Presentation given at SciPy 2003, September 2003
Presentation given at SciPy 2003, September 2003
Presentation given at OSCON 2003, July 2003
Python has long had an array module (Numeric) for science and engineering applications; why a replacement? We explain the motivations for developing numarray, which are primarily, though not entirely focused on enabling the use of larger arrays and data sets that Numeric has difficulty handling. We also describe the design issues in its development and its new features and capabilities. Numarray is highly compatible with Numeric, including the C-API, though there are some differences that are discussed. Numarray is sufficiently well developed that it is being used in production pipelines to reduce and calibrate Hubble Space Telescope (HST) data and is being distributed to HST users along with applications for data reduction. Finally, we outline planned enhancements and improvements. Numarray is available from the Sourceforge numpy project page.
From PyCon, March 2003.
Although Python has long had a module for numeric array manipulations, it has had some shortcomings that prevent it from being as useful for astronomy applications as it could be. We have re-implemented the module to handle large arrays in a more memory-efficient manner, and to support direct access of data in tables and non-native data formats. The new module has been implemented mostly in Python although the core computational loops are performed in C for efficiency. The new approach allows arrays to be subclassed as well as new kinds of array objects, such as record arrays, to be created. This paper discusses the design and implementation issues that were addressed in the development of the new array module for Python and gives examples of its use.
From Astronomical Data Analysis Software and Systems XI, October 2001
Full text in html