Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Fri, 11 Jun 2004 14:37:17 -0500

Pete R Jemian wrote:

But someone can always come up with a standard situation where
a single format specifier for all fields is problematic.
How can we handle some fields differently from others?

This can already be done for the precision value, if it is implemented in
the record support code - there is no requirement that there be just one
field for setting precision, but I'm not aware of any records that have
more than one; many record types don't have a PREC field at all. There's
nothing stopping something like the motor record from providing a
different precision for some fields than it does for others. Here's how:
There is a routine in the record support, called get_precision() which is
responsible for returning the precision value for any record field. Most
record types just check if the field they're being asked about is VAL and
if it is they return PREC; if not they just ask recGblGetPrec() which sets
a default for the particular field type. Any record that needs multiple
precision settings can just test for multiple field addresses and return
different precision values as it sees appropriate - presumably it would
group different fields together and have a precision field for each group.
I agree that Stephanie's .FMT would be a nice idea if we were starting
from scratch, but right now we would have to replace the numeric precision
field that currently exists in all of the struct dbr_gr_<xxx> definitions
in cadef.h and we'd have to modify the CA protocol and all of the clients
that use it. In the future we'll be able to do that if we decide it's the
best solution, but I can't really see us moving away from a single number
defining precision for quite a while.
We already have an implementation that copes with multiple fields if we
can agree to redefine what the precision value actually means. It would
probably be simplest and best for the next major release branch (R3.15) to
change the definition of precision to be that of the ANSI C standard's %g
format string - does anyone particularly object to that idea?
- Andrew
--
Dear God, I didn't think orange went with purple until I saw
the sunset you made last night.  That was really cool. - Caro

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System