Experimental Physics and
Industrial Control System

Matt Newville <[email protected]> · Wed, 20 Apr 2011 21:36:25 -0500

Hi Mark, Andrew,

There are ca.context_create(), ca.context_destroy(),
ca.attach_context() functions that go directly to the C library
functions.

In all (both) examples I had of using put() with threads, a new
context was explicitly created, then PVs were created and used within
that thread, and finally the context was destroyed.

Andrews suggestion of attaching to the "main thread" context works
too.  I think there is a mistake in ca.py, as
ca.attach_context(context) raises an exception if the return value
from ca_attach_context() is not ECA_ISATTACHED (I think I read the doc
incorrectly -- I get ECA_NORMAL).  I pushed the change to ca.py to run
PySEVCHK expecting expect ECA_NORMAL instead of ECA_ISATTACHED to
github.  Ignoring that bug for now, attaching to the existing context
like this does work:

#
import time, epics, threading

def threaded_putwait(pv, context, value):
    "put-with-wait for calling in a thread"
    t0 = time.time()
    try:
        epics.ca.attach_context(context)
    except:
        pass
    pv.put(value, wait=True, timeout=10.0)
    print ' - threaded_putwait done (%.3f sec)' % (time.time()-t0)

if __name__ == '__main__':
    pvname = 'XXX.VAL'
    target = 10

    pv = epics.PV(pvname)
    pv.put(-target, wait=True)
    context = epics.ca.current_context()

    th = threading.Thread(target=threaded_putwait,
                          args=(pv, context, target))
    th.start()
    th.join()
    print 'All Done.'

Mark, is that approximately what you found as well?

Perhaps it would be better for a PV to save its context at creation,
and automatically know to switch to its context before doing any work?
 That would make the saving and attaching of the context in the above
script unnecessary, so that "plain looking" code could more easily be
made multi-threaded without having to worry about contexts.   I'm not
sure what the negative consequences of that would be.

--Matt Newville

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System