var.put.nc {RNetCDF}

Write Data to a NetCDF Variable

Description

Write the contents of a NetCDF variable.

Usage

var.put.nc(ncfile, variable, data, start=NA, count=NA, na.mode=4, pack=FALSE,
  cache_bytes=NA, cache_slots=NA, cache_preemption=NA)

Arguments

Arguments marked "netcdf4" are optional for datasets in that format and ignored for other formats.

ncfile

Object of class NetCDF which points to the NetCDF dataset (as returned from open.nc).

variable

ID or name of the NetCDF variable.

data

An R vector or array of data to be written to the NetCDF variable. Values are taken from data in the order of R vector elements, so that leftmost indices vary fastest over an array.

start

A vector of indices specifying the element where writing starts along each dimension of variable. Indices are numbered from 1 onwards, and the order of dimensions is shown by print.nc (array elements are stored sequentially with leftmost indices varying fastest). By default (start=NA), all dimensions of variable are written from the first element onwards. Otherwise, start must be a vector whose length is not less than the number of dimensions in variable (excess elements are ignored). Any NA values in vector start are set to 1.

count

A vector of integers specifying the number of values to write along each dimension of variable. The order of dimensions is the same as for start. By default (count=NA), count is set to dim(data) for an array or length(data) for a vector. Otherwise, count must be a vector whose length is not less than the number of dimensions in variable (excess elements are ignored). Any NA value in vector count indicates that the corresponding dimension should be written from the start index to the end of the dimension. Note that an unlimited dimension initially has zero length, and the dimension is extended by setting the corresponding element of count greater than the current length.

na.mode

NA values in data are converted to a missing value in the NetCDF dataset. The missing value is defined by attributes of the NetCDF variable, which are selected by the following modes:

mode	data type	attribute(s)
0	numeric	_FillValue, then missing_value
1	numeric	_FillValue only
2	numeric	missing_value only
3	any	no conversion
4	numeric	valid_range valid_min, valid_max, _FillValue
5	any	same as mode 4 for numeric types;
		_FillValue for other types

For explanation of attribute conventions used by mode 4, please see: https://docs.unidata.ucar.edu/nug/current/attribute_conventions.html

pack

Variables are packed if pack=TRUE and the attributes add_offset and/or scale_factor are defined. Default is FALSE.

cache_bytes

("netcdf4") Size of chunk cache in bytes. Value of NA (default) implies no change.

cache_slots

("netcdf4") Number of slots in chunk cache. Value of NA (default) implies no change.

cache_preemption

("netcdf4") Value between 0 and 1 (inclusive) that biases the cache scheme towards eviction of chunks that have been fully read. Value of NA (default) implies no change.

Details

This function writes values to a NetCDF variable. Data values in R are automatically converted to the correct type of NetCDF variable.

Text represented by R type character can be written to NetCDF types NC_CHAR and NC_STRING, and R type raw can be written to NetCDF type NC_CHAR. When writing to NC_CHAR variables, character variables have an implied dimension corresponding to the string length. This implied dimension must be defined explicitly as the fastest-varying dimension of the NC_CHAR variable, and it must be included as the first element of arguments start and count taken by this function.

Due to the lack of native support for 64-bit integers in R, NetCDF types NC_INT64 and NC_UINT64 require special attention. This function accepts the usual R integer (signed 32-bit) and numeric (double precision) types, but to represent integers larger than about 53-bits without truncation, integer64 vectors are also supported.

NetCDF numeric variables cannot portably represent NA values from R. NetCDF does allow attributes to be defined for variables, and several conventions exist for attributes that define missing values and valid ranges. The convention in use can be specified by argument na.mode. Values of NA in argument data are converted to a missing or fill value before writing to the NetCDF variable. Unusual cases can be handled directly in user code by setting na.mode=3.

Variables of user-defined types are supported, subject to conditions on the corresponding data structures in R. "compound" arrays must be stored in R as lists, with items named for the compound fields; items of base NetCDF data types are stored as R arrays, with leading dimensions from the field dimensions (if any) and trailing dimensions from the NetCDF variable. "enum" arrays are stored in R as factor arrays. "opaque" arrays are stored in R as raw (byte) arrays, with a leading dimension for bytes of the opaque type and trailing dimensions from the NetCDF variable. "vlen" arrays are stored in R as a list with dimensions of the NetCDF variable; items in the list may have different lengths; base NetCDF data types are stored as R vectors.

To reduce the storage space required by a NetCDF file, numeric variables can be packed into types of lower precision. The packing operation involves subtraction of attribute add_offset before division by attribute scale_factor. This packing operation is performed automatically for variables defined with the attributes add_offset and/or scale_factor if argument pack is set to TRUE. If pack is FALSE, data values are assumed to be packed correctly and are written to the variable without alteration.

Data in a NetCDF variable is represented as a multi-dimensional array. The number and length of dimensions is determined when the variable is created. The start and count arguments of this routine indicate where the writing starts and the number of values to write along each dimension.

Awkwardness arises mainly from one thing: NetCDF data are written with the last dimension varying fastest, whereas R works opposite. Thus, the order of the dimensions according to the CDL conventions (e.g., time, latitude, longitude) is reversed in the R array (e.g., longitude, latitude, time).

The arguments marked for "netcdf4" format refer to the chunk cache used for reading and writing variables. Default cache settings are defined by the NetCDF library, and they can be adjusted for each variable to improve performance in some applications.

Note

NC_BYTE is always interpreted as signed. For best performance, it is recommended that the definition of dimensions, variables and attributes is completed before variables are read or written.

Author(s)

Pavel Michna, Milton Woods

References

https://www.unidata.ucar.edu/software/netcdf/

Examples

##  Create a new NetCDF dataset and define two dimensions
file1 <- tempfile("var.put_", fileext=".nc")
nc <- create.nc(file1)

dim.def.nc(nc, "station", 5)
dim.def.nc(nc, "time", unlim=TRUE)
dim.def.nc(nc, "max_string_length", 32)

##  Create three variables, one as coordinate variable
var.def.nc(nc, "time", "NC_INT", "time")
var.def.nc(nc, "temperature", "NC_DOUBLE", c(0,1))
var.def.nc(nc, "name", "NC_CHAR", c("max_string_length", "station"))

##  Put some _FillValue attribute for temperature
att.put.nc(nc, "temperature", "_FillValue", "NC_DOUBLE", -99999.9)

##  Define variable values
mytime        <- c(1:2)
mytemperature <- c(1.1, 2.2, 3.3, 4.4, 5.5, 6.6, 7.7, NA, NA, 9.9)
myname        <- c("alfa", "bravo", "charlie", "delta", "echo")

dim(mytemperature) <- c(5,2)

##  Put subsets of the data:
var.put.nc(nc, "time", mytime, start=2, count=1)
var.put.nc(nc, "temperature", mytemperature[3:4,2], start=c(3,2), count=c(2,1))
var.put.nc(nc, "name", myname[3:4], start=c(NA,3), count=c(NA,2))
sync.nc(nc)

##  Put all of the data:
var.put.nc(nc, "time", mytime)
var.put.nc(nc, "temperature", mytemperature)
var.put.nc(nc, "name", myname)

close.nc(nc)
unlink(file1)