INSTALL

If you have the "numpy", "numarray" or "Numeric" python packages, and
PGPLOT installed, and the PGPLOT_DIR environmental variable is
pointing to the directory where PGPLOT is installed, it should be
possible to install ppgplot with one command (issued as user "root"):

 # python setup.py install

New in 2018:
    Now also support python3

If the pgplot libraries are in some other directory, or you don't feel
like setting the PGPLOT_DIR, try this (again as user "root"):

 # python setup.py build_ext -L/usr/local/pgplot
 # python setup.py install

Assuming "/usr/local/pgplot" is the directory where PGPLOT is
installed.

New in 2018:
    The "-L/.../" trick does not work if > 1 'libcpgplot.{so|dylib}' are
    installed on the system.

    If linkage to an alternative, co-existing, PGPLOT library is required
    (e.g. 'giza' - http://giza.sourceforge.net/):
        export PGPLOT_DIR=/path/to/giza-root

    The linker will be instructed to choose the library(ies) from
    /path/to/giza-root/lib/ over those found in the system paths 
    in such a way the user will not have to tinker with their
    LD_LIBRARY_PATH variable to make ld.so find and load 
    he correct shared libraries.

Depending on how you compiled PGPLOT, you may need to link ppgplot
with additional runtime libraries. If compilation (linking) of the
extension fails due to unresolved symbols, then this is probably the
case. A common culprit for this is the PNG library, which is needed if
you compiled PGPLOT with PNG driver support. It this case comment-in
the line:
#    libraries.append("png")

in "setup.py", and try again.

If you are compiling with an older fortran compiler (e.g. not gfortran),
then you may need to comment-in the following line in setup.py:
#    libraries.append("g2c")

Since 1.4 ppgplot is configured to prefer "numpy" over "numarray" over
"Numeric".  If for some reason you want to use "numarray" or
"Numeric", then uncomment the appropriate "raise ImportError" lines in
setup.py

New in 2018:
    In stead of having to edit the setup.py script it is now possible to
    pass '--no-Numeric', '--no-numpy' and/or '--no-numarray' on the
    commandline to prevent checking for a specific num* implementation

Have fun ! :)