Experimental Physics and
Industrial Control System

Noboru Yamamoto <[email protected]> · Mon, 07 Feb 2000 21:26:24 +0900

Geoff Savage wrote:
> 
> Hi,
> 
> I have followed the recent postings on using CA from Python with
> interest.  We here at Fermilab make extensive use of our version of a
> Python/CA interface.  You can get the source and documentation from
> http://www-d0.fnal.gov/~savage/caPython/caPython.html
> Along with the source code I have included Python modules for Linux,
> IRIX, and OSF1.  The wrapper class, CaChannel, is pure Python and will
> also run on all these platforms.
I'm very glad to know there is another group of people supporting Python
in EPICS.

> 
> Highlights:
> + Straight conversion between C functions and Python functions using
> SWIG.  This let's you program in Python as if you were programming in
> C.  Some of the more advanced functions have not been implemented due to
> constraints with how callbacks are handled.  I haven't completed the
Can you describe it in detail ? (may be in the private mail because the
discussion 
can be too technical for the others in this mailing list.)

> scan group functionality as we don't use it here.
> + Included is a Python class that wraps the functions.  This makes CA in
> Python easier to use.
> + Many macros from db_access.h and cadef.h are included along with the
> data type identifiers (DBR_XXXX and DBF_XXXX).

When I started to write Python/CA interface, there were no "%apply" or
"%typemap" 
mechanism in SWIG ( or I just did not recognized them). So for me, it
seemed very difficult 
to map CA library interface to Python interface. So I decided to write a
C module from scratch.
I also assumed that the main user of Python/CA is an accelerator
physicist, so I prepared 
rather simple interface in Python/CA. 

> 
> Unfortunately we have not had the time to evaluate the CA/Python
> interfaces developed by Noboru Yamamoto and Matthieu Bec.  Maybe we can
> combine the strongest parts from each project into a standard CA/Python
> interface.
Our user uses just Python module but never access C-module in there
program directly.
So I think it is possible to replace C-module without affecting user
program, in principle.
But I need to check it.
I would be very happy to discuss CA/Python interface, anyway.

> 
> Please contact me with any questions or concerns.
> 
> Geoff

Noboru Yamamoto
KEKB control group
KEK, JAPAN

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System