Milestone 1.2.0

This minor release comes almost four months after the 1.1.0 release. The major features of this release are a new testing framework and huge amount of documentation work. It also includes a some minor API breakage scheduled in the 1.1 release.

Please note that NumPy 1.2.0 requires Python 2.4 or greater.

Major documentation improvements

This summer saw the first NumPy Documentation Marathon, during which many thousands of lines of documentation were written. In addition, a web framework was developed which allows the community to contribute docstrings in a wiki-like fashion, without needing access to the source repository. The new reference guide, which is based on these contributions, was built using the popular Sphinx tool. While the documentation coverage is now better than ever, there is still a lot of work to be done, and we encourage interested parties to register and contribute further.

New testing framework

The NumPy test framework now relies on the nose testing framework version 0.10 or later. Much of the work on the testing framework was carried out as part of a Google Summer of Code project.

Buffered iterator

Roberto de Almeida's buffered iterator has been added as "numpy.lib.Arrayterator". The buffered iterator reads large arrays in small blocks, allowing operations over very large data-sets while maintaining a small memory footprint. Any object that supports multidimensional slices can be used, including variables from NetCDF, pynetcdf and ndarrays.

Median

This release is the second step in adding 'axis' support for numpy.median(); the first step was included in the 1.1 release. This is not just useful on its own, but it is also makes median() consistent with mean(), min(), max(), var(), std(), sum(), prod(), etc.

The new call signature is:

def median(a, axis=None, out=None, overwrite_input=False)

This returns the median of the array elements taken over a flattened version of the array by default, otherwise over the specified axis.

Histogram

The semantics of histogram has been modified to fix long-standing issues with outliers handling. The main changes concern

1. the definition of the bin edges, now including the rightmost edge, and
2. the handling of upper outliers, now ignored rather than tallied in the rightmost bin.

The previous behavior is still accessible using new=False, but is scheduled to be deprecated in the next release (1.3).

Random number generator improvements

Several improvements and bug-fixes to random number generator distributions. The negative binomial generator now accepts floating point arguments. The Zipf generator now handles an integer wraparound problem. Fixed a scaling factor for the F distribution. The logarithmic series now rejects out of bounds results due to integer conversions.

Set operations

unique1d will now return reverse indices, i.e., the indices required to build the input array from the output set.

Optimizations

Numpy import times have been heavily optimized, and should now be at least twice faster on most platforms.

Bug-fixes