ds format

Documentation version: 2.0.1 3.0.0 3.1.0 3.2.0 3.3.3 3.4.0 3.5.2 3.6.1 3.7.0 4.0.1 4.1.1 4.1.2 (latest) master
You are viewing documentation for version 3.7.0. The latest release is 4.1.2.
GitHub

Python API

The ds-format Python package provides API for reading, writing and manipulating data files. The library can be imported with:

import ds_format as ds

Contents

Function Description
attr Get or set a dataset or variable attribute.
attrs Get or set variable or dataset attributes.
dim Get a dimension size.
dims Get dataset or variable dimensions or set variable dimensions.
find Find a variable, dimension or attribute matching a glob pattern in a dataset.
findall Find variables, dimensions or attributes matching a glob pattern in a dataset.
group_by Group values along a dimension.
merge Merge datasets along a dimension.
meta Get or set dataset or variable metadata.
read Read dataset from a file.
readdir Read multiple files in a directory.
rename Rename a variable.
rename_attr Rename a dataset or variable attribute.
rename_attr_m Rename one or more dataset or variable attributes.
rename_dim Rename a dimension.
rename_dim_m Rename one or more dimensions.
rename_m Rename one or more variables.
require Require that a variable, dimension or attribute is defined in a dataset.
rm Remove a variable.
rm_attr Remove a dataset or variable attribute.
select Filter dataset by a selector.
size Get variable size.
type Get or set variable type.
var Get or set variable data.
vars Get all variable names in a dataset.
with_mode Context manager which temporarily changes ds.mode.
write Write dataset to a file.

Constants

ds.drivers.netcdf.JD_UNITS

days since -4713-11-24 12:00 UTC

NetCDF units for storing Julian date time variables.

ds.drivers.netcdf.JD_CALENDAR

proleptic_greogorian

NetCDF calendar for storing Julian date time variables.

Variables

mode

Error handling mode. If “strict”, handle missing variables, dimensions and attributes as errors. If “moderate”, report a warning. If “soft”, ignore missing items. Overrides the environment variable DS_MODE.

Examples:

Set error handling mode to strict.

ds.mode = 'strict'

Environment variables

DS_MODE

The same as mode.

Functions

attr

Get or set a dataset or variable attribute.

Usage: attr(d, attr, *value, var=None)

Arguments:

Options:

Return value:

Attribute value if value is not set, otherwise None.

Examples:

Get an attribute long_name of a variable temperature in dataset.nc.

$ d = ds.read('dataset.nc')
ds.attr(d, 'long_name', var='temperature')
'temperature'

Get a dataset attribute title of dataset.nc.

$ ds.attr(d, 'title')
'Temperature data'

Set an attribute units of a variable temperature to K.

$ ds.attr(d, 'units', 'K', var='temperature')
$ ds.attr(d, 'units', var='temperature')
'K'

attrs

Get or set variable or dataset attributes.

Usage: attrs(d, var=None, *value)

Arguments:

Options:

Return value:

Attributes (dict).

Examples:

Get attributes of a variable temperature in a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.attrs(d, 'temperature')
{'long_name': 'temperature', 'units': 'celsius'}

Get dataset attributes.

$ ds.attrs(d)
{'title': 'Temperature data'}

Set attributes of a variable temperature.

$ ds.attrs(d, 'temperature', {'long_name': 'new temperature', 'units': 'K'})
$ ds.attrs(d, 'temperature')
{'long_name': 'new temperature', 'units': 'K'}

dim

Get a dimension size.

Usage: dim(d, dim, full=None)

Arguments:

Options:

Return value:

Dimension size or 0 if the dimension does not exist (int).

Examples:

Get the size of a dimension time in dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.dim(d, 'time')
3

Get the size of a dimension time in dataset.nc without reading data.

$ d = ds.read('dataset.nc', full=True)
$ ds.dim(d, 'time', full=True)
3

dims

Aliases: get_dims

Get dataset or variable dimensions or set variable dimensions.

Usage:

dims(d, var=None, *value, full=False, size=False)
get_dims(d, var=None, full=False, size=False)

The function get_dims (deprecated) is the same as dims, but assumes that size is True if var is None and does not allow setting of dimensions.

Arguments:

Options:

Return value:

If size is False, a list of dataset or variable dimension names (list of str). If size is True, a dictionary of dataset or variable dimension names and sizes (dict), where a key is a dimension name (str) and the value is the dimension size (int). The order of keys in the dictionary is not guaranteed. Dataset dimensions are the dimensions of all variables together.

Examples:

Get dimensions of a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.dims(d)
['time']

Get dimension sizes.

$ ds.dims(d, size=True)
{'time': 3}

Get dimensions of a variable temperature.

$ ds.dims(d, 'temperature')
['time']

find

Find a variable, dimension or attribute matching a glob pattern in a dataset.

Usage: find(d, what, name, var=None)

If more than one name matches the pattern, raises ValueError.

Arguments:

Options:

Return value:

A variable, dimension or attribute name matching the pattern, or name if no matching name is found (str).

Examples:

Find a variable matching the glob pattern temp* in a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.find(d, 'var', 'temp*')
'temperature'

findall

Find variables, dimensions or attributes matching a glob pattern in a dataset.

Usage: findall(d, what, name, var=None)

Arguments:

Options:

Return value:

A list of variables, dimensions or attributes matching the pattern, or [name] if no matching names are found (list of str).

Examples:

Find all variables matching the glob pattern t* in a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.findall(d, 'var', 't*')
['temperature', 'time']

attrs

Get or set variable or dataset attributes.

Usage: attrs(d, var=None, *value)

Arguments:

Options:

Return value:

Attributes (dict).

Examples:

Get attributes of a variable temperature in a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.attrs(d, 'temperature')
{'long_name': 'temperature', 'units': 'celsius'}

Get dataset attributes.

$ ds.attrs(d)
{'title': 'Temperature data'}

Set attributes of a variable temperature.

$ ds.attrs(d, 'temperature', {'long_name': 'new temperature', 'units': 'K'})
$ ds.attrs(d, 'temperature')
{'long_name': 'new temperature', 'units': 'K'}

group_by

Group values along a dimension.

Usage: group_by(d, dim, group, func)

Each variable with a given dimension dim is split by group into subsets. Each subset is replaced with a value computed by func.

Arguments:

Return value:

None

Examples:

Calculate mean along a dimension time for a group where time <= 2 and a group where time > 2.

$ d = {
	'time': np.array([1., 2., 3., 4.]),
	'temperature': np.array([1., 3., 4., 6.]),
	'.': {
		'time': { '.dims': ['time'] },
		'temperature': { '.dims': ['time'] },
	}
}
$ ds.group_by(d, 'time', d['time'] > 2,  np.mean)
$ print(d['time'])
[1.5 3.5]
$ print(d['temperature'])
[1.5 3.5]

merge

Merge datasets along a dimension.

Usage: merge(dd, dim, new=None, variables=None, jd=True)

Merge datasets along a dimension dim. If the dimension is not defined in the dataset, merge along a new dimension dim. If new is None and dim is not new, variables without the dimension dim are set with the first occurrence of the variable. If new is not None and dim is not new, variables without the dimension dim are merged along a new dimension new. If variables is not None, only those variables are merged along a new dimension, and other variables are set to the first occurrence of the variable. Variables which are merged along a new dimension and are not present in all datasets have their subsets corresponding to the datasets where they are missing filled with missing values. Dataset and variable metadata are merged sequentially from all datasets, with metadata from later datasets overriding metadata from the former ones. When merging time variables whose units are not equal and jd is True, they are first converted to Julian date and then merged.

Arguments:

Options:

Return value:

A dataset (dict).

Examples:

Merge datasets d1 and d2 along a dimension time.

$ d1 = {'time': [1, 2, 3], 'temperature': [16., 18., 21.], '.': {
	'time': { '.dims': ['time'] },
	'temperature': { '.dims': ['time'] },
}}
$ d2 = { 'time': [4, 5, 6], 'temperature': [23., 25., 28.], '.': {
	'time': { '.dims': ['time'] },
	'temperature': { '.dims': ['time'] },
}}
$ d = ds.merge([d1, d2], 'time')
$ print(d['time'])
[1 2 3 4 5 6]
$ print(d['temperature'])
[16. 18. 21. 23. 25. 28.]

meta

Aliases: get_meta

Get or set dataset or variable metadata.

Usage: meta(d, var=None, *value, create=False)

Arguments:

Options:

Return value:

Metadata (dict).

Examples:

Get metadata of a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ print(ds.meta(d))
{'.': {'title': 'Temperature data'}, 'temperature': {'long_name': 'temperature', 'units': 'celsius', '.dims': ('time',), '.size': (3,), '.type': 'float64'}, 'time': {'long_name': 'time', 'units': 's', '.dims': ('time',), '.size': (3,), '.type': 'int64'}}

Get metadata of a variable temperature.

$ ds.meta(d, 'temperature')
{'long_name': 'temperature', 'units': 'celsius', '.dims': ('time',), '.size': (3,), '.type': 'float64'}

Set metadata of a variable temperature.

$ ds.meta(d, 'temperature', { '.dims': ['new_time'], 'long_name': 'new temperature', 'units': 'K'})
$ ds.meta(d, 'temperature')
ds.meta(d, 'temperature', { '.dims': ['new_time'], 'long_name': 'new temperature', 'units': 'K'})

read

Read dataset from a file.

Usage: read(filename, variables=None, sel=None, full=False, jd=False)

Arguments:

Options:

Return value:

Dataset (dict).

Supported formats:

Examples:

Read a file dataset.nc.

$ d = ds.read('dataset.nc')
$ print(d.keys())
dict_keys(['.', 'temperature', 'time'])
$ print(d['temperature'])
[16. 18. 21.]
$ d['.']['temperature']
{'long_name': 'temperature', 'units': 'celsius', '.dims': ('time',), '.size': (3,), '.type': 'float64'}

Read a variable temperature at an index 0 of the dimension time from dataset.nc.

$ d = ds.read('dataset.nc', 'temperature', sel={'time': 0})
$ d.keys()
dict_keys(['.', 'temperature'])
$ print(d['temperature'])
16.0

Read only the metadata of dataset.nc.

$ d = ds.read('dataset.nc', [], full=True)
$ d.keys()
dict_keys(['.'])
$ print(d['.'])
{'.': {'title': 'Temperature data'}, 'temperature': {'long_name': 'temperature', 'units': 'celsius', '.dims': ('time',), '.size': (3,), '.type': 'float64'}, 'time': {'long_name': 'time', 'units': 's', '.dims': ('time',), '.size': (3,), '.type': 'int64'}}

readdir

Read all data files in a directory.

Usage: readdir(dirname, variables=None, merge=None, warnings=[], recursive=False, parallel=False, executor=None, njobs=None, …)

Only files with known extensions are read. Files are read in an alphabetical order. Variable filename is added to the output datasets, containing the name of the file. If merge is not None, variables i and n are added to the resulting dataset, containing the index within the input dataset and a file index referring to the filename variable, respectively. They are defined along the dimension merge and are zero-indexed.

Arguments:

Options:

Return value:

A list of datasets (list of dict) if merge is None or a merged dataset (dict) if merge is a dimension name.

Supported formats:

Examples:

Read datasets dataset1.nc and dataset2.nc in the current directory (.).

$ ds.write('dataset1.nc', { 'time': [1, 2, 3], 'temperature': [16., 18., 21.], '.': {
	'time': { '.dims': ['time'] },
	'temperature': { '.dims': ['time'] },
}})
$ ds.write('dataset2.nc', { 'time': [4, 5, 6], 'temperature': [23., 25., 28.], '.': {
	'time': { '.dims': ['time'] },
	'temperature': { '.dims': ['time'] },
}})
$ dd = ds.readdir('.')
$ for d in dd: print(d['time'])
[1 2 3]
[4 5 6]

Read datasets in the current directory and merge them by a dimension time.

$ d = ds.readdir('.', merge='time')
$ print(d['time'])
[1 2 3 4 5 6]
$ print(d['temperature'])
[16. 18. 21. 23. 25. 28.]

rename

Rename a variable.

Usage: rename(d, old, new)

Any dimension with the same name is also renamed.

Arguments:

Return value:

None

Examples:

Rename a variable temperature to new_temperature in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.vars(d)
['temperature', 'time']
$ ds.rename(d, 'temperature', 'new_temperature')
$ ds.vars(d)
['new_temperature', 'time']

rename_attr

Rename a dataset or variable attribute.

Usage: rename_attr(d, old, new, var=None)

Arguments:

Options:

Return value:

None

Examples:

Rename an attribute units of a variable temperature to new_units in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.attrs(d, 'temperature')
{'long_name': 'temperature', 'units': 'celsius'}
$ ds.rename_attr(d, 'units', 'new_units', var='temperature')
$ ds.attrs(d, 'temperature')
{'long_name': 'temperature', 'new_units': 'celsius'}

Rename a dataset attribute title to new_title.

$ ds.attrs(d)
{'title': 'Temperature data'}
$ ds.rename_attr(d, 'title', 'new_title')
$ ds.attrs(d)
{'new_title': 'Temperature data'}

rename_attr_m

Rename one or more dataset or variable attributes.

Arguments:

Options:

Return value:

None

Examples:

Rename an attribute long_name to new_long_name and units to new_units of a variable temperature in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.attrs(d, 'temperature')
{'long_name': 'temperature', 'units': 'celsius'}
$ ds.rename_attr_m(d, {'long_name': 'new_long_name', 'units': 'new_units'}, var='temperature')
$ ds.attrs(d, 'temperature')
{'new_long_name': 'temperature', 'new_units': 'celsius'}

Rename a dataset attribute title to new_title.

$ ds.attrs(d)
{'title': 'Temperature data'}
$ ds.rename_attr_m(d, {'title': 'new_title'})
$ ds.attrs(d)
{'new_title': 'Temperature data'}

rename_dim

Rename a dimension.

Usage: rename_dim(d, old, new)

Arguments:

Return value:

None

Examples:

Rename a dimension time to new_time in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.dims(d)
['time']
$ ds.rename_dim(d, 'time', 'new_time')
$ ds.dims(d)
['new_time']

rename_dim_m

Rename one or more dimensions.

Usage: rename_dim_m(d, mapping)

Arguments:

Return value:

None

Examples:

Rename a dimension time to new_time in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.dims(d)
['time']
$ ds.rename_dim_m(d, {'time': 'new_time'})
$ ds.dims(d)
['new_time']

rename_m

Rename one or more variables.

Usage: rename_m(d, mapping)

Any dimension with the same name is also renamed.

Arguments:

Return value:

None

Examples:

Rename a variable time to new_time and temperature to new_temperature in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.vars(d)
['temperature', 'time']
$ ds.rename(d, {'time': 'new_time', 'temperature': 'new_temperature'})
$ ds.vars(d)
['new_temperature', 'new_time']

require

Require that a variable, dimension or attribute is defined in a dataset.

Usage: require(d, what, name, var=None, full=False)

If the item is not found and the mode is “soft”, returns False. If the mode is “strict”, raises NameError. If the mode is “moderate”, produces a warning and returns False.

Arguments:

Options:

Return value:

true if the required item is defined in the dataset, otherwise false or raises an exception depending on the mode.

Examples:

Require that a variable temperature is defined in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.require(d, 'var', 'temperature')
True

rm

Remove a variable.

Usage: rm(d, var)

Arguments:

Return value:

None

Examples:

Remove a variable temperature in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.vars(d)
['temperature', 'time']
$ ds.rm(d, 'temperature')
$ ds.vars(d)
['time']

rm_attr

Remove a dataset or variable attribute.

Usage: rm_attr(d, attr, var)

Arguments:

Options:

Return value:

None

Examples:

Remove an attribute long_name of a variable temperature in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.attrs(d, 'temperature')
{'long_name': 'temperature', 'units': 'celsius'}
$ ds.rm_attr(d, 'long_name', var='temperature')
$ ds.attrs(d)
{'title': 'Temperature data'}

Remove a dataset attribute title in a dataset read from dataset.nc.

$ ds.attrs(d)
{'title': 'Temperature data'}
$ ds.rm(d, 'title')
$ ds.attrs(d)
{}

select

Filter dataset by a selector.

Usage: select(d, sel)

The function subsets data of all variables in a dataset d by a selector sel. Data can be subset by a mask or a list of indexes along one or more dimensions.

Arguments:

Return value:

None

Examples:

Subset index 0 a along dimension time in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.var(d, 'temperature')
print(ds.var(d, 'temperature'))
$ ds.select(d, {'time': 0})
$ ds.var(d, 'temperature')
16

Subset by a mask along a dimension time in a dataset read from dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.select(d, {'time': [False, True, True]})
$ ds.var(d, 'temperature')
[18. 21.]

size

Get variable size.

Usage: size(d, var)

Variable size is determined based on the size of the variable data if defined, or by variable metadata attribute .size.

Arguments:

Return value:

Variable size (list) or None if not defined.

Examples:

Get the size of a variable temperature in dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.size(d, 'temperature')
[3]

type

Get or set variable type.

Usage: type(d, var, *value)

Variable type is determined based on the type of the variable data if defined, or by variable metadata attribute .type.

Arguments:

Return value:

Variable type (str) or None if not defined.

Examples:

Get the type of a variable temperature in dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.type(d, 'temperature')
'float64'

Set the type of a variable temperature to int64.

$ ds.type(d, 'temperature', 'int64')
$ ds.type(d, 'temperature')
'int64'
$ print(ds.var(d, 'temperature'))
[16 18 21]

var

Get or set variable data.

Usage: var(d, var, *value)

Arguments:

Return value:

Variable data (np.ndarray or np.generic) or None if the variable data are not defined or value is supplied. If the variable data are a list or tuple, they are converted to np.ndarray, or to np.ma.MaskedArray if they contain None, which is masked. If the variable data are int, float, bool, str or bytes, they are converted to np.generic. Raises ValueError if the output dtype is not one of float32, float64, int8, int16, int32, int64, uint8, uint16, uint32, uint64, bool, bytes<n>, str<n>, or object for which all items are an instance of str or bytes.

Examples:

Get data of a variable temperature in a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ print(ds.var(d, 'temperature'))
[16. 18. 21.]

Set data of a variable temperature.

$ ds.var(d, 'temperature', [17, 18, 22])
$ ds.var(d, 'temperature')
array([17, 18, 22])

vars

Aliases: get_vars

Get all variable names in a dataset.

Usage: get_vars(d, full=False)

Arguments:

Options:

Return value:

Variable names (list of str).

Examples:

List variables in a dataset dataset.nc.

$ d = ds.read('dataset.nc')
$ ds.vars(d)
['temperature', 'time']

List variables in a dataset dataset.nc without reading the data.

$ d = ds.read('dataset.nc', [], full=True)
$ ds.vars(d, full=True)
['temperature', 'time']

with_mode

Context manager which temporarily changes ds.mode.

Arguments:

Examples:

A block of code in which ds.mode is set to “soft”.

with ds.with_mode('soft'):
	...

write

Write dataset to a file.

Usage: write(filename, d)

The file type is determined from the file extension.

Arguments:

Return value:

None

Supported formats:

Examples:

Write a dataset to a file dataset.nc.

$ ds.write('dataset.nc', {
	'time': [1, 2, 3],
	'temperature': [16. 18. 21.],
})

Write a dataset with metadata to a file dataset.nc.

$ ds.write('dataset.nc', {
	'time': [1, 2, 3],
	'temperature': [16. 18. 21.],
	'.': {
		'.': { 'title': 'Temperature data' },
		'time': { '.dims': ['time'] },
		'temperature': { '.dims': ['time'], 'units': 'degree_celsius' },
	}
})