Re: Long string command to device

[Bo Jakobsen <lister@f77.dk>]

I finally got a chance to test the suggested solutions.

I now have a record reading:

record(waveform, "$(P):RAW2") {
  field(DTYP, "stream")
  field(FTVL, "CHAR")
  field(DESC, "RAW waveform lakeshore record")
  field(INP, "@Huginn.proto setRAWOUT $(PORT) $(ADDR)")
}

however, when I try to send a value from caput I get the following error:

[labuser@localhost m-epics-Huginn]$  caput Huginn:RAW2 "A"
Old : Huginn:RAW2                    0
New : Huginn:RAW2                    0
CA.Client.Exception...............................................
    Warning: "Channel write request failed"
    Context: "op=1, channel=Huginn:RAW2, type=DBR_STRING, count=1, 
ctx="Huginn:RAW2""
    Source File: ../oldChannelNotify.cpp line 160
    Current Time: Mon Sep 26 2016 10:52:43.487194595
..................................................................


It is consistent with the manual for waveform records which reads
"The values retrieved from the input link are placed in an array 
referenced by VAL. (If the INP link is a constant, elements can be 
placed in the array via dbPuts.)"
So in conclusion I am stuck on how to use waveform records for sending 
long strings via streamDevice
Any suggestions are welcome

    Best
         Bo



On 2016-09-21 12:00, Mark Rivers wrote:
> Your proposed solution seems reasonable, and it should work. You should be able to use caput to write to the record when its INP field is configured for StreamDevice. If you get an error message when attempting this please send it to tech-talk because it must be some other problem.
>
> Mark
>
>
> Sent from my iPhone
>
> > On Sep 21, 2016, at 3:52 AM, Bo Jakobsen <lister@f77.dk> wrote:
> >
> > Thanks for the comment. My use case is a bit clumsy.
> >
> > I am working on a project using the existing LakeShore temperature controller EPICS module.
> > Unfortunately this module does not implement uploading temperature calibration curves to the controller.
> >
> > As stated earlier I am rather new to EPICS (and EPICS implementation is not my main job) so doing a full implementation of the curves uploading mechanism seemed to big a job.
> >
> > Inspired by the streamDevice manual, I just implemented a "RAW" record which forwards strings to the apparatus.
> > The  control strings are then generated in the scripting language I use (Matlab) and send over CA. This works fine for this special case.
> > However, for defining the "curve header" the string can get longer than the 40 characters allowed for stringout, as it contains "names" and "serial numbers" to be associated with the curve. Therefore the need for a longer string.
> >
> > I will try ones more with the waveform record. However, I seems to remember that if I associate the streamDevice with the INP field, I could not write to the record using "caput" (which would be consistent with the manual)
> >
> > I total agree this is not a pretty solution, but it could be very useful in prototyping work.
> >
> > Best
> >     Bo
> >
> >
> >
> > > On 2016-09-20 17:40, Rod Nussbaumer wrote:
> > > Bob:
> > >
> > > I've never found a need to use string data stored in any kind of array record or string record for output to a device using streamDevice. Normally, the strings that are sent to a device are static command or query strings embedded as format strings in a streamDevice protocol file. streamDevice will send these strings, perhaps modified by data expanded from an EPICS record, to the device. So, no need to stuff string data into array records, which is a bit clumsy to do.
> > >
> > > Do you have a different use case that dictates that you must send string data that originates in a record?
> > >
> > > Note also that streamDevice essentially ignores the input/output nature of IO records. You can send the value of an input record to a device, and you can fetch the value of a device parameter into an output record.
> > >
> > > Rod Nussbaumer
> > > TRIUMF,
> > > Vancouver, Canada
> > >
> > >
> > > > On 09/20/2016 04:10 AM, Mark Rivers wrote:
> > > > Hi Bo,
> > > >
> > > > The waveform record can be used for both input and output.
> > > >
> > > > This is from the documentation for StreamDevice for the waveform record:
> > > >
> > > > ************************
> > > > If FTVL=="CHAR" or FTVL="UCHAR":
> > > >      In this case, the complete waveform is treated as a large single string of size NORD. No separators are printed or expected.
> > > >      Output: x=range(VAL,0,NORD)
> > > >         The first NORD characters are printed, which might be less than NELM.
> > > >      Input: VAL=x, NORD=length(x)
> > > >          A maximum of NELM-1 characters can be read. NORD is updated to the index of the first of the trailing zeros. Usually, this is the same as the string length.
> > > > ************************
> > > >
> > > > Mark
> > > >
> > > > ______
> > > > __________________________________
> > > > From: tech-talk-bounces@aps.anl.gov [tech-talk-bounces@aps.anl.gov] on behalf of Bo Jakobsen [lister@f77.dk]
> > > > Sent: Tuesday, September 20, 2016 3:34 AM
> > > > To: tech-talk@aps.anl.gov
> > > > Subject: Long string command to device
> > > >
> > > > Hi,
> > > >
> > > > I am working on modifying an EPICS interface for a LakeShore 336 (LS336)
> > > > temperature controller.
> > > > In that connection I need to send strings to the LS336 which extends 40
> > > > characters.
> > > >
> > > > I found the following thread:
> > > > http://www.aps.anl.gov/epics/tech-talk/2016/msg00887.php
> > > > which discusses a similar problem, and suggests to use waveform records
> > > > and StreamDevice (which I am already using).
> > > >
> > > > However, I am unable to get this to work.
> > > > As I read the documentation of waveform records they are only for input,
> > > > and hence does not allow to set the OUT field (I am using EPICS 3.14)
> > > >
> > > > Can anyone help with a solution to this problem?
> > > >
> > > > Best
> > > > Bo Jakobsen