Joins between data.frames and sf objects

[Nowosad]

Second part of this issue - r-spatial/sf#372

Lately I've been trying to use joins for pairs of sf objects and data.frames and I came across a problematic joins. When a data.frame is the main object in join., then a new object has a geometry column, but doesn't have a sf class.
My expectation is that a geom column should be removed in these cases. I'm not sure, if my expectation is the best. What do you think?

library(tidyverse)
library(sf)

sf_obj = st_read(system.file("shape/nc.shp", package = "sf")) %>% filter(NAME %in% 
  c("Ashe", "Surry")) %>% select(NAME)
#> Reading layer `nc' from data source `/home/jn/R/x86_64-redhat-linux-gnu-library/3.3/sf/shape/nc.shp' using driver `ESRI Shapefile'
#> Simple feature collection with 100 features and 14 fields
#> geometry type:  MULTIPOLYGON
#> dimension:      XY
#> bbox:           xmin: -84.32385 ymin: 33.88199 xmax: -75.45698 ymax: 36.58965
#> epsg (SRID):    4267
#> proj4string:    +proj=longlat +datum=NAD27 +no_defs

df_obj = data.frame(NAME = c("Ashe", "Surry", "Rowan"), VALUE = c(1, 4, 6))

left_join1 = df_obj %>% left_join(sf_obj, by = "NAME")
#> Warning: Column `NAME` joining factors with different levels, coercing to
#> character vector
left_join1
#>    NAME VALUE                       geometry
#> 1  Ashe     1 MULTIPOLYGON(((-81.47275543...
#> 2 Surry     4 MULTIPOLYGON(((-80.45634460...
#> 3 Rowan     6                           NULL

right_join2 = df_obj %>% right_join(sf_obj, by = "NAME")
#> Warning: Column `NAME` joining factors with different levels, coercing to
#> character vector
right_join2
#>    NAME VALUE                       geometry
#> 1  Ashe     1 MULTIPOLYGON(((-81.47275543...
#> 2 Surry     4 MULTIPOLYGON(((-80.45634460...

inner_join1 = df_obj %>% inner_join(sf_obj, by = "NAME")
#> Warning: Column `NAME` joining factors with different levels, coercing to
#> character vector
inner_join1
#>    NAME VALUE                       geometry
#> 1  Ashe     1 MULTIPOLYGON(((-81.47275543...
#> 2 Surry     4 MULTIPOLYGON(((-80.45634460...

full_join2 = df_obj %>% full_join(sf_obj, by = "NAME")
#> Warning: Column `NAME` joining factors with different levels, coercing to
#> character vector
full_join2
#>    NAME VALUE                       geometry
#> 1  Ashe     1 MULTIPOLYGON(((-81.47275543...
#> 2 Surry     4 MULTIPOLYGON(((-80.45634460...
#> 3 Rowan     6                           NULL

[hadley]

Could you please use the reprex package so I can see the outputs?

[Nowosad]

Yes, of course. I've updated my question.

[hadley]

Keeping the geometry column seems like the right behaviour to me.