Re: Channel Archiver

[Kay-Uwe Kasemir <kasemirk@ornl.gov>]

On May 16, 2007, at 16:20 , Szalata, Zenon M. wrote:
> I am learning how the SNS Channel Archiver works.  I am using  
> Vlinac as channel server.  In particular I am archiving the  
> cathodeTempM quantity.  Typical value is 156.6 degF.  When I look  
> at the index file with ArchiveExport tool the value is in an  
> undesirable format, as follows:
>
> 05/16/2007 12:16:20.202837000   2e+02
>
> I am pouring through the Channel Archiver Manuals
Always a good idea.

> for information on how to change the format of the data.  Actually,  
> I do not know if the data is written to the data file in this  
> uninformative format
As explained in the manual, the raw data as received from Channel Access
is written. For example a 'double' value.
Together with meta information like suggested display precision etc.

> or it is the ArchiveExport tool that presents the data in that format.
Yes, it's a matter of presentation.

And there the problem is:
If ArchiveExport (or any other tool that looks at the data)
actually uses the display precision received via ChannelAccess,
about 45% of all users are unhappy because they never configured
the PREC fields of the records, and then miss the digits after
the decimal point.

If ArchiverExport on the other hand just uses some default precision,
the other 45% of all the users who actually configured PREC are unhappy.

Then there are the remaining 10% who deal with vacuum readings
and want exponential notation, or other cases where the meta data
received via ChannelAccess is insufficient.

The exact percentages change every few month. As did ArchiveExport.
Recent versions use the printf "%g" format as the default,
because at some point in time that seemed like a good idea.

Depending on your version, there should be some command-line switches
to select various more specific output formats as well as an option  
to override the
default precision:

USAGE: ArchiveExport [Options] <index file> {channel}

Options:
...
  -format <decimal|engineering|exponential>  Use specific format for  
numbers
  -precision <int>                           Precision of numbers
...

-Kay