Grouper Users - Open Discussion List

[Julio Polo <>]

Hi Nathan,

Yes, I ended up doing something similar.  Our users can build their groups from a menu of institution-defined groups which include departments, faculty, staff, students at specific campus, programs of study, and so on.  I cannot predict the complexity of our user-defined groups, so I ended up created a recursive method that parses a group _expression_ and creates each intermediate group on the fly.

Julio Polo
University of Hawaii

On Fri, Sep 20, 2013 at 4:34 AM, Nathan Weyenberg <> wrote:

Hi Julio, 

It's funny you bring up automatically creating hidden groups for composite groups, because we are implementing this for our grouping system (Manifest) that utilizes Grouper. We essentially allow a different group type (controlled by attribute at this moment) that allows a user to see the final rollup group and all of the rollup and leaf/terminal groups that make up its membership all in one screen. This screen will also be fully editable and change the appropriate membership when the logical structure was changed. We plan on calling these 'complex' groups. 

How we have decided to do it is by creating an _admin folder in every folder created. In here, we plan on leveraging the uniqueness of a groups name and using a naming convention in the admin folder; every 'hidden' and all rollup composite groups will be created here with the naming convention of 'mygroup_composite#'. 

--Nathan Weyenberg

University of Wisconsin - Madison

On Sep 19, 2013, at 11:18 PM, Julio Polo <> wrote:

Hi Chris,

Sorry I'm chiming in late here. I understand all the technical reasons you present for not implementing union, but from a functional point of view, it would be nice not to have to create an intermediate group that won't be used (outside of being a composite factor, of course).

We are planning to provide a service whereby every group (from the end-user's point of view) will be defined as a composite C = ((X union Y) complement Z).  It would be nice not to be have to create thousands of intermediate groups for X+Y for each of the thousands of C composite groups we expect to implement.

Like Brad, I was hoping that we could define a composite group with more than two factors.  Perhaps the key to this is for Grouper to automatically create hidden groups as needed.  For my situation above, Grouper could automatically create a hidden group with X and Y as members when it is parsing a composite group definition.  Creating hidden groups on the fly would allow you to have more than two factors on any composite group without giving up the efficiency you want.

Thanks,

Julio Polo

University of Hawaii

On Tue, Sep 3, 2013 at 3:50 PM, Chris Hyzer <> wrote:

well, someone else from the list can chime in if I am missing something, but from my experience, a union has no advantage over a group as a member.  but a group as a member has the advantage that it is implemented such that each membership does not require another distinct row in the membership table, and of course, you dont have use a composite with just two factors, you can add as many groups or individuals to a group as you want.  For intersection and minus composites, more membership table rows is the only way to implement it efficiently.  Therefore, if we were going to design Grouper from scratch, I would recommend not having a union composite.  It was implemented originally probably because it could be.  It still exists just because we havent written a migration upgrade to get away from it.  Maybe we should have a switch so no new groups can be composite unions...

Thanks,

Chris

---

From: Omaraie, Brad []
Sent: Tuesday, September 03, 2013 8:02 PM
To: Chris Hyzer
Cc:
Subject: Re: Composite group with more than two groups contained

Hi Chris,

Yes, we are talking union here. I tested your suggestion and it worked just fine. But when I brought up the subject with our management team here, they wanted to know what is the difference between union and group of groups then? If they are the same, why do we have them separately? In other word, what are the pros and cons of using group of groups solution vs cascaded composites?

Thanks,

Brad

From: "" <>
Date: Monday, September 2, 2013 8:08 PM
To: ucla <>
Cc: "" <>
Subject: RE: Composite group with more than two groups contained

Are you talking about union?  If it is union, you should never use union, just add the groups as members of the parent group.  Ok?

 

Are you talking about intersection?  If it is intersection, then you either need a lot of composites, or you need rules.

 

Thanks,

Chris

 

From: Omaraie, Brad []
Sent: Thursday, August 29, 2013 6:29 PM
To: Chris Hyzer
Subject: Composite group with more than two groups contained

 

Hi Chris,

Just a quick question. We need to create a composite group for one of our clients that has about twelve groups included in it. By that I mean a member of each of these twelve groups should also be a member of our composite group. Currently grouper UI allows us to create a composite group from a logical combination of two groups at a time. I know we can cascade multiple composite groups and finally get to a group that has all the members of the twelve groups in it, but I was wondering is there a way that we can directly create only one composite group  containing all these groups?

 

Thanks,

Brad

 

Nathan Weyenberg

UW-Madison DoIT - Middleware