writeHDF5Array: Write an array-like object to an HDF5 file

View source: R/writeHDF5Array.R

writeHDF5ArrayR Documentation

Write an array-like object to an HDF5 file

Description

A function for writing an array-like object to an HDF5 file.

Usage

writeHDF5Array(x, filepath=NULL, name=NULL,
                  H5type=NULL, chunkdim=NULL, level=NULL, as.sparse=NA,
                  with.dimnames=FALSE, verbose=NA)

Arguments

x

The array-like object to write to an HDF5 file.

If x is a DelayedArray object, writeHDF5Array realizes it on disk, that is, all the delayed operations carried by the object are executed while the object is written to disk. See "On-disk realization of a DelayedArray object as an HDF5 dataset" section below for more information.

filepath

NULL or the path (as a single string) to the (new or existing) HDF5 file where to write the dataset. If NULL, then the dataset will be written to the current HDF5 dump file i.e. to the file whose path is getHDF5DumpFile.

name

NULL or the name of the HDF5 dataset to write. If NULL, then the name returned by getHDF5DumpName will be used.

H5type

The H5 datatype to use for the HDF5 dataset to be written to the HDF5 file is automatically inferred from the type of x (type(x)). Advanced users can override this by specifying the H5 datatype they want via the H5type argument.

See rhdf5::h5const("H5T") for a list of available H5 datatypes. See References section below for the link to the HDF Group's Support Portal where H5 predefined datatypes are documented.

A typical use case is to use a datatype that is smaller than the automatic one in order to reduce the size of the dataset on disk. For example you could use "H5T_IEEE_F32LE" when type(x) is "double" and you don't care about preserving the precision of 64-bit floating-point numbers (the automatic H5 datatype used for "double" is "H5T_IEEE_F64LE"). Another example is to use "H5T_STD_U16LE" when x contains small non-negative integer values like counts (the automatic H5 datatype used for "integer" is "H5T_STD_I32LE").

chunkdim

The dimensions of the chunks to use for writing the data to disk. By default (i.e. when chunkdim is set to NULL), getHDF5DumpChunkDim(dim(x)) will be used. See ?getHDF5DumpChunkDim for more information.

Set chunkdim to 0 to write unchunked data (a.k.a. contiguous data).

level

The compression level to use for writing the data to disk. By default, getHDF5DumpCompressionLevel() will be used. See ?getHDF5DumpCompressionLevel for more information.

as.sparse

Whether the data in the returned HDF5Array object should be flagged as sparse or not. If set to NA (the default), then is_sparse(x) is used.

IMPORTANT NOTE: This only controls the as.sparse flag of the returned HDF5Array object. See man page of the HDF5Array() constructor for more information. In particular this does NOT affect how the data will be laid out to the HDF5 file in any way (HDF5 doesn't natively support sparse storage at the moment). In other words, the data will always be stored in a dense format, even when as.sparse is set to TRUE.

with.dimnames

By default the dimnames on x are not written to the HDF5 file. Set with.dimnames to TRUE to also have them written.

Note that h5writeDimnames is used internally to write the dimnames to disk. Setting with.dimnames to FALSE and calling h5writeDimnames is another way to write the dimnames on x to disk that gives more control. See ?h5writeDimnames for more information.

verbose

Whether block processing progress should be displayed or not. If set to NA (the default), verbosity is controlled by DelayedArray:::get_verbose_block_processing(). Setting verbose to TRUE or FALSE overrides this.

Details

Please note that, depending on the size of the data to write to disk and the performance of the disk, writeHDF5Array() can take a long time to complete. Use verbose=TRUE to see its progress.

Use setHDF5DumpFile and setHDF5DumpName to control the location of automatically created HDF5 datasets.

Use setHDF5DumpChunkLength, setHDF5DumpChunkShape, and setHDF5DumpCompressionLevel, to control the physical properties of automatically created HDF5 datasets.

Value

An HDF5Array object pointing to the newly written HDF5 dataset on disk.

On-disk realization of a DelayedArray object as an HDF5 dataset

When passed a DelayedArray object, writeHDF5Array realizes it on disk, that is, all the delayed operations carried by the object are executed on-the-fly while the object is written to disk. This uses a block-processing strategy so that the full object is not realized at once in memory. Instead the object is processed block by block i.e. the blocks are realized in memory and written to disk one at a time.

In other words, writeHDF5Array(x, ...) is semantically equivalent to writeHDF5Array(as.array(x), ...), except that as.array(x) is not called because this would realize the full object at once in memory.

See ?DelayedArray for general information about DelayedArray objects.

References

Documentation of the H5 predefined datatypes on the HDF Group's Support Portal: https://portal.hdfgroup.org/display/HDF5/Predefined+Datatypes

See Also

  • HDF5Array objects.

  • h5writeDimnames for writing the dimnames of an HDF5 dataset to disk.

  • saveHDF5SummarizedExperiment and loadHDF5SummarizedExperiment in this package (the HDF5Array package) for saving/loading an HDF5-based SummarizedExperiment object to/from disk.

  • HDF5-dump-management to control the location and physical properties of automatically created HDF5 datasets.

  • h5ls to list the content of an HDF5 file.

Examples

## ---------------------------------------------------------------------
## WRITE AN ORDINARY ARRAY TO AN HDF5 FILE
## ---------------------------------------------------------------------
m <- matrix(runif(364, min=-1), nrow=26,
            dimnames=list(letters, LETTERS[1:14]))

h5file <- tempfile(fileext=".h5")
M1 <- writeHDF5Array(m, h5file, name="M1", chunkdim=c(5, 5))
M1
chunkdim(M1)

## By default, writeHDF5Array() does not write the dimnames to the HDF5
## file so they are lost:
dimnames(M1)   # no dimnames

## Set 'with.dimnames' to TRUE to write them to the file:
M1b <- writeHDF5Array(m, h5file, name="M1b", with.dimnames=TRUE)
dimnames(M1b)  # dimnames are back

## With sparse data:
sm <- rsparsematrix(20, 8, density=0.1)
M2 <- writeHDF5Array(sm, h5file, name="M2", chunkdim=c(5, 5))
M2
is_sparse(M2)  # TRUE

## ---------------------------------------------------------------------
## WRITE A DelayedArray OBJECT TO AN HDF5 FILE
## ---------------------------------------------------------------------
M3 <- log(t(DelayedArray(m)) + 1)
M3 <- writeHDF5Array(M3, h5file, name="M3", chunkdim=c(5, 5))
M3
chunkdim(M3)

library(h5vcData)
tally_file <- system.file("extdata", "example.tally.hfs5",
                          package="h5vcData")
h5ls(tally_file)

cvg0 <- HDF5Array(tally_file, "/ExampleStudy/16/Coverages")

cvg1 <- cvg0[ , , 29000001:29000007]

writeHDF5Array(cvg1, h5file, "cvg1")
h5ls(h5file)

Bioconductor/HDF5Array documentation built on Aug. 2, 2022, 11:22 p.m.