Nexus Python API

Overview

NeXus Python Api binds the NeXus libraries to Python. It brings functionality of the NeXus API to Python for reading, writing and modifying NeXus Files. Python NeXus API imitates the functionality NeXus API though with a more object oriented flavour.

• Information on NeXus Dataformat: http://www.nexusformat.org/Introduction

Installation

Requirements

The following software has to be installed on your system in order to successfully install and use the Python bindings for NAPI

• Python 2.5 or higher
• setuptools (replaces the old distutils code)
• sphinx to build the documentation
• a working installation of the runtime binaries of libNeXus
• numpy-package

Supported operating systems are: Windows, OS X and Linux.

The bindings should be easily modified for any version of Python which supports ctypes and numpy. In order to build the documentation sphinx is required.

Building and Installing

This package uses the standard distutils installer for python

$ python setup.py install

You will also need to make sure that libNeXus can be found. In order to build the documentation use

$ python setup.py build_sphinx

To run the tests use

$ python setup.py test

Using API from Python

Test Files

The Python NeXus-API includes nxstest.py, which provides similar tests to the original C api file napi_test.c.

After installing, you can run the test using:

$ python [options] [formats]

where options are -q for quiet and -x for external, and formats are hdf4, hdf5 and xml. The default is to test hdf5 format read/write.

Using The API And An Example

The API's functions aim to reproduce the funtionality of the C API closely. Some low level functionality has been hidden from the user. Memory allocation functions NXmalloc and NXfree are done automatically in the API when needed. The file handle is an object with methods rather than a parameter to functions. Instead of checking status codes, errors raise exceptions.

The input and returned values match the format of the data in the files. On return, python creates values of the correct type. However on input, numeric types must be created correctly using numpy.array(...,dtype='type'). The matching datatypes are:

NeXus datatype	Python Datatype
NX_CHAR	char
NX_FLOAT32	float32
NX_FLOAT64	float64
NX_UINT8	uint8
NX_INT16	int16
NX_UINT16	uint16
NX_INT32	int32
NX_UINT32	uint32

Here is simple example program that demonstrates the basic functions and most important differences between the C Nexus Api and the Python Nexus API.

• Creates a NeXus file with access method HDF5
• adds datagroups
• makes a data array of data type NX_INT32
• puts data to the array
• reads the data and attributes
• prints data and attribute value<
• closes the groups and the file.

import nxs,numpy

# Access method accepts strings or integer (e.g., nxs.ACC_CREATE5)
f = nxs.open("test.h5", 'w5')
f.makegroup("testgroup", "NXentry")
f.opengroup("testgroup", "NXentry")
f.makegroup("anothergroup", "NXentry")

# Some data to store in the file, this of type int16
data = numpy.array([[0,1,2,3],[4,5,6,7],[8,9,10,11],[12,13,14,15] ],'int16')

# Make a data set for the array. Note that this could also
# be done as f.makedata('data1','int16',[4,4])
f.makedata('data1', dtype=data.dtype, shape=data.shape)
f.opendata("data1")
f.putdata(data)

# Attribute type can be inferred from the data or specified.  If inferred, it
# must match the type of the data.  Attributes are scalars or strings, with
# string length inferred from value.
f.putattr('integer-attribute', 42, 'int16')
f.putattr('double-attribute', 3.14159)
f.closedata()
# NeXus returns arrays from getattr/getdata/getslab
f.opendata("data1")
print 'data :',f.getdata()

# getnext functions return tuples
attrname,length,type = f.getnextattr ()
value = f.getattr(attrname, length, type)
print 'first attribute: ', value

# ... or you can use iterators for attrs and entries
print 'all attributes'
for attr,value in f.attrs():
    print "  %s: %s"%(attr,value)

f.closedata()
f.closegroup()
f.close()

NeXus API Routines

Documentation for the individual methods, and how they differ from the basic NAPI methods is available from the Python command line. Rather than duplicate it here, use the following in Python:

import nxs
help(nxs)