filter_() methods for Spatial*DataFrame objects

filter_-sp-S3-methods.R

#' Filter (subset) a Spatial*DataFrame object
#'
#' @param object a \code{Spatial*DataFrame}
#' @param ... see \link{subset}
#' @param .dots
#'
#' @importFrom lazyeval all_dots lazy_eval
#'
#' @return a subset of the original \code{Spatial*DataFrame}
#'
#' @seealso subset
#'
#' @examples
#' library(dplyr)
#' library(sp)
#' data(meuse, package = "sp")
#' j <- which(names(meuse) %in% c("x", "y"))
#' meuse_spdf <- SpatialPointsDataFrame(meuse[, j], meuse[, -j])
#' high_copper <- filter(meuse_spdf, copper > median(copper))
#' stopifnot(inherits(high_copper, class(meuse_spdf)))
#'
#' @export
filter_.SpatialDataFrame <- function (object, ..., .dots) {
  dots <- lazyeval::all_dots(.dots, ..., all_named = TRUE)
  masks <- lazyeval::lazy_eval(dots, data = as.data.frame(object@data))
  subset(object, Reduce(`&`, masks))
}

# FIXME: @describeIn makes roxygen complain "Don't know how to describe s3method in s3method"

# @describeIn filter_.SpatialDataFrame
filter_.SpatialLinesDataFrame <- filter_.SpatialDataFrame

# @describeIn filter_.SpatialDataFrame
filter_.SpatialPointsDataFrame <- filter_.SpatialDataFrame

# @describeIn filter_.SpatialDataFrame
filter_.SpatialPolygonsDataFrame <- filter_.SpatialDataFrame

Suspected issues

Please suggest fixes in the form of comments! :-)

This code should maybe:

• invoke NextMethod and/or UseMethod
• contain @method in the roxygen skeleton

Let's figure out whose repository it should live in, and then fix it up! :-)

I don't think your example here applies; as meuse is already a data frame

Hi @dgrtwo! Almost --- meuse is a data.frame, but the statement:

coordinates(meuse) <- ~x+y

... promotes meuse to a SpatialPointsDataFrame. (Same syntax found in help(meuse)).

dplyr::filter doesn't work out-of-the-box with Spatial*DataFrame objects. What will work is subset, but it needs a little help in the absence of non-standard evaluation.

I'll edit the example to make this a little more clear. Thanks for the comment!

Edited. Also fixed && where & should have been. World of difference.

I find this function super useful and wish dplyr would support Spatial*DataFrame directly! Would be great to have the same for all the dplyr verbs. Particularly mutate, transmute, select, slice, rename. Others don't really make sense or are more complicated. group_by, for example, would be nice in combination with mutate but summarise would require joining polygons, which could be useful but more complicated.

Edit: All the joins as well (left_join etc)

I've done mutate, transmute, select, slice, rename, arrange and distinct. Shall we create a simple package for just these methods? Here's what I have: https://gist.github.com/mdsumner/85d76f8b33e26a427c3d Originally I was seeing this through a different lens of flattening to a table - but that might be a good way to handle joins.

I think summarise is best left undefined for now.

There's probably a bunch of things I'm not doing properly here, so happy for any feedback / discussion.

And here's my packaging of it: https://github.com/mdsumner/spbabel