handle various column classes

[mdsumner]

This PR adds a slightly safer way of comparing column values to "{ }" and adds a curly_brace_na.data.frame method that only compares character and factor columns.

The primary motivation was to enable round-tripping with Date and POSIXct columns, which error currently (5a293b0):

library(sf)
example(st_read)
x <- st_sf(date = Sys.Date() + seq_len(2), geometry = st_geometry(nc)[1:2])
rmapshaper::ms_simplify(x)
# Error in charToDate(x) : 
 # character string is not in a standard unambiguous format

y <- st_sf(date = Sys.time() + seq_len(2), geometry = st_geometry(nc)[1:2])
rmapshaper::ms_simplify(y)
#Error in as.POSIXlt.character(x, tz, ...) : 
# character string is not in a standard unambiguous format 

I had to remove the Date metadata here in order to run ms_simplify, as a real world example:

https://github.com/mdsumner/ozplot/blame/master/README.Rmd#L100

I kind of remember discussing this at some point, so apologies if I've missed something! I think there will be other details and perhaps problematic and column types, but I feel like this is a reasonable PR without being comprehensive yet. :)

There's a further problem in that date-time types don't always(?) round-trip completely, and I assume that's a time zone thing - but it wouldn't have been noticed. I'll wait until this is accepted before pursuing those details, for now I've left out the round-trip test, commented out here:

https://github.com/ateucher/rmapshaper/pull/68/files#diff-792448277e53b51a3b944f1b1873e7f8R515

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@5a293b0). Click here to learn what that means.
The diff coverage is 100%.

@@            Coverage Diff            @@
##             master      #68   +/-   ##
=========================================
  Coverage          ?   96.76%           
=========================================
  Files             ?       11           
  Lines             ?      587           
  Branches          ?        0           
=========================================
  Hits              ?      568           
  Misses            ?       19           
  Partials          ?        0

Impacted Files	Coverage Δ
R/utils.R	94.69% <100%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5a293b0...89ea2e2. Read the comment docs.

[ateucher]

Thanks @mdsumner this is great! I had actually just made an initial attempt at this in another branch, but not as comprehensive as yours: https://github.com/ateucher/rmapshaper/blob/mapshaper_v0.4.x/R/utils.R#L262-L270. The other thing I did was rather than use as.data.frame I just removed the "sf" class, because with the former it also removed tibble classes, so one would input a tibble sf object, and output a regular data.frame sf object.

Thanks so much for the tests as well. I will take this PR here and meld the best parts of our two solutions together.

Yeah, the date stuff is something I haven't tackled yet... so if you 're game, please feel free!

[ateucher]

@mdsumner FYI I merged your solution into mine, with a few tweaks, on the mapshaper_v0.4x branch here:

rmapshaper/R/utils.R

Lines 260 to 282 in b3d7c33

	curly_brace_na.data.frame <- function(x) {
	chr_or_factor <- vapply(x, inherits, c("character", "factor"), FUN.VALUE = logical(1))
	if (any(chr_or_factor)) {
	x[, chr_or_factor][x[, chr_or_factor] == "{ }"] <- NA
	}
	x
	}
	curly_brace_na.Spatial <- function(x) {
	x@data <- curly_brace_na(x@data)
	x
	}
	# This method basically just removes the sf class and then
	# restores it after the data.frame method does its work, because
	# the sf column is 'sticky' with `[`.sf methods, so would be
	# included in the { } substitution if the sf class was kept
	curly_brace_na.sf <- function(x) {
	sf_col <- attr(x, "sf_column")
	class(x) <- setdiff(class(x), "sf")
	x <- curly_brace_na(x)
	sf::st_as_sf(x, sf_column_name = sf_col)
	}