IDLSave - a python module to read IDL 'save' files
News!
The IDLSave code is now in
scipy.io! At the moment, it is only available through the svn version of SciPy, but it should be included in future stable releases (from 0.9.0 onwards). To use the version in scipy.io, you can use from scipy.io.idl import readsav and then use readsav as you would use idlsave.read.At this time, the standalone
idlsave package on this page will be maintained in parallel with the scipy.io.idl version.Introduction
IDLSave is a pure python module to import variables from IDL 'save' files (e.g. .sav) into python, and does not require IDL to work. It has a very simple command-line interface, and converts all IDL variables to python types. Arrays are converted to Numpy arrays, and Structures are converted to Numpy record arrays.
(0.9.7, released 18 August 2010)
This program is distributed with permission from ITT Visual Information Systems. To report bugs and request features, please use the issue tracker. To contact me directly, please use thomas dot robitaille [at] gmail dot com.
To follow the development and get a copy of the latest development code, visit the GitHub pages.
Changes
- 0.9.7: Code style changes to conform to scipy coding guidelines. This version is identical to the initial version committed to scipy.io.idl
- 0.9.6: Added unit tests. Results are now returned as a dictionary of variables. Variables can be added to an existing dictionary. Scalars are returned with correct Numpy type. Code re-released under MIT license.
- 0.9.5: Implemented reading in IDL .sav files written with the
/COMPRESSoption - 0.9.4: Fixed bug with pre-defined IDL structures
- 0.9.3: Implemented reading in of IDL pointers
- 0.9.2: Fixed reading of byte values/arrays
- 0.9.1: Implemented reading in of IDL structures as record arrays
- 0.9.0: Initial release
Installation
To install, simply run
python setup.py install inside
the IDLSave-x.x.x directory.
Alternatively, IDLSave can be installed using easy_install idlsave if you have setuptools installed.The only dependency for IDLSave is Numpy (1.3.0 or later)
Quick start
The following example demonstrates how to read a
.sav file into
python. This is done using the idlsave.read method,
which returns an IDLSaveFile instance. The variables
are then accessible as attributes to the IDLSaveFile
instance. Variable names are not case-sensitive. For structures (i.e. recarrays),
variable names can be access either lower or upper case, but not
mixed-case.Python 2.6.1 (r261:67515, Jul 7 2009, 23:51:51)
[GCC 4.2.1 (Apple Inc. build 5646)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import idlsave
>>> s = idlsave.read('varsandstructs.sav')
--------------------------------------------------
Date: Tue Sep 22 11:15:11 2009
User: johndoe
Host: hal9000
--------------------------------------------------
Format: 9
Architecture: x86_64
Operating System: linux
IDL Version: 7.0
--------------------------------------------------
Successfully read 11 records of which:
- 7 are of type VARIABLE
- 1 are of type TIMESTAMP
- 1 are of type NOTICE
- 1 are of type VERSION
--------------------------------------------------
Available variables:
- nan [<type 'numpy.ndarray'>]
- nstruct [<class 'numpy.core.records.recarray'>]
- floatarray [<type 'numpy.ndarray'>]
- astruct [<class 'numpy.core.records.recarray'>]
- journalver [<type 'int'>]
- stringarray [<type 'numpy.ndarray'>]
- zstruct [<class 'numpy.core.records.recarray'>]
--------------------------------------------------
>>> s.journalver
800
>>> s.floatarray
array([[ 0.00000000e+00, 1.00000000e+00, 2.00000000e+00, ...,
9.70000000e+01, 9.80000000e+01, 9.90000000e+01],
[ 1.00000000e+02, 1.01000000e+02, 1.02000000e+02, ...,
1.97000000e+02, 1.98000000e+02, 1.99000000e+02],
[ 2.00000000e+02, 2.01000000e+02, 2.02000000e+02, ...,
2.97000000e+02, 2.98000000e+02, 2.99000000e+02],
...,
[ 9.70000000e+03, 9.70100000e+03, 9.70200000e+03, ...,
9.79700000e+03, 9.79800000e+03, 9.79900000e+03],
[ 9.80000000e+03, 9.80100000e+03, 9.80200000e+03, ...,
9.89700000e+03, 9.89800000e+03, 9.89900000e+03],
[ 9.90000000e+03, 9.90100000e+03, 9.90200000e+03, ...,
9.99700000e+03, 9.99800000e+03, 9.99900000e+03]], dtype=float32)
>>> s.nstruct
rec.array([ (array([[ 0.00000000e+00, 1.00000000e+00, 2.00000000e+00, ...,
9.70000000e+01, 9.80000000e+01, 9.90000000e+01],
[ 1.00000000e+02, 1.01000000e+02, 1.02000000e+02, ...,
1.97000000e+02, 1.98000000e+02, 1.99000000e+02],
[ 2.00000000e+02, 2.01000000e+02, 2.02000000e+02, ...,
2.97000000e+02, 2.98000000e+02, 2.99000000e+02],
...,
[ 9.70000000e+03, 9.70100000e+03, 9.70200000e+03, ...,
9.79700000e+03, 9.79800000e+03, 9.79900000e+03],
[ 9.80000000e+03, 9.80100000e+03, 9.80200000e+03, ...,
9.89700000e+03, 9.89800000e+03, 9.89900000e+03],
[ 9.90000000e+03, 9.90100000e+03, 9.90200000e+03, ...,
9.99700000e+03, 9.99800000e+03, 9.99900000e+03]], dtype=float32),
array([[a, a, a, ..., a, a, a],
[a, a, a, ..., a, a, a],
[a, a, a, ..., a, a, a],
...,
[a, a, a, ..., a, a, a],
[a, a, a, ..., a, a, a],
[a, a, a, ..., a, a, a]], dtype=object), 'named structure')],
dtype=[(('floatarray', 'FLOATARRAY'), '|O8'),
(('stringarray', 'STRINGARRAY'), '|O8'),
(('comment', 'COMMENT'), '|O8')])
>>> s.nstruct.stringarray
array([ [[a a a ..., a a a]
[a a a ..., a a a]
[a a a ..., a a a]
...,
[a a a ..., a a a]
[a a a ..., a a a]
[a a a ..., a a a]]], dtype=object)
Many thanks to Craig Markwardt for publishing the Unofficial Format Specification for IDL
.sav files, without which this Python module would not exist. IDL® is a registered trademark of ITT Visual Information Systems, Inc. for their Interactive Data Language software.