Fetching latest commit…
Cannot retrieve the latest commit at this time
|mach||rename trunk/Lib to trunk/scipy to conform to convention (see ticket #…|
|Makefile||rename trunk/Lib to trunk/scipy to conform to convention (see ticket #…|
|README||rename trunk/Lib to trunk/scipy to conform to convention (see ticket #…|
# Author: Travis E. Oliphant # email@example.com # # This can be freely distributed provided this notification remain and it is # understood that no warranty is expressed or implied. CEPHES module for Numerical Python DESCRIPTION: ===================== This package contains all of the important functions from the cephes library with some complex extensions from the amos library in a python module. There are now some hooks for including toms algorithms as well (thanks Lorenzo Catucci). The module is patterned after the umath module and uses the ufunc object as the interface so that all of the functions are accessible on multidimensional arrays. The ufunc methods (reduce, outer, etc.) are also available to the functions which take two arguments (the need for these methods is dubious for these functions, they are just along for the ride as part of the ufunc object). The ufunc object interface is general enough to allow even functions which take more than two arguments and return more than one argument to be implemented. In the ./docs directory is the documentation for the all of the cephes library, and it includes an HTML page showing brief documentation for all included functions calls. INSTALLATION: Part of SciPy CHANGES: =========== 0.5 Initial Release 0.51 Fixed the name of a subroutine that was conflicting with one from umath so that Numeric and cephes work together. 0.52 Fixed the real reason Numeric and cephes weren't working together: I had copied a setup routine from umathmodule.c that wasn't necessary and shouldn't have been called. 0.60 Added kolmogorov and kolmogi functions from cephes file. These functions were not really functions in the cephes library, but they existed in a standalone program. They may not be as robust. Kolmogorov hangs if given 0 as an argument. Someone could probably fix this. 0.61 Changed so Kolmogorov doesn't hang on 0 argument. Returns 1 Also kolmogi returns 0 on 1.0 as argument. 0.70 Added many more functions from the library. Wrote wrappers so that functions which took an integer as one of two arguments could be added. Also added wrappers around functions which returned more than one value: I made a new function for each returned value. yn,kn,jn Student T, and Poisson distribution functions are notable inclusions. 0.71 Added info about mconf.h in the README and created a patchfile to patch mconf.h to a generic BIGENDIAN computer from an IBM_PC Changed gamma to Gamma in the c-code so that Solaris's gamma which actually computes lgamma is not called. Removed log2 from library as it gave poor results on UNK machines. 1.0 Added all of the remaining cephes calls to the library and eliminated the wrapper functions after learning how to really use the ufunc object correctly to handle generic functions. 1.1 Added complex extensions to bessel and airy functions by incorporating the amos libraries into the module with some wrappers around the fortran code. 1.11 Fixed small bugs. 1.15 Added changes from Lorenzo Catucci: a new wofz function and a testing script. 1.2 Added a loopfunc method to the cephes module. This method is a kind of generalized map that takes a Python function with scalar inputs and outputs and a tuple of input arrays or nested sequence objects and returns an array created by applying the Python function in turn to input tuples created from the elements of the input arrays. It uses the same broadcasting rules as are available with ufuncs. Basically, this turns a regular Python function into the equivalent of a ufunc. 1.3 Added besselpoly function. Added empty docstrings to compile with Numeric 15.3 (thanks Warren Focke). Changed makefiles to compile shared libraries with libtool.