|
From: Greg W. <gr...@sl...> - 2012-01-23 22:25:18
|
EPICS V4 Meeting Minutes 23-Jan-2012 (Monday)
======Gather Service, MK======
Impl: 2 parts: createGather, gather
AJ: blocking a call?
GS: createGather to create an instance of a new V4 record to do gather job
Gather methods:
get() simply gets present (unrecord-processed) values
processGet() process records before get
createGather takes 2 seconds is caused by 2 seconds hardcoding. It does not really take so long time, and can be solved easily.
ISSUE: should gather diff language implementations be in one repo or separated repo?
ISSUE: gather in pvService or elsewhere?
Thinking about this tonight, and discuss it tomorrow
ISSUE: should the data package for Gather be NTTable NTMultiChannelArray, or both?
Client Gather API in Matlab, cothread, API like that
return a vertor of double
Looking to cothread code, DBR type support already in cothread (but cothread is based on Python which supports putting heterogenous data into lists)
separate ezpva and gather
ezpva in C++ first
AI - MK by ?? :complish gather doc, make a simple complete example: orbit service based on gather. (GS, MK, LY) [First just publish a simple written example of what a user would type to use the gather library to acqure the values of 3 (homogenous) V3 PVs (just literally the 10 minute job of writign an example in the document), then do the bigger implementation example.
RESOLUTION: ezpva should be developed independently of Gather requirements.
Gather functionality implemented as a library should be published first and is our priority. Gather as a service will come later.
PD: What about atomaticity and synchronization of acquisitions?
GW: The gather library in itself doesn't have support for synchronization. It would be up to a service that uses the Gather library to ensure that what is returned to a user is synchronized data, using whatever facilities exist at each institution.
NEW TOPIC: Normative Types review
NTNameValue
structure NTxxx
string ref (Greg agrees this is a crappy name)
alarm and timestamp: returned value should always have this setting correctly.
modify section 4: Services SHOULD provide alarm and timestamp information
NTMutiChannelArray: have a pvname [] field
string [] pvnames
NTTimeDomainArray and NTFreqDomainArray, the value should be double array
Change structure NTHistogram
rename long[] values => long[] counts
enum_t => double [] binIntervals
double firstValue => gone
double lastValue => gone
NTImage: Simple display tools should check colorMode to ensure that they can actually display a particular image.
variant_t as a field
variant_t variant_field
BD: write Mark River an email whether NTImage is suitable
=======unsigned integer=====
MK: present only
===end of Monday===
|