#' @param e2 numeric, or object of class \code{sfg}; in case \code{e1} is of class \code{sfc} also an object of class \code{sfc} is allowed
#'
#' @details in case \code{e2} is numeric, +, -, *, /, %% and %/% add, subtract, multiply, divide, modulo, or integer-divide by \code{e2}. In case \code{e2} is an n x n matrix, * matrix-multiplies and / multiplies by its inverse. If \code{e2} is an \code{sfg} object, |, /, & and %/% result in the geometric union, difference, intersection and symmetric difference respectively, and \code{==} and \code{!=} return geometric (in)equality, using \link{st_equals}.
#'
#' If \code{e1} is of class \code{sfc}, and \code{e2} is a length 2 numeric, then it is considered a two-dimensional point (and if needed repeated as such) only for operations \code{+} and \code{-}, in other cases the individual numbers are repeated; see commented examples.
#' @param x object of class \code{sf}, \code{sfc} or \code{sfg}
#' @param y object of class \code{sf}, \code{sfc} or \code{sfg}; if missing, \code{x} is used
#' @param sparse logical; should a sparse index list be returned (TRUE) or a dense logical matrix? See below.
#' @param prepared logical; prepare geometry for x, before looping over y? See Details.
#' @details If \code{prepared} is \code{TRUE}, and \code{x} contains POINT geometries and \code{y} contains polygons, then the polygon geometries are prepared, rather than the points.
#' @return If \code{sparse=FALSE}, \code{st_predicate} (with \code{predicate} e.g. "intersects") returns a dense logical matrix with element \code{i,j} \code{TRUE} when \code{predicate(x[i], y[j])} (e.g., when geometry of feature i and j intersect); if \code{sparse=TRUE}, an object of class \code{\link{sgbp}} with a sparse list representation of the same matrix, with list element \code{i} an integer vector with all indices j for which \code{predicate(x[i],y[j])} is \code{TRUE} (and hence \code{integer(0)} if none of them is \code{TRUE}). From the dense matrix, one can find out if one or more elements intersect by \code{apply(mat, 1, any)}, and from the sparse list by \code{lengths(lst) > 0}, see examples below.
#' @details For most predicates, a spatial index is built on argument \code{x}; see \url{http://r-spatial.org/r/2017/06/22/spatial-index.html}.
#' Specifically, \code{st_intersects}, \code{st_disjoint}, \code{st_touches} \code{st_crosses}, \code{st_within}, \code{st_contains}, \code{st_contains_properly}, \code{st_overlaps}, \code{st_equals}, \code{st_covers} and \code{st_covered_by} all build spatial indexes for more efficient geometry calculations. \code{st_relate}, \code{st_equals_exact}, and \code{st_is_within_distance} do not.
#' Geometric unary operations on simple feature geometry sets
#'
#' Geometric unary operations on simple feature geometry sets. These are all generics, with methods for \code{sfg}, \code{sfc} and \code{sf} objects, returning an object of the same class.
#' Geometric unary operations on simple feature geometries. These are all generics, with methods for \code{sfg}, \code{sfc} and \code{sf} objects, returning an object of the same class. All operations work on a per-feature basis, ignoring all other features.
#' @name geos_unary
#' @param x object of class \code{sfg}, \code{sfg} or \code{sf}
#' @param dist numeric; buffer distance for all, or for each of the elements in \code{x}; in case
#' @details \code{st_simplify} simplifies lines by removing vertices
#' @param preserveTopology logical; carry out topology preserving simplification? May be specified for each, or for all feature geometries.
#' @param preserveTopology logical; carry out topology preserving simplification? May be specified for each, or for all feature geometries. Note that topology is preserved only for single feature geometries, not for sets of them.
#' @param dTolerance numeric; tolerance parameter, specified for all or for each feature geometry.
#' @details \code{st_segmentize} adds points to straight lines
#' @export
#' @param dfMaxLength maximum length of a line segment. If \code{x} has geographical coordinates (long/lat), \code{dfMaxLength} is either a numeric expressed in meter, or an object of class \code{units} with length units or unit \code{rad} or \code{degree}; segmentation takes place along the great circle, using \link[lwgeom]{st_geod_segmentize}.
#' @param dfMaxLength maximum length of a line segment. If \code{x} has geographical coordinates (long/lat), \code{dfMaxLength} is either a numeric expressed in meter, or an object of class \code{units} with length units \code{rad} or \code{degree}; segmentation in the long/lat case takes place along the great circle, using \link[lwgeom]{st_geod_segmentize}.
#' @param ... ignored
#' @examples
#' sf = st_sf(a=1, geom=st_sfc(st_linestring(rbind(c(0,0),c(1,1)))), crs = 4326)
#' @param precision numeric; see \link{st_as_binary} for how to do this.
#' @details Setting a \code{precision} has no direct effect on coordinates of geometries, but merely set an attribute tag to an \code{sfc} object. The effect takes place in \link{st_as_binary} or, more precise, in the C++ function \code{CPL_write_wkb}, where simple feature geometries are being serialized to well-known-binary (WKB). This happens always when routines are called in GEOS library (geometrical operations or predicates), for writing geometries using \link{st_write} or \link{write_sf}, \code{st_make_valid} in package \code{lwgeom}; also \link{aggregate} and \link{summarise} by default union geometries, which calls a GEOS library function. Routines in these libraries receive rounded coordinates, and possibly return results based on them. \link{st_as_binary} contains an example of a roundtrip of \code{sfc} geometries through WKB, in order to see the rounding happening to R data.
#' @param precision numeric, or object of class \code{units} with distance units (but see details); see \link{st_as_binary} for how to do this.
#' @details If \code{precision} is a \code{units} object, the object on which we set precision must have a coordinate reference system with compatible distance units.
#'
#' Setting a \code{precision} has no direct effect on coordinates of geometries, but merely set an attribute tag to an \code{sfc} object. The effect takes place in \link{st_as_binary} or, more precise, in the C++ function \code{CPL_write_wkb}, where simple feature geometries are being serialized to well-known-binary (WKB). This happens always when routines are called in GEOS library (geometrical operations or predicates), for writing geometries using \link{st_write} or \link{write_sf}, \code{st_make_valid} in package \code{lwgeom}; also \link{aggregate} and \link{summarise} by default union geometries, which calls a GEOS library function. Routines in these libraries receive rounded coordinates, and possibly return results based on them. \link{st_as_binary} contains an example of a roundtrip of \code{sfc} geometries through WKB, in order to see the rounding happening to R data.
#'
#' The reason to support precision is that geometrical operations in GEOS or liblwgeom may work better at reduced precision. For writing data from R to external resources it is harder to think of a good reason to limiting precision.
#' @seealso \link{st_as_binary} for an explanation of what setting precision actually does.
#'
#' @seealso \link{st_as_binary} for an explanation of what setting precision does, and the examples therein.
