Experimental Physics and
Industrial Control System

Dirk Zimoch <[email protected]> · Thu, 14 Jun 2012 13:46:32 +0200

Hi Judy,

Rock, Judith E. wrote:

Hi -
We have a device that returns multiple lines of input from one command.  We're successfully using streamdevice to capture the inputs, which are of mixed type.  The issue at hand is error handling for the float inputs.  Upon error, any of the float fields can return an error string.  I tried setting up an @mismatch for the prototype, and this works...the problem is that the prototype stops processing inputs at the error: subsequent inputs in the list are all invalid which could mask some important conditions.  I also tried a separate prototype for each input with its own @mismatch; however I observed the same behavior: inputs following the error condition were all invalid.

This is indeed a problem yet unsolved with redirection. The protocol

stops upon error and only the active record becomes invalid.

Our solution at the moment is to read everything in as strings and use scalcouts to translate to numbers - that way each input is converted independently.  However, if streamdevice itself has a more elegant way of handling the situation we would be very happy to adopt it!

I can offer you two other solutions:

1) Use "I/O Intr". Like before, you have one active record. But you

don't redirect other values, you simply ignore them. The other records

are "I/O Intr". Their protocol consists only of an 'in' command and they

parse their input and ignore everything else. So if you have a format

like "%*s %*s %f %*s %*s" i.e. read the 3rd of 5 words as a float and

you get an error message as word 3 then this record will simply ignore

the input. If the error strings are enums (i.e. limited set of strings),

you can have the error record for this value trying to read the message

in parallel to the "normal" float input: "%*s %*s

%{error1|error2|error3} %*s %*s".

2) Use a new feature: optional input. With a format like "%?f" you can

read a float or get 0.0 if there is no float.

Something like this:
in "%?f%(e1)?{error1|error2|error3} "
   "%(v2)?f%(e2)?{error1|error2|error3}";

If you can give me some example input, I may give you a more detailed

answer.

Drik

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System