gdsfmt (version 1.8.3)

clusterApply.gdsn: Apply functions over matrix margins in parallel

Description

Return a vector or list of values obtained by applying a function to margins of a GDS matrix in parallel.

Usage

clusterApply.gdsn(cl, gds.fn, node.name, margin, FUN, selection=NULL, as.is=c("list", "none", "integer", "double", "character", "logical", "raw"), var.index=c("none", "relative", "absolute"), .useraw=FALSE, .value=NULL, .substitute=NULL, ...)

Arguments

cl

a cluster object, created by this package or by the package parallel

gds.fn

the file name of a GDS file

node.name

a character vector indicating GDS node path

margin

an integer giving the subscripts which the function will be applied over. E.g., for a matrix 1 indicates rows, 2 indicates columns

FUN

the function to be applied

selection

a list or NULL; if a list, it is a list of logical vectors according to dimensions indicating selection; if NULL, uses all data

as.is

returned value: a list, an integer vector, etc

var.index

if "none", call FUN(x, ...) without an index; if "relative" or "absolute", add an argument to the user-defined function FUN like FUN(index, x, ...) where index in the function is an index starting from 1: "relative" for indexing in the selection defined by selection, "absolute" for indexing with respect to all data

.useraw

use R RAW storage mode if integers can be stored in a byte, to reduce memory usage

.value

a vector of values to be replaced in the original data array, or NULL for nothing

.substitute

a vector of values after replacing, or NULL for nothing; length(.substitute) should be one or length(.value); if length(.substitute) = length(.value), it is a mapping from .value to .substitute

...

optional arguments to FUN

Value

A vector or list of values.

Details

The algorithm of applying is optimized by blocking the computations to exploit the high-speed memory instead of disk.

References

http://github.com/zhengxwen/gdsfmt

See Also

apply.gdsn

Examples

###########################################################
# prepare a GDS file

# cteate a GDS file
f <- createfn.gds("test1.gds")

(n <- add.gdsn(f, "matrix", val=matrix(1:(10*6), nrow=10)))
read.gdsn(index.gdsn(f, "matrix"))

closefn.gds(f)


# cteate the GDS file "test2.gds"
(f <- createfn.gds("test2.gds"))

X <- matrix(1:50, nrow=10)
Y <- matrix((1:50)/100, nrow=10)
Z1 <- factor(c(rep(c("ABC", "DEF", "ETD"), 3), "TTT"))
Z2 <- c(TRUE, FALSE, TRUE, FALSE, TRUE)

node.X <- add.gdsn(f, "X", X)
node.Y <- add.gdsn(f, "Y", Y)
node.Z1 <- add.gdsn(f, "Z1", Z1)
node.Z2 <- add.gdsn(f, "Z2", Z2)
f

closefn.gds(f)



###########################################################
# apply in parallel

library(parallel)

# Use option cl.core to choose an appropriate cluster size.
cl <- makeCluster(getOption("cl.cores", 2L))


# Apply functions over rows or columns of matrix

clusterApply.gdsn(cl, "test1.gds", "matrix", margin=1, FUN=function(x) x)

clusterApply.gdsn(cl, "test1.gds", "matrix", margin=2, FUN=function(x) x)

clusterApply.gdsn(cl, "test1.gds", "matrix", margin=1,
    selection = list(rep(c(TRUE, FALSE), 5), rep(c(TRUE, FALSE), 3)),
    FUN=function(x) x)

clusterApply.gdsn(cl, "test1.gds", "matrix", margin=2,
    selection = list(rep(c(TRUE, FALSE), 5), rep(c(TRUE, FALSE), 3)),
    FUN=function(x) x)



# Apply functions over rows or columns of multiple data sets

clusterApply.gdsn(cl, "test2.gds", c("X", "Y", "Z1"), margin=c(1, 1, 1),
    FUN=function(x) x)

# with variable names
clusterApply.gdsn(cl, "test2.gds", c(X="X", Y="Y", Z="Z2"), margin=c(2, 2, 1),
    FUN=function(x) x)


# stop clusters
stopCluster(cl)


# delete the temporary file
unlink(c("test1.gds", "test2.gds"), force=TRUE)