Fortran 77 Extension Builder for Python
Photon Research Associates, Inc.
The use of Fortran codes in scientific, research and development environments,
such computational chemistry, radiation transport, and modeling and simulation
is well established. A flexible paradigm of code development and scientific
research is possible making use of interactive, scripting languages controlling
and "steering" computational-intensive codes developed in traditional,
compiled languages. A general-purpose tool (Extension Builder) has
been developed to build Fortran 77 extension modules and to define public
interfaces exposing Fortran 77 variables and functions.
Fortran 77 developers and researchers use Extension Builder to declare
a public interface to Fortran 77 procedures and variables. Variables
of all common Fortran 77 data types (up to 2 dimensions) can be declared
using Extension Builder. Function prototype definitions describe the Fortran
procedures to be made public. Function prototype definitions can
include parameters of all common types. Extension Builder was developed
in C++ using a cross-platform graphical user interface tool. Currently,
Extension Builder generates MAKE files compatible with Silicon Graphics
systems (IRIX 6.4 and later) and Win32 Microsoft Visual Studio IDE (MSVC5.0
and Digital Visual Fortran).
Developers specify function prototypes for Fortran 77 procedures to be
made public to Python users. Function parameters can be declared with "input",
"output", or "inout" access, as defined in Extension Builder's SWIG (Simplified
Wrapper and Interface Generator) typemaps. Likewise, Fortran variables
can be declared with default values with minimum and maximum range values.
Based on declarations, a code generator creates wrapper files, a SWIG interface
file, and a MAKE file. The MAKE file can be executed from within Extension
builder, from within an IDE, or stand-alone. A target dynamic link
library or static python executable (Unix only) is then built.
Extension Builder utilizes SWIG extensively to build the wrapper interfaces.
Customized SWIG typemaps are used to map standard data types and Numeric
Python (NumPy) array types from the python user environment to the Fortran
extension library via the public functions and variables.