Hi, Holly.
There's a utility Driver you can use now for copying any cluster collection into a new one via your steering job....
<driver name="CopyClusters" type="org.hps.recon.ecal.cluster.CopyClusterCollectionDriver">
<inputCollectionName>EcalClusters</inputCollectionName>
<outputCollectionName>EcalClustersCopy</outputCollectionName>
<storeHits>true</storeHits>
</driver>
The "storeHits" argument controls whether or not the hits in the new collection are persisted in the LCIO output, but it does not affect the in-memory hit list which is always copied to the new cluster. (You can just ignore that hit list in your Driver if you don't need to use it.)
I added a copy constructor to BaseCluster in lcsim in order to more easily support this. The data structures (lists and arrays) from the source clusters are themselves copied to avoid unwanted direct references to objects in the original cluster. So you should be able to change the data in the new cluster collection including all its lists and arrays, without it affecting the old clusters. For energy and other primitives this wouldn't be a problem as these are not referenced objects. The hits themselves are *not* copied so they remain references but are put into a new list. I hope that all makes sense.
I verified in a simple test that I can run this Driver using the LCSimLoop on some recon output and get the output clusters with and without the hits stored, so it seems to work fine.
Can you confirm if this works for you?
--Jeremy
-----Original Message-----
From: Holly Vance [mailto:[log in to unmask]]
Sent: Friday, May 22, 2015 12:49 PM
To: McCormick, Jeremy I.
Subject: Re: collection issue
Thanks!
On Fri, May 22, 2015 at 3:43 PM, McCormick, Jeremy I. <[log in to unmask]> wrote:
There is an added complication here, because you can't just copy the Cluster objects into a new list. You need to perform a more low-level copy that actually makes new lists/arrays/etc. in the new object because otherwise the old objects are referenced directly from the new object, and you will really confuse yourself e.g. you'll modify something in the new cluster and it will also affect the old one!
I can handle this though...just give me a day or two to code something up and put it into the BaseCluster class which will handle this correctly. Then writing the copy Driver will be straightforward.
-----Original Message-----
From: Holly Vance [mailto:[log in to unmask]]
Sent: Friday, May 22, 2015 12:29 PM
To: McCormick, Jeremy I.
Subject: Re: collection issue
Is it just not possible to be done in the ClusterDriver? I think the way I wrote it in there, it should work, but for some reason it always outputs only 1 collection. Do you know why that behavior happens? It's not clear to me.
On Fri, May 22, 2015 at 3:25 PM, McCormick, Jeremy I. <[log in to unmask]> wrote:
ClusterDriver probably has enough functionality in it already without adding more to it....
I would keep it separate and have a new CopyClusterCollectionDriver that does this which can be run after the clustering is done. It could have the input and output cluster collection names. That's a more flexible way to do it.
-----Original Message-----
From: Holly Vance [mailto:[log in to unmask]]
Sent: Friday, May 22, 2015 12:23 PM
To: McCormick, Jeremy I.
Subject: Re: collection issue
Ok I thought it would be simlpler to add this feature to overall clustering.
I made my changes to ClusterDriver in my users folder. I can move it to a new driver, but I strongly feel like this is a feature that should not become too complicated. It makes sense that when you run a particular cluster driver, you could choose to output it its results to another collection for use in cluster track matching (effectively choosing this cluster collection).
On Fri, May 22, 2015 at 2:53 PM, McCormick, Jeremy I. <[log in to unmask]> wrote:
I think what you actually want to do here is not directly modify ClusterDriver itself but add a downstream Driver which knows the original cluster collection name and can read/copy it. That way we can leave ClusterDriver alone.
-----Original Message-----
From: Holly Vance [mailto:[log in to unmask]]
Sent: Friday, May 22, 2015 11:46 AM
To: McCormick, Jeremy I.
Subject: collection issue
Hi Jeremy,
I am trying to add a second cluster collection output in the simplest way possible such that it is a copy of what is in ReconCluster output and the position and energy information can be modified at cluster-track matching without hurting the orignal information.
So I added a new boolean variable to ClusterDriver and a new cluster collection output name.
Where we write: event.put ()
I added an if statement after this (if boolean, then) where this same cluster list can be added to this new output collection.
Unfortunately, it will only write one cluster collection when the driver is run. I don't know where this behavior is restricted, and I was wondering if you had advice.
-Holly
--
--
--
BEGIN-ANTISPAM-VOTING-LINKS
------------------------------------------------------
NOTE: This message was trained as non-spam. If this is wrong,
please correct the training as soon as possible.
Teach CanIt if this mail (ID 01Ov7J7w5) is spam:
Spam: https://www.spamtrap.odu.edu/canit/b.php?i=01Ov7J7w5&m=bd9621efc037&t=20150522&c=s
Not spam: https://www.spamtrap.odu.edu/canit/b.php?i=01Ov7J7w5&m=bd9621efc037&t=20150522&c=n
Forget vote: https://www.spamtrap.odu.edu/canit/b.php?i=01Ov7J7w5&m=bd9621efc037&t=20150522&c=f
------------------------------------------------------
END-ANTISPAM-VOTING-LINKS
########################################################################
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
