Nothing
R/slice.R
In MatrixExtra: Extra Methods for Sparse Matrices
Defines functions
subset_csc_masked
subset_generic_with_vector
subset_csr
drop_slice
inject_NAs
get_ij_properties
ij_is_in_lower_triangle
get_indices_integer
get_indices_integer <- function(i, max_i, index_names) {
if (inherits(i, c("lsparseVector", "nsparseVector"))) {
i <- sort_sparse_indices(i, copy=!getOption("MatrixExtra.inplace_sort", default=FALSE))
}
if (inherits(i, "lsparseVector")) {
if (anyNA(i@x))
i <- seq(1L, max_i)[as.logical(i)]
else
i <- as(i, "nsparseVector")
}
if (inherits(i, "nsparseVector")) {
if (i@length > max_i) {
stop("Dimension of indexing vector is larger than matrix to subset.")
} else if (i@length == max_i) {
i <- as.integer(i@i)
} else { ### mimic of R base's recycling
full_repeats <- as.integer(max_i %/% i@length)
remainder <- as.integer(max_i - i@length*full_repeats)
if (remainder != 0)
i <- repeat_indices_n_times(as.integer(i@i), as.integer(i[seq(1L, remainder)]@i), as.integer(i@length), max_i)
else
i <- repeat_indices_n_times(as.integer(i@i), integer(), as.integer(i@length), max_i)
}
if (typeof(i) != "integer")
i <- as.integer(i)
}
if (is.character(i)) {
if (!is.null(index_names))
i <- match(i, index_names)
else
i <- as.integer(i)
}
if (is.logical(i)) {
if (length(i) != max_i) {
i <- seq(1L, max_i)[i]
} else {
i <- which(i)
}
}
if (typeof(i) != "integer")
i <- as.integer(i)
if (any(!is.na(i) & i <= 0))
i <- seq(1L, max_i)[i]
if(NROW(i) && any(i > max_i, na.rm=TRUE))
stop("some of row subset indices are not present in matrix")
return(as.integer(i))
}
ij_is_in_lower_triangle <- function(i, j) {
return(i > j)
}
get_ij_properties <- function(x, i, j) {
row_names <- rownames(x)
col_names <- colnames(x)
all_i <- FALSE
all_j <- FALSE
i_is_seq <- FALSE
j_is_seq <- FALSE
i_is_rev_seq <- FALSE
j_is_rev_seq <- FALSE
i_has_NA <- FALSE
j_has_NA <- FALSE
if (missing(j)) {
all_j <- TRUE
j <- seq_len(ncol(x))
n_col <- ncol(x)
} else {
j <- get_indices_integer(j, ncol(x), col_names)
j_has_NA <- anyNA(j)
if (!j_has_NA) {
if (length(j) == ncol(x) && j[1L] == 1L && j[length(j)] == ncol(x)) {
if (check_is_seq(j))
all_j <- TRUE
} else {
j_is_seq <- check_is_seq(j)
}
}
n_col <- length(j)
}
if (missing(i)) {
i <- seq_len(nrow(x))
all_i <- TRUE
i_is_seq <- TRUE
n_row <- nrow(x)
} else {
i <- get_indices_integer(i, nrow(x), row_names)
i_has_NA <- anyNA(i)
if (!i_has_NA)
i_is_seq <- check_is_seq(i)
n_row <- length(i)
if (i_is_seq && length(i) == nrow(x) && i[1L] == 1L && i[length(i)] == nrow(x)) {
if (check_is_seq(i))
all_i <- TRUE
}
}
if (!all_i && !i_is_seq && !i_has_NA)
i_is_rev_seq <- check_is_rev_seq(i)
if (!all_j && !j_is_seq && !j_has_NA)
j_is_rev_seq <- check_is_rev_seq(j)
i_all_NA <- FALSE
j_all_NA <- FALSE
i_NAs <- NULL
j_NAs <- NULL
if (i_has_NA) {
i_temp <- find_first_non_na(i)
if (is.na(i_temp)) {
i_all_NA <- TRUE
} else {
i_NAs <- which(is.na(i))
i[i_NAs] <- i_temp
}
}
if (j_has_NA) {
j_temp <- find_first_non_na(j)
if (is.na(j_temp)) {
j_all_NA <- TRUE
} else {
j_NAs <- which(is.na(j))
j[j_NAs] <- j_temp
}
}
return(list(
i = i, all_i = all_i, i_is_seq = i_is_seq, i_is_rev_seq = i_is_rev_seq,
j = j, all_j = all_j, j_is_seq = j_is_seq, j_is_rev_seq = j_is_rev_seq,
i_has_NA = i_has_NA, i_all_NA = i_all_NA, i_NAs = i_NAs,
j_has_NA = j_has_NA, j_all_NA = j_all_NA, j_NAs = j_NAs,
row_names = row_names, col_names = col_names,
n_row = n_row, n_col = n_col
))
}
inject_NAs <- function(x, i_NAs, j_NAs) {
if (NROW(i_NAs) || NROW(j_NAs)) {
if (inherits(x, "sparseMatrix")) {
if (NROW(i_NAs) && !is.null(x@Dimnames[[1L]]))
x@Dimnames[[1L]][i_NAs] <- NA_character_
if (NROW(j_NAs) && !is.null(x@Dimnames[[2L]]))
x@Dimnames[[2L]][j_NAs] <- NA_character_
} else {
if (NROW(i_NAs) && !is.null(rownames(x)))
rownames(x)[i_NAs] <- NA
if (NROW(j_NAs) && !is.null(rownames(x)))
rownames(x)[j_NAs] <- NA
}
output_logical <- inherits(x, c("nsparseMatrix", "lsparseMatrix"))
if (inherits(x, "RsparseMatrix")) {
x <- as.csr.matrix(x, logical=output_logical)
} else if (inherits(x, "TsparseMatrix")) {
x <- as.coo.matrix(x, logical=output_logical)
}
if (inherits(x, "RsparseMatrix")) {
x <- as.csr.matrix(x)
if (NROW(i_NAs))
x <- assign_csr_internal(x, i_NAs, seq(1L, NCOL(x)), NA_real_)
if (NROW(j_NAs))
x <- assign_csr_internal(x, seq(1L, NROW(x)), j_NAs, NA_real_)
if (output_logical)
x <- as.csr.matrix(x, logical=TRUE)
}
else if (inherits(x, "TsparseMatrix")) {
if (!NROW(i_NAs))
i_NAs <- integer()
if (!NROW(j_NAs))
j_NAs <- integer()
if (output_logical) {
out <- new("lgTMatrix")
res <- inject_NAs_inplace_coo_logical(x@i, x@j, x@x, i_NAs-1L, j_NAs-1L, nrow(x), ncol(x))
} else {
out <- new("dgTMatrix")
res <- inject_NAs_inplace_coo_numeric(x@i, x@j, x@x, i_NAs-1L, j_NAs-1L, nrow(x), ncol(x))
}
out@i <- res$ii
out@j <- res$jj
out@x <- res$xx
out@Dim <- x@Dim
x <- out
}
else if (inherits(x, "float32")) {
x[i_NAs, ] <- float::fl(NA_real_)
x[, j_NAs] <- float::fl(NA_real_)
}
else {
x[i_NAs, ] <- NA
x[, j_NAs] <- NA
}
}
return(x)
}
drop_slice <- function(x, drop, i_NAs=NULL, j_NAs=NULL) {
x <- inject_NAs(x, i_NAs, j_NAs)
if ((missing(drop) || isTRUE(drop)) && inherits(x, c("sparseMatrix", "float32"))) {
if (nrow(x) == 1L || ncol(x) == 1L) {
if (!(nrow(x) == 1L && ncol(x) == 1L) && !inherits(x, "float32") &&
getOption("MatrixExtra.drop_sparse", default=FALSE)
) {
x <- as(x, "sparseVector")
} else {
x <- as.vector(x)
}
}
}
if (inherits(x, "sparseMatrix")) {
Dimnames <- list(NULL, NULL)
if (NROW(x@Dimnames[[1L]]))
Dimnames[[1L]] <- x@Dimnames[[1L]]
if (NROW(x@Dimnames[[2L]]))
Dimnames[[2L]] <- x@Dimnames[[2L]]
x@Dimnames <- Dimnames
}
return(x)
}
#' @title Sparse Matrices Slicing
#' @description Natively slice CSR/CSC/COO matrices without changing the storage order.
#' @details \bold{Important:} When slicing sparse matrices with `drop=TRUE` (the default),
#' `Matrix` will drop 1-d matrices to \bold{dense} dense vectors, whereas
#' this package allows dropping them to either dense or \bold{sparse} vectors,
#' the latter of which is more efficient and is the default option.
#'
#' The `drop` behavior can be changed back to dense vectors like `Matrix` does,
#' through \link{restore_old_matrix_behavior} or through the package options
#' (e.g. `options("MatrixExtra.drop_sparse" = FALSE)` - see \link{MatrixExtra-options}).
#'
#' \bold{Note:} Trying to slice a sparse matrix without supplying any parameter for the
#' second axis (e.g. `X[1:10]`) will \bold{select whole rows} (as if it were `X[1:10,]`)
#' instead of selecting entries as if the input were a flattened array (which is what
#' `Matrix` and base R do).
#'
#' This package will override the subsetting methods from `Matrix` for all
#' sparse matrix types. It is usually much faster for all three storage orders (especially
#' CSR) but in some situations could end up being slightly slower. Be aware that, in the
#' case of COO matrices (a.k.a. "TsparseMatrix"), the resulting object will \bold{not}
#' have sorted indices, which `Matrix` will oftentimes do in addition to subsetting,
#' at a large speed penalty.
#'
#' In general, it's much faster to select rows when the input is a CSR matrix ("RsparseMatrix"),
#' and much faster to select columns when the input is a CSC matrix ("CsparseMatrix").
#' Slicing COO matrices is typically not efficient, but could end up being faster when
#' the slice involves random rows and random columns with repeated entries.
#' @param x A sparse matrix to subset, in any format.
#' @param i row indices to subset.
#' @param j column indices to subset.
#' @param drop whether to simplify 1d matrix to a vector
#' @return A sparse matrix with the same storage order and dtype as `x`.
#' @name slice
#' @examples
#' library(Matrix)
#' library(MatrixExtra)
#' m <- rsparsematrix(20, 20, 0.1, repr="R")
#' inherits(m[1:2, ], "RsparseMatrix")
#' inherits(m[1:2, 3:4], "RsparseMatrix")
#' inherits(as.coo.matrix(m)[1:2, 3:4], "TsparseMatrix")
#' inherits(as.csc.matrix(m)[1:2, 3:4], "CsparseMatrix")
#'
#' ### New: slice with a sparse vector
#' m[as(c(TRUE,FALSE), "sparseVector"), ]
#'
#' ### Important!!!
#' ### This differs from Matrix
#' set_new_matrix_behavior()
#' inherits(m[1,,drop=TRUE], "sparseVector")
#'
#' ### To bring back the old behavior:
#' restore_old_matrix_behavior()
#' inherits(m[1,,drop=TRUE], "numeric")
NULL
subset_csr <- function(x, i, j, drop) {
check_valid_matrix(x)
if (missing(i) && missing(j)) {
x <- drop_slice(x, drop)
return(x)
}
if (!missing(i) && !missing(j) &&
NROW(i) == 1L && NROW(j) == 1L &&
typeof(i) %in% c("integer", "numeric") &&
typeof(j) %in% c("integer", "numeric")
) {
i <- as.integer(i)
j <- as.integer(j)
if (is.na(i) || is.na(j)) {
res <- NA_real_
} else if (i > nrow(x) || j > ncol(x)) {
stop("Subscript out of bounds.")
} else {
### TODO: add tests for these
if (inherits(x, "triangularMatrix") &&
((x@uplo == "U" && ij_is_in_lower_triangle(i, j)) ||
(x@uplo == "L" && ij_is_in_lower_triangle(j, i)))
) {
res <- 0
} else if (.hasSlot(x, "diag") && x@diag == "U" && i == j) {
res <- 1.
} else {
if (inherits(x, "symmetricMatrix")) {
if ((x@uplo == "U" && ij_is_in_lower_triangle(i, j)) ||
(x@uplo == "L" && ij_is_in_lower_triangle(j, i))
) {
temp <- i
i <- j
j <- temp
}
}
if (inherits(x, "dsparseMatrix")) {
res <- extract_single_val_csr_numeric(x@p, x@j, x@x, i-1L, j-1L)
} else if (inherits(x, "lsparseMatrix")) {
res <- as.logical(extract_single_val_csr_logical(x@p, x@j, x@x, i-1L, j-1L))
} else if (inherits(x, "nsparseMatrix")) {
res <- extract_single_val_csr_binary(x@p, x@j, i-1L, j-1L)
} else {
throw_internal_error()
}
}
}
if (missing(drop) || isTRUE(drop)) {
if (!is.null(colnames(x)))
names(res) <- colnames(x)[j]
return(res)
}
if (inherits(x, "dsparseMatrix")) {
out <- new("dgRMatrix")
} else if (inherits(x, "lsparseMatrix")) {
out <- new("lgRMatrix")
} else if (inherits(x, "nsparseMatrix")) {
out <- new("ngRMatrix")
} else {
throw_internal_error()
}
out@Dim <- c(1L, 1L)
if (!is.null(x@Dimnames[[1L]]))
out@Dimnames[[1L]] <- x@Dimnames[[1L]][i]
if (!is.null(x@Dimnames[[2L]]))
out@Dimnames[[2L]] <- x@Dimnames[[2L]][j]
if (res == 0) {
out@p <- c(0L, 0L)
return(out)
} else {
out@p <- c(0L, 1L)
out@j <- 0L
if (.hasSlot(out, "x")) {
if (inherits(out, "dsparseMatrix"))
out@x <- as.numeric(res)
else
out@x <- as.logical(res)
}
return(out)
}
}
### TODO: should also add paths for sequences with a gap (e.g. 3, 5, 7, 9...)
temp <- get_ij_properties(x, i, j)
i <- temp$i
j <- temp$j
all_i <- temp$all_i
all_j <- temp$all_j
i_is_seq <- temp$i_is_seq
j_is_seq <- temp$j_is_seq
i_is_rev_seq <- temp$i_is_rev_seq
j_is_rev_seq <- temp$j_is_rev_seq
i_has_NA <- temp$i_has_NA
j_has_NA <- temp$j_has_NA
i_all_NA <- temp$i_all_NA
j_all_NA <- temp$j_all_NA
i_NAs <- temp$i_NAs
j_NAs <- temp$j_NAs
n_row <- temp$n_row
n_col <- temp$n_col
row_names <- temp$row_names
col_names <- temp$col_names
if (!NROW(i) || !NROW(j) || length(x@j) == 0L) {
if (inherits(x, "dsparseMatrix")) {
res <- new("dgRMatrix")
} else if (inherits(x, "lsparseMatrix")) {
res <- new("lgRMatrix")
} else if (inherits(x, "nsparseMatrix")) {
res <- new("ngRMatrix")
} else {
throw_internal_error()
}
res@p <- integer(NROW(i) + 1L)
res@Dim <- c(NROW(i), NROW(j))
row_names <- if(is.null(row_names) || !NROW(row_names)) NULL else row_names[i]
col_names <- if(is.null(col_names) || !NROW(col_names)) NULL else col_names[j]
res@Dimnames <- list(row_names, col_names)
res <- drop_slice(res, drop)
return(res)
}
if (all_i && all_j) {
return(drop_slice(x, drop))
}
if ((i_all_NA || all_i) && (j_all_NA || all_j)) {
output_logical <- inherits(x, c("nsparseMatrix", "lsparseMatrix"))
out <- matrix(NA_real_, nrow=NROW(i), ncol=NROW(j))
out <- as.csr.matrix(out, logical=output_logical)
if (!i_all_NA && !is.null(rownames(x)))
rownames(out) <- rownames(x)[i]
if (!j_all_NA && !is.null(colnames(x)))
colnames(out) <- colnames(x)[j]
return(drop_slice(out, drop))
}
if (!all_i && !i_is_seq)
i_is_rev_seq <- check_is_rev_seq(i)
if (!all_j && !j_is_seq)
j_is_rev_seq <- check_is_rev_seq(j)
if (i_is_rev_seq && j_is_rev_seq && length(i) == nrow(x) && length(j) == ncol(x)) {
if (inherits(x, "dsparseMatrix")) {
temp <- reverse_rows_numeric(x@p, x@j, x@x)
reverse_columns_inplace_numeric(temp$indptr, temp$indices, temp$values, ncol(x))
} else if (inherits(x, "lsparseMatrix")) {
temp <- reverse_rows_logical(x@p, x@j, x@x)
reverse_columns_inplace_logical(temp$indptr, temp$indices, temp$values, ncol(x))
} else if (inherits(x, "nsparseMatrix")) {
temp <- reverse_rows_binary(x@p, x@j)
reverse_columns_inplace_binary(temp$indptr, temp$indices, ncol(x))
} else {
throw_internal_error()
}
X_attr <- attributes(x)
X_attr$p <- temp$indptr
X_attr$j <- temp$indices
if ("x" %in% names(X_attr))
X_attr$x <- temp$values
if (!is.null(X_attr$Dimnames[[1L]]))
X_attr$Dimnames[[1L]] <- rev(X_attr$Dimnames[[1L]])
if (!is.null(X_attr$Dimnames[[2L]]))
X_attr$Dimnames[[2L]] <- rev(X_attr$Dimnames[[2L]])
if ("uplo" %in% names(X_attr))
X_attr$uplo = ifelse(X_attr$uplo == "U", "L", "U")
attributes(x) <- X_attr
return(drop_slice(x, drop, i_NAs, j_NAs))
}
if (inherits(x, c("symmetricMatrix", "triangularMatrix")) && !(length(x@j) == 0L))
x <- as.csr.matrix(x, logical=inherits(x, "lsparseMatrix"), binary=inherits(x, "nsparseMatrix"))
has_x <- .hasSlot(x, "x")
if (i_is_seq && all_j) {
first <- x@p[i[1L]] + 1L
last <- x@p[i[n_row] + 1L]
indptr <- x@p[seq(i[1L], i[n_row]+1L)] - x@p[i[1L]]
col_indices <- x@j[first:last]
if (has_x)
x_values <- x@x[first:last]
} else if (i_is_rev_seq && all_j) {
first <- x@p[i[n_row]] + 1L
last <- x@p[i[1L] + 1L]
indptr <- x@p[seq(i[n_row], i[1L]+1L)] - x@p[i[n_row]]
col_indices <- x@j[first:last]
if (has_x)
x_values <- x@x[first:last]
if (!has_x) {
temp <- reverse_rows_binary(indptr, col_indices)
} else if (typeof(x_values) == "logical") {
temp <- reverse_rows_logical(indptr, col_indices, x_values)
} else {
temp <- reverse_rows_numeric(indptr, col_indices, as.numeric(x_values))
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else if (!i_is_seq && all_j) {
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_rows_numeric(x@p, x@j, x@x, i-1L)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_rows_logical(x@p, x@j, x@x, i-1L)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_rows_binary(x@p, x@j, i-1L)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else if (j_is_seq) {
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_rows_col_seq_numeric(x@p, x@j, x@x, i-1L, j, TRUE)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_rows_col_seq_logical(x@p, x@j, x@x, i-1L, j, TRUE)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_rows_col_seq_binary(x@p, x@j, i-1L, j, TRUE)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else if (j_is_rev_seq) {
new_ncol <- j[1L] - j[length(j)] + 1L
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_rows_col_seq_numeric(x@p, x@j, x@x, i-1L, j, TRUE)
reverse_columns_inplace_numeric(temp$indptr, temp$indices, temp$values, new_ncol)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_rows_col_seq_logical(x@p, x@j, x@x, i-1L, j, TRUE)
reverse_columns_inplace_logical(temp$indptr, temp$indices, temp$values, new_ncol)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_rows_col_seq_binary(x@p, x@j, i-1L, j, TRUE)
reverse_columns_inplace_binary(temp$indptr, temp$indices, new_ncol)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else {
### FAIL: this ended up being much slower than the original Matrix
### when the columns don't get pre-discarded through the
### condition min(j) <= col < max(j)
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_arbitrary_numeric(x@p, x@j, x@x, i-1L, j-1L)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_arbitrary_logical(x@p, x@j, x@x, i-1L, j-1L)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_arbitrary_binary(x@p, x@j, i-1L, j-1L)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
}
res <- new(class(x)[1L])
res@p <- indptr
res@j <- col_indices
if (has_x) {
if (inherits(x, "lsparseMatrix")) {
res@x <- as.logical(x_values)
} else {
res@x <- x_values
}
}
res@Dim <- c(n_row, n_col)
row_names <- if (is.null(row_names) || !NROW(row_names)) NULL else row_names[i]
col_names <- if (is.null(col_names) || !NCOL(col_names)) NULL else col_names[j]
res@Dimnames <- list(row_names, col_names)
res <- drop_slice(res, drop, i_NAs, j_NAs)
return(res)
}
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="index", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="index", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="missing", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="missing", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="index", drop="missing"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="index", drop="missing"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="missing", drop="missing"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="missing", drop="missing"), subset_csr)
subset_generic_with_vector <- function(x, i, j, drop) {
if (inherits(x, "CsparseMatrix"))
return(t_shallow(subset_csr(t_shallow(x), j, i, drop)))
if (inherits(x, "RsparseMatrix"))
return(subset_csr(x, i, j, drop))
if (missing(i) && missing(j))
return(drop_slice(x, drop))
if (!missing(i))
i <- get_indices_integer(i, NROW(x), rownames(x))
if (!missing(j))
j <- get_indices_integer(j, NCOL(x), colnames(x))
if (!is.null(dim(x))) {
if (!missing(i) && !missing(j)) {
return(x[i, j, drop=drop])
} else if (!missing(i)) {
return(x[i, , drop=drop])
} else {
return(x[, j, drop=drop])
}
} else {
if (missing(i))
return(x)
return(x[i])
}
}
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="nsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="nsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="missing", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="nsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="index", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="nsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="nsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="missing", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="nsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="index", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="lsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="lsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="missing", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="lsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="index", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="lsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="lsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="missing", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="lsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="index", drop="missing"), subset_generic_with_vector)
### The method here is usually faster than CSC slicing from Matrix, so it will replace it too
subset_csc_masked <- function(x, i, j, drop) {
x <- subset_csr(t_shallow(x), j, i, drop)
if (inherits(x, "sparseMatrix")) {
x <- t_shallow(x)
} else if (is.matrix(x)) {
x <- t(x)
}
return(x)
}
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="index", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="index", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="missing", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="missing", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="index", drop="missing"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="index", drop="missing"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="missing", drop="missing"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="missing", drop="missing"), subset_csc_masked)
Try the MatrixExtra package in your browser
Any scripts or data that you put into this service are public.
MatrixExtra documentation built on Aug. 21, 2023, 1:08 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions subset_csc_masked subset_generic_with_vector subset_csr drop_slice inject_NAs get_ij_properties ij_is_in_lower_triangle get_indices_integer
get_indices_integer <- function(i, max_i, index_names) {
if (inherits(i, c("lsparseVector", "nsparseVector"))) {
i <- sort_sparse_indices(i, copy=!getOption("MatrixExtra.inplace_sort", default=FALSE))
}
if (inherits(i, "lsparseVector")) {
if (anyNA(i@x))
i <- seq(1L, max_i)[as.logical(i)]
else
i <- as(i, "nsparseVector")
}
if (inherits(i, "nsparseVector")) {
if (i@length > max_i) {
stop("Dimension of indexing vector is larger than matrix to subset.")
} else if (i@length == max_i) {
i <- as.integer(i@i)
} else { ### mimic of R base's recycling
full_repeats <- as.integer(max_i %/% i@length)
remainder <- as.integer(max_i - i@length*full_repeats)
if (remainder != 0)
i <- repeat_indices_n_times(as.integer(i@i), as.integer(i[seq(1L, remainder)]@i), as.integer(i@length), max_i)
else
i <- repeat_indices_n_times(as.integer(i@i), integer(), as.integer(i@length), max_i)
}
if (typeof(i) != "integer")
i <- as.integer(i)
}
if (is.character(i)) {
if (!is.null(index_names))
i <- match(i, index_names)
else
i <- as.integer(i)
}
if (is.logical(i)) {
if (length(i) != max_i) {
i <- seq(1L, max_i)[i]
} else {
i <- which(i)
}
}
if (typeof(i) != "integer")
i <- as.integer(i)
if (any(!is.na(i) & i <= 0))
i <- seq(1L, max_i)[i]
if(NROW(i) && any(i > max_i, na.rm=TRUE))
stop("some of row subset indices are not present in matrix")
return(as.integer(i))
}
ij_is_in_lower_triangle <- function(i, j) {
return(i > j)
}
get_ij_properties <- function(x, i, j) {
row_names <- rownames(x)
col_names <- colnames(x)
all_i <- FALSE
all_j <- FALSE
i_is_seq <- FALSE
j_is_seq <- FALSE
i_is_rev_seq <- FALSE
j_is_rev_seq <- FALSE
i_has_NA <- FALSE
j_has_NA <- FALSE
if (missing(j)) {
all_j <- TRUE
j <- seq_len(ncol(x))
n_col <- ncol(x)
} else {
j <- get_indices_integer(j, ncol(x), col_names)
j_has_NA <- anyNA(j)
if (!j_has_NA) {
if (length(j) == ncol(x) && j[1L] == 1L && j[length(j)] == ncol(x)) {
if (check_is_seq(j))
all_j <- TRUE
} else {
j_is_seq <- check_is_seq(j)
}
}
n_col <- length(j)
}
if (missing(i)) {
i <- seq_len(nrow(x))
all_i <- TRUE
i_is_seq <- TRUE
n_row <- nrow(x)
} else {
i <- get_indices_integer(i, nrow(x), row_names)
i_has_NA <- anyNA(i)
if (!i_has_NA)
i_is_seq <- check_is_seq(i)
n_row <- length(i)
if (i_is_seq && length(i) == nrow(x) && i[1L] == 1L && i[length(i)] == nrow(x)) {
if (check_is_seq(i))
all_i <- TRUE
}
}
if (!all_i && !i_is_seq && !i_has_NA)
i_is_rev_seq <- check_is_rev_seq(i)
if (!all_j && !j_is_seq && !j_has_NA)
j_is_rev_seq <- check_is_rev_seq(j)
i_all_NA <- FALSE
j_all_NA <- FALSE
i_NAs <- NULL
j_NAs <- NULL
if (i_has_NA) {
i_temp <- find_first_non_na(i)
if (is.na(i_temp)) {
i_all_NA <- TRUE
} else {
i_NAs <- which(is.na(i))
i[i_NAs] <- i_temp
}
}
if (j_has_NA) {
j_temp <- find_first_non_na(j)
if (is.na(j_temp)) {
j_all_NA <- TRUE
} else {
j_NAs <- which(is.na(j))
j[j_NAs] <- j_temp
}
}
return(list(
i = i, all_i = all_i, i_is_seq = i_is_seq, i_is_rev_seq = i_is_rev_seq,
j = j, all_j = all_j, j_is_seq = j_is_seq, j_is_rev_seq = j_is_rev_seq,
i_has_NA = i_has_NA, i_all_NA = i_all_NA, i_NAs = i_NAs,
j_has_NA = j_has_NA, j_all_NA = j_all_NA, j_NAs = j_NAs,
row_names = row_names, col_names = col_names,
n_row = n_row, n_col = n_col
))
}
inject_NAs <- function(x, i_NAs, j_NAs) {
if (NROW(i_NAs) || NROW(j_NAs)) {
if (inherits(x, "sparseMatrix")) {
if (NROW(i_NAs) && !is.null(x@Dimnames[[1L]]))
x@Dimnames[[1L]][i_NAs] <- NA_character_
if (NROW(j_NAs) && !is.null(x@Dimnames[[2L]]))
x@Dimnames[[2L]][j_NAs] <- NA_character_
} else {
if (NROW(i_NAs) && !is.null(rownames(x)))
rownames(x)[i_NAs] <- NA
if (NROW(j_NAs) && !is.null(rownames(x)))
rownames(x)[j_NAs] <- NA
}
output_logical <- inherits(x, c("nsparseMatrix", "lsparseMatrix"))
if (inherits(x, "RsparseMatrix")) {
x <- as.csr.matrix(x, logical=output_logical)
} else if (inherits(x, "TsparseMatrix")) {
x <- as.coo.matrix(x, logical=output_logical)
}
if (inherits(x, "RsparseMatrix")) {
x <- as.csr.matrix(x)
if (NROW(i_NAs))
x <- assign_csr_internal(x, i_NAs, seq(1L, NCOL(x)), NA_real_)
if (NROW(j_NAs))
x <- assign_csr_internal(x, seq(1L, NROW(x)), j_NAs, NA_real_)
if (output_logical)
x <- as.csr.matrix(x, logical=TRUE)
}
else if (inherits(x, "TsparseMatrix")) {
if (!NROW(i_NAs))
i_NAs <- integer()
if (!NROW(j_NAs))
j_NAs <- integer()
if (output_logical) {
out <- new("lgTMatrix")
res <- inject_NAs_inplace_coo_logical(x@i, x@j, x@x, i_NAs-1L, j_NAs-1L, nrow(x), ncol(x))
} else {
out <- new("dgTMatrix")
res <- inject_NAs_inplace_coo_numeric(x@i, x@j, x@x, i_NAs-1L, j_NAs-1L, nrow(x), ncol(x))
}
out@i <- res$ii
out@j <- res$jj
out@x <- res$xx
out@Dim <- x@Dim
x <- out
}
else if (inherits(x, "float32")) {
x[i_NAs, ] <- float::fl(NA_real_)
x[, j_NAs] <- float::fl(NA_real_)
}
else {
x[i_NAs, ] <- NA
x[, j_NAs] <- NA
}
}
return(x)
}
drop_slice <- function(x, drop, i_NAs=NULL, j_NAs=NULL) {
x <- inject_NAs(x, i_NAs, j_NAs)
if ((missing(drop) || isTRUE(drop)) && inherits(x, c("sparseMatrix", "float32"))) {
if (nrow(x) == 1L || ncol(x) == 1L) {
if (!(nrow(x) == 1L && ncol(x) == 1L) && !inherits(x, "float32") &&
getOption("MatrixExtra.drop_sparse", default=FALSE)
) {
x <- as(x, "sparseVector")
} else {
x <- as.vector(x)
}
}
}
if (inherits(x, "sparseMatrix")) {
Dimnames <- list(NULL, NULL)
if (NROW(x@Dimnames[[1L]]))
Dimnames[[1L]] <- x@Dimnames[[1L]]
if (NROW(x@Dimnames[[2L]]))
Dimnames[[2L]] <- x@Dimnames[[2L]]
x@Dimnames <- Dimnames
}
return(x)
}
#' @title Sparse Matrices Slicing
#' @description Natively slice CSR/CSC/COO matrices without changing the storage order.
#' @details \bold{Important:} When slicing sparse matrices with `drop=TRUE` (the default),
#' `Matrix` will drop 1-d matrices to \bold{dense} dense vectors, whereas
#' this package allows dropping them to either dense or \bold{sparse} vectors,
#' the latter of which is more efficient and is the default option.
#'
#' The `drop` behavior can be changed back to dense vectors like `Matrix` does,
#' through \link{restore_old_matrix_behavior} or through the package options
#' (e.g. `options("MatrixExtra.drop_sparse" = FALSE)` - see \link{MatrixExtra-options}).
#'
#' \bold{Note:} Trying to slice a sparse matrix without supplying any parameter for the
#' second axis (e.g. `X[1:10]`) will \bold{select whole rows} (as if it were `X[1:10,]`)
#' instead of selecting entries as if the input were a flattened array (which is what
#' `Matrix` and base R do).
#'
#' This package will override the subsetting methods from `Matrix` for all
#' sparse matrix types. It is usually much faster for all three storage orders (especially
#' CSR) but in some situations could end up being slightly slower. Be aware that, in the
#' case of COO matrices (a.k.a. "TsparseMatrix"), the resulting object will \bold{not}
#' have sorted indices, which `Matrix` will oftentimes do in addition to subsetting,
#' at a large speed penalty.
#'
#' In general, it's much faster to select rows when the input is a CSR matrix ("RsparseMatrix"),
#' and much faster to select columns when the input is a CSC matrix ("CsparseMatrix").
#' Slicing COO matrices is typically not efficient, but could end up being faster when
#' the slice involves random rows and random columns with repeated entries.
#' @param x A sparse matrix to subset, in any format.
#' @param i row indices to subset.
#' @param j column indices to subset.
#' @param drop whether to simplify 1d matrix to a vector
#' @return A sparse matrix with the same storage order and dtype as `x`.
#' @name slice
#' @examples
#' library(Matrix)
#' library(MatrixExtra)
#' m <- rsparsematrix(20, 20, 0.1, repr="R")
#' inherits(m[1:2, ], "RsparseMatrix")
#' inherits(m[1:2, 3:4], "RsparseMatrix")
#' inherits(as.coo.matrix(m)[1:2, 3:4], "TsparseMatrix")
#' inherits(as.csc.matrix(m)[1:2, 3:4], "CsparseMatrix")
#'
#' ### New: slice with a sparse vector
#' m[as(c(TRUE,FALSE), "sparseVector"), ]
#'
#' ### Important!!!
#' ### This differs from Matrix
#' set_new_matrix_behavior()
#' inherits(m[1,,drop=TRUE], "sparseVector")
#'
#' ### To bring back the old behavior:
#' restore_old_matrix_behavior()
#' inherits(m[1,,drop=TRUE], "numeric")
NULL
subset_csr <- function(x, i, j, drop) {
check_valid_matrix(x)
if (missing(i) && missing(j)) {
x <- drop_slice(x, drop)
return(x)
}
if (!missing(i) && !missing(j) &&
NROW(i) == 1L && NROW(j) == 1L &&
typeof(i) %in% c("integer", "numeric") &&
typeof(j) %in% c("integer", "numeric")
) {
i <- as.integer(i)
j <- as.integer(j)
if (is.na(i) || is.na(j)) {
res <- NA_real_
} else if (i > nrow(x) || j > ncol(x)) {
stop("Subscript out of bounds.")
} else {
### TODO: add tests for these
if (inherits(x, "triangularMatrix") &&
((x@uplo == "U" && ij_is_in_lower_triangle(i, j)) ||
(x@uplo == "L" && ij_is_in_lower_triangle(j, i)))
) {
res <- 0
} else if (.hasSlot(x, "diag") && x@diag == "U" && i == j) {
res <- 1.
} else {
if (inherits(x, "symmetricMatrix")) {
if ((x@uplo == "U" && ij_is_in_lower_triangle(i, j)) ||
(x@uplo == "L" && ij_is_in_lower_triangle(j, i))
) {
temp <- i
i <- j
j <- temp
}
}
if (inherits(x, "dsparseMatrix")) {
res <- extract_single_val_csr_numeric(x@p, x@j, x@x, i-1L, j-1L)
} else if (inherits(x, "lsparseMatrix")) {
res <- as.logical(extract_single_val_csr_logical(x@p, x@j, x@x, i-1L, j-1L))
} else if (inherits(x, "nsparseMatrix")) {
res <- extract_single_val_csr_binary(x@p, x@j, i-1L, j-1L)
} else {
throw_internal_error()
}
}
}
if (missing(drop) || isTRUE(drop)) {
if (!is.null(colnames(x)))
names(res) <- colnames(x)[j]
return(res)
}
if (inherits(x, "dsparseMatrix")) {
out <- new("dgRMatrix")
} else if (inherits(x, "lsparseMatrix")) {
out <- new("lgRMatrix")
} else if (inherits(x, "nsparseMatrix")) {
out <- new("ngRMatrix")
} else {
throw_internal_error()
}
out@Dim <- c(1L, 1L)
if (!is.null(x@Dimnames[[1L]]))
out@Dimnames[[1L]] <- x@Dimnames[[1L]][i]
if (!is.null(x@Dimnames[[2L]]))
out@Dimnames[[2L]] <- x@Dimnames[[2L]][j]
if (res == 0) {
out@p <- c(0L, 0L)
return(out)
} else {
out@p <- c(0L, 1L)
out@j <- 0L
if (.hasSlot(out, "x")) {
if (inherits(out, "dsparseMatrix"))
out@x <- as.numeric(res)
else
out@x <- as.logical(res)
}
return(out)
}
}
### TODO: should also add paths for sequences with a gap (e.g. 3, 5, 7, 9...)
temp <- get_ij_properties(x, i, j)
i <- temp$i
j <- temp$j
all_i <- temp$all_i
all_j <- temp$all_j
i_is_seq <- temp$i_is_seq
j_is_seq <- temp$j_is_seq
i_is_rev_seq <- temp$i_is_rev_seq
j_is_rev_seq <- temp$j_is_rev_seq
i_has_NA <- temp$i_has_NA
j_has_NA <- temp$j_has_NA
i_all_NA <- temp$i_all_NA
j_all_NA <- temp$j_all_NA
i_NAs <- temp$i_NAs
j_NAs <- temp$j_NAs
n_row <- temp$n_row
n_col <- temp$n_col
row_names <- temp$row_names
col_names <- temp$col_names
if (!NROW(i) || !NROW(j) || length(x@j) == 0L) {
if (inherits(x, "dsparseMatrix")) {
res <- new("dgRMatrix")
} else if (inherits(x, "lsparseMatrix")) {
res <- new("lgRMatrix")
} else if (inherits(x, "nsparseMatrix")) {
res <- new("ngRMatrix")
} else {
throw_internal_error()
}
res@p <- integer(NROW(i) + 1L)
res@Dim <- c(NROW(i), NROW(j))
row_names <- if(is.null(row_names) || !NROW(row_names)) NULL else row_names[i]
col_names <- if(is.null(col_names) || !NROW(col_names)) NULL else col_names[j]
res@Dimnames <- list(row_names, col_names)
res <- drop_slice(res, drop)
return(res)
}
if (all_i && all_j) {
return(drop_slice(x, drop))
}
if ((i_all_NA || all_i) && (j_all_NA || all_j)) {
output_logical <- inherits(x, c("nsparseMatrix", "lsparseMatrix"))
out <- matrix(NA_real_, nrow=NROW(i), ncol=NROW(j))
out <- as.csr.matrix(out, logical=output_logical)
if (!i_all_NA && !is.null(rownames(x)))
rownames(out) <- rownames(x)[i]
if (!j_all_NA && !is.null(colnames(x)))
colnames(out) <- colnames(x)[j]
return(drop_slice(out, drop))
}
if (!all_i && !i_is_seq)
i_is_rev_seq <- check_is_rev_seq(i)
if (!all_j && !j_is_seq)
j_is_rev_seq <- check_is_rev_seq(j)
if (i_is_rev_seq && j_is_rev_seq && length(i) == nrow(x) && length(j) == ncol(x)) {
if (inherits(x, "dsparseMatrix")) {
temp <- reverse_rows_numeric(x@p, x@j, x@x)
reverse_columns_inplace_numeric(temp$indptr, temp$indices, temp$values, ncol(x))
} else if (inherits(x, "lsparseMatrix")) {
temp <- reverse_rows_logical(x@p, x@j, x@x)
reverse_columns_inplace_logical(temp$indptr, temp$indices, temp$values, ncol(x))
} else if (inherits(x, "nsparseMatrix")) {
temp <- reverse_rows_binary(x@p, x@j)
reverse_columns_inplace_binary(temp$indptr, temp$indices, ncol(x))
} else {
throw_internal_error()
}
X_attr <- attributes(x)
X_attr$p <- temp$indptr
X_attr$j <- temp$indices
if ("x" %in% names(X_attr))
X_attr$x <- temp$values
if (!is.null(X_attr$Dimnames[[1L]]))
X_attr$Dimnames[[1L]] <- rev(X_attr$Dimnames[[1L]])
if (!is.null(X_attr$Dimnames[[2L]]))
X_attr$Dimnames[[2L]] <- rev(X_attr$Dimnames[[2L]])
if ("uplo" %in% names(X_attr))
X_attr$uplo = ifelse(X_attr$uplo == "U", "L", "U")
attributes(x) <- X_attr
return(drop_slice(x, drop, i_NAs, j_NAs))
}
if (inherits(x, c("symmetricMatrix", "triangularMatrix")) && !(length(x@j) == 0L))
x <- as.csr.matrix(x, logical=inherits(x, "lsparseMatrix"), binary=inherits(x, "nsparseMatrix"))
has_x <- .hasSlot(x, "x")
if (i_is_seq && all_j) {
first <- x@p[i[1L]] + 1L
last <- x@p[i[n_row] + 1L]
indptr <- x@p[seq(i[1L], i[n_row]+1L)] - x@p[i[1L]]
col_indices <- x@j[first:last]
if (has_x)
x_values <- x@x[first:last]
} else if (i_is_rev_seq && all_j) {
first <- x@p[i[n_row]] + 1L
last <- x@p[i[1L] + 1L]
indptr <- x@p[seq(i[n_row], i[1L]+1L)] - x@p[i[n_row]]
col_indices <- x@j[first:last]
if (has_x)
x_values <- x@x[first:last]
if (!has_x) {
temp <- reverse_rows_binary(indptr, col_indices)
} else if (typeof(x_values) == "logical") {
temp <- reverse_rows_logical(indptr, col_indices, x_values)
} else {
temp <- reverse_rows_numeric(indptr, col_indices, as.numeric(x_values))
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else if (!i_is_seq && all_j) {
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_rows_numeric(x@p, x@j, x@x, i-1L)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_rows_logical(x@p, x@j, x@x, i-1L)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_rows_binary(x@p, x@j, i-1L)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else if (j_is_seq) {
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_rows_col_seq_numeric(x@p, x@j, x@x, i-1L, j, TRUE)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_rows_col_seq_logical(x@p, x@j, x@x, i-1L, j, TRUE)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_rows_col_seq_binary(x@p, x@j, i-1L, j, TRUE)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else if (j_is_rev_seq) {
new_ncol <- j[1L] - j[length(j)] + 1L
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_rows_col_seq_numeric(x@p, x@j, x@x, i-1L, j, TRUE)
reverse_columns_inplace_numeric(temp$indptr, temp$indices, temp$values, new_ncol)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_rows_col_seq_logical(x@p, x@j, x@x, i-1L, j, TRUE)
reverse_columns_inplace_logical(temp$indptr, temp$indices, temp$values, new_ncol)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_rows_col_seq_binary(x@p, x@j, i-1L, j, TRUE)
reverse_columns_inplace_binary(temp$indptr, temp$indices, new_ncol)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
} else {
### FAIL: this ended up being much slower than the original Matrix
### when the columns don't get pre-discarded through the
### condition min(j) <= col < max(j)
if (inherits(x, "dsparseMatrix")) {
temp <- copy_csr_arbitrary_numeric(x@p, x@j, x@x, i-1L, j-1L)
} else if (inherits(x, "lsparseMatrix")) {
temp <- copy_csr_arbitrary_logical(x@p, x@j, x@x, i-1L, j-1L)
} else if (inherits(x, "nsparseMatrix")) {
temp <- copy_csr_arbitrary_binary(x@p, x@j, i-1L, j-1L)
} else {
throw_internal_error()
}
indptr <- temp$indptr
col_indices <- temp$indices
if (has_x)
x_values <- temp$values
}
res <- new(class(x)[1L])
res@p <- indptr
res@j <- col_indices
if (has_x) {
if (inherits(x, "lsparseMatrix")) {
res@x <- as.logical(x_values)
} else {
res@x <- x_values
}
}
res@Dim <- c(n_row, n_col)
row_names <- if (is.null(row_names) || !NROW(row_names)) NULL else row_names[i]
col_names <- if (is.null(col_names) || !NCOL(col_names)) NULL else col_names[j]
res@Dimnames <- list(row_names, col_names)
res <- drop_slice(res, drop, i_NAs, j_NAs)
return(res)
}
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="index", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="index", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="missing", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="missing", drop="logical"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="index", drop="missing"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="index", drop="missing"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="index", j="missing", drop="missing"), subset_csr)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="RsparseMatrix", i="missing", j="missing", drop="missing"), subset_csr)
subset_generic_with_vector <- function(x, i, j, drop) {
if (inherits(x, "CsparseMatrix"))
return(t_shallow(subset_csr(t_shallow(x), j, i, drop)))
if (inherits(x, "RsparseMatrix"))
return(subset_csr(x, i, j, drop))
if (missing(i) && missing(j))
return(drop_slice(x, drop))
if (!missing(i))
i <- get_indices_integer(i, NROW(x), rownames(x))
if (!missing(j))
j <- get_indices_integer(j, NCOL(x), colnames(x))
if (!is.null(dim(x))) {
if (!missing(i) && !missing(j)) {
return(x[i, j, drop=drop])
} else if (!missing(i)) {
return(x[i, , drop=drop])
} else {
return(x[, j, drop=drop])
}
} else {
if (missing(i))
return(x)
return(x[i])
}
}
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="nsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="nsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="missing", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="nsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="index", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="nsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="nsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="missing", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="nsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="nsparseVector", j="index", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="lsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="lsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="missing", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="lsparseVector", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="index", drop="logical"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="lsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="missing", j="lsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="missing", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="index", j="lsparseVector", drop="missing"), subset_generic_with_vector)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="ANY", i="lsparseVector", j="index", drop="missing"), subset_generic_with_vector)
### The method here is usually faster than CSC slicing from Matrix, so it will replace it too
subset_csc_masked <- function(x, i, j, drop) {
x <- subset_csr(t_shallow(x), j, i, drop)
if (inherits(x, "sparseMatrix")) {
x <- t_shallow(x)
} else if (is.matrix(x)) {
x <- t(x)
}
return(x)
}
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="index", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="index", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="missing", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="missing", drop="logical"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="index", drop="missing"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="index", drop="missing"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="index", j="missing", drop="missing"), subset_csc_masked)
#' @rdname slice
#' @export
setMethod(`[`, signature(x="CsparseMatrix", i="missing", j="missing", drop="missing"), subset_csc_masked)
Try the MatrixExtra package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.