[Hps-analysis] datacat example

[McCormick, Jeremy I.]

Thanks for the detailed feedback, Maurik.

> 1.	If I add up Single0+Singe1+Pair0+Pair1+Pulser, I would expect to recover the number of events in the file. There are 261442 events in the file, but the output shows only 5893 Single0 triggers. I think there is something wrong here.

There is something wrong here.  The bit masking I'm using to detect the trigger types is maybe not working right.  I'll look into it.

> 2.	It would be nice to re-order the output. The Trigger info should be listed in order, right below each other: Single0,Single1,Pair0,Pair1, the first and last physics events should follow each other, scaler banks and epics events should follow each other.

I'm not sure it is possible to do this in the generic web interface that is provided.  Internally, these key-value pairs are essentially unordered on the keys.  I know alphabetical sorting is being added in a future version to this browser interface.  

I can ask about it.  Or we can make our own custom webpages based on the output from the datacat.

> 3.	The only time you now show is when the data was crawled. It is much more relevant when the data was taken, i.e. timestamp of the first event and timestamp of the last event. Perhaps also the timestamps of the prestart and go events if they are present, but less important.

These are all present...

LAST_PHYSICS_EVENT - last physics event number
FIRST_PHYSICS_EVENT - last physics event number
FIRST_TIMESTAMP - first timestamp in file
LAST_TIMESTAMP - last timestamp in file

The first timestamp will come from GO if present and the last from END if present.  They could be split out into separate fields if necessary, but that mean they would be '0' for most data files.

> 4.	As this is a data catalog, it would be great if there were links to the processed datafiles and the DQM info, if available, and some status indicator if these are not available. I hope that such info would then allow us to query which files have or have not been processed already. 

I was planning to put links on the run database web frontend to all the relevant EVIO, LCIO, ROOT DST and DQM files.  These would however not show individual file associations.

In order to tell if an EVIO file has been processed, I believe you'd want to check for the existence of its corresponding LCIO file in the datacat, not put a flag on the EVIO file itself.  Because there are multiple passes, so that approach doesn't work.

> 5.	Also, will there be a way to incorporate the MC output into this data catalog? There, some query features would be very useful. MC data would need to include what the type of MC it was (tri-trig-beam, A’, pure-moller, etc etc) 

Yes, most definitely we can do this.  If the relevant metadata cannot be assigned automatically by reading through the file, then it can be assigned later in a script via the python API to the datacat.

> This is a nice start, but I don’t think we are close to a re-crawl. Would it be more efficient to combine crawling with the next pass of cooking?

They are entirely separate processing steps so I'm not sure that buys us anything.

I'm continuing to work on this in the meantime....

	On Nov 7, 2015, at 10:26 AM, Graf, Norman A. <ngraf at slac.stanford.edu> wrote:

	Jeremy,
	I'd like to see formal sign-off from the principals before proceeding to crawl through all the data again. Nathan, Matt, can you arrange this?
	Thanks,
	Norman
	________________________________________
	From: hps-software at slac.stanford.edu <hps-software at slac.stanford.edu> on behalf of McCormick, Jeremy I. <jeremym at slac.stanford.edu>
	Sent: Friday, November 6, 2015 3:35 PM
	To: hps-software
	Cc: hps-analysis at jlab.org
	Subject: datacat example
	
	Hi,
	
	I'm close to starting the Auger jobs that will populate the JLAB datacat with the EVIO files from the Engineering Run.
	
	Here is an example...
	
	http://hpsweb.jlab.org/datacat/display/datasets/HPS/data/raw/hps_005688.evio.0
	
	The built-in metadata fields are:
	
	file format - always set to 'EVIO'
	
	data type - always set to 'RAW'
	
	size - file size on disk
	
	master site - always set to 'JLAB'
	
	master resource - path of EVIO file on MSS
	
	run min/max - set to the file's run number
	
	events - count of physics events (does not include control or EPICS events)
	
	The per file custom metadata fields I have added thusfar are the following:
	
	BAD_EVENTS  - number of bad (unparseable) events
	
	BLINDED - flag if file is blinded
	
	END - flag if END event exists
	
	EPICS_EVENTS - number of EPICS events
	
	FILE - file sequence number
	
	FIRST_PHYSICS_EVENT - first physics event ID
	
	FIRST_TIMESTAMP - first timestamp
	
	GO - flag if GO event is present
	
	LAST_PHYSICS_EVENT - last physics event number
	
	LAST_TIMESTAMP - last event timestamp
	
	LED_COSMIC - number of cosmic triggers
	
	PAIRS0  - number of PAIRS0 triggers
	
	PAIRS1 - number of PAIRS1 triggers
	
	PRESTART - flag if PRESTART event exists
	
	PULSER  0 - number of pulser triggers
	
	SCALER_BANKS - number of scaler banks found
	
	SINGLE0 - number of SINGLE0 triggers
	
	SINGLE1 - number of SINGLE1 triggers
	
	Please tell me if there is any other metadata that should be included here.  I would prefer not to go back and assign additional metadata later unless it becomes absolutely necessary.  (It would have to be done in a separate step or the files would need to be deleted/re-added.)
	
	Also, if you want any of these metadata field names modified for clarity of meaning, let me know.
	
	Thanks.
	
	--Jeremy
	
	########################################################################
	Use REPLY-ALL to reply to list
	
	To unsubscribe from the HPS-SOFTWARE list, click the following link:
	https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1
	
	_______________________________________________
	Hps-analysis mailing list
	Hps-analysis at jlab.org
	https://mailman.jlab.org/mailman/listinfo/hps-analysis