flt("importance", learner = lrn)

filter_sp$calculate(task_sp)
#> Error: Filter 'importance' does not support the type 'classif_st' of task 'ecuador_task'

^{Created on 2023-01-16 by the reprex package (v2.0.1)}

Any help would be awesome - thanks so much!

[pat-s]

Thanks for the reprex!

This is due to the recent mlr3filters 0.7.0 release.
Installing mlr3filters dev should solve the issue.
I've only checked briefly though, it might be that more work is needed.

cc @mllg

[h-a-graham]

Thanks @pat-s - yes you're correct - I was running ‘0.7.0’. Installing the dev version from Github (‘0.7.0.9000’) solves this!

Cheers!