Indexes

[George James]

George,
Further to your ponderings about what indexes are needed, I gleaned
the following from the AllegroGraph website (http://www.franz.com/
agraph/allegrograph/):

Triple-indices are user configurable, or index management can be taken
care of entirely by AllegroGraph. By default, all committed triples
are always indexed (default: 7 indices). AllegroGraph now supports any
index combination of S, P, O, G. The default indices are:

S, P, O, G, I - Subject, Predicate, Object, Named Graph, ID
P, O, S, G, I
O, S, P, G, I
G, S, P, O, I
G, P, O, S, I
G, O, S, P, I
I

I don't know if this helps any, but it's interesting to me that
Allegro allows indexing to be user configurable, which suggests that
there is sometimes a need to optimize the storage. I wonder if this
optimization is per graph or per graph database...

George

[Tom Munnecke]

I ranted a bit about semantic web at
http://www.osehra.org/discussion/visual-cross-reference-packages-routines-globals-now-available#comment-81

I'm very interested in getting smarter in all this, and am
particularly looking for a sandbox I can either install or play
with... is anyone connecting http://www.smartplatforms.org/ to VistA?

I'm still fuzzy how one extracts meaning from triple stores... for
example, in http://dbpedia.org/page/The_Lord_of_the_Rings, how does
one extract a given language from the list? do you have to add a
SparQL term for language? what if that tuple is missing - does it
then get included?

Where does the time of observation fit into a triple store of say, lab
data. Is it held in another tuple, or linked somehow?

This is all very interesting to me, but there is quite a steep
learning curve. I would appreciate anything folks can do to help me
with it. I'm wondering if I scheduled a Google Plus hangout, anyone
would be interested in talking with me?

[George Lilly]

It is interesting (and to me daunting !) that they let the users specify what indexes there will be.. I know fileman does that, but it seems to me that it's hard to do.

Yikes! I just looked at the indexes I put on the initial F2N prototype and they are probably wrong... For one thing, I forgot the PSO and POS indexes completely, and I've left off the ID completely from the indexes for bulkload. 

Here's what we currently have:

SPO

SOP

OSP

OPS

GSPO

GSPO

GOSP

GOPS

G

B (which is the index to the node)

I need to add the following I think:

POS

PSO

GPOS

GPSO 

[George James]

Are indexes that span graphs actually useful?  I can see that GPOS would be useful, but what kind of query would span a graph and benefit from a bare POS index?

 

---

From: fileman-tr...@googlegroups.com [mailto:fileman-tr...@googlegroups.com] On Behalf Of George Lilly
Sent: 23 November 2011 17:54
To: fileman-tr...@googlegroups.com
Subject: Re: Indexes

[George Lilly]

I think the indexes that cross all graphs will be very powerful to have.. here's a couple of examples that come to mind:

if there are a number of graphs for each patient's summary clinical data (ala smart of hData).. and one wants to look across all patients for certain codes (all patients with this LOINC code, or RxNorm drug, or Snomed Code. for example). it would probably be theses indexes you would use. This might have a practical application in the computation of quality measures based on the qds (quality data set).. 

if certain types of encounters are recorded as graphs, it will be the cross-graph indexes that will provide answers to questions like "show me all encounters for a given patient" or "show me all encounters for a given provider" 

gpl

[George Lilly]

Tom:

I'd love to join any Google+ group you start to talk about these things. 

Also, with regard to "Smart Enabling VistA".. we are focused so far on the incoming scenarios .. the ability to use Smart Enabled incoming data in VistA. To implement a Smart Container with VistA, we would need to work on the outgoing scenarios. If we can get far enough on the path we are on, the infrastructure we are building would certainly be very useful in that effort. 

So... are we working on Smart-enabling VistA? Yes and no.

gpl

[Tom Munnecke]

Well, I'll be hanging out on Google+ at 9AM PST this Friday, nov 25... you'll need a video connection to Google, and I think we need to become acquaintences on Google + to make it happen.  

It seems like connecting to Smart would be a good demo of the technology and ideas...

[Tom Munnecke]

I've installed the Allegro Graph program, and played with the Gruff program... 

I'm stuck trying to generate my own RDF, though.  I am thinking of taking the XINDEX file and turning it into a file of programs and what they are called by... thought it would be a fun file to play with.  any suggestions?

[glilly]

Tom:

I can think of three approaches to generate the RDF file you want:

1. modify XINDEX to do insertions to the triple store (ADD^C0XF2N and
UPDIE^C0XF2N) and then run the yet-to-be-written rdf-out routine to
generate an rdf file from a triple store graph.

2. modify XINDEX to populate the MXMLDOM and then use the kernal
routines to generate XML from the dom. I wrote a recursive routine to
do this which you can just copy.

the kernal XML output routine is here:
https://trac.opensourcevista.net/browser/ccr/trunk/p/C0CMXMLB.m
(I put it in the C0C namespace because at the time we were not up to
patch level to install the kernal version)

the routine I wrote to use it to generate XML is here:
(OUTXML^C0CMXML)
https://trac.opensourcevista.net/browser/ccr/trunk/p/C0CMXML.m

3. Don't bother populating the DOM and just modify XINDEX to output
the rdf xml directly using the kernal utilities (C0CMXMLB.m as above)

Hope this helps.

gpl

On Nov 23, 10:44 pm, Tom Munnecke <munne...@gmail.com> wrote:
> I've installed the Allegro Graph program, and played with the Gruff
> program...
>
> I'm stuck trying to generate my own RDF, though.  I am thinking of taking
> the XINDEX file and turning it into a file of programs and what they are
> called by... thought it would be a fun file to play with.  any suggestions?
>
>
>
>
>
>
>
> On Wed, Nov 23, 2011 at 12:48 PM, Tom Munnecke <munne...@gmail.com> wrote:
> > Well, I'll be hanging out on Google+ at 9AM PST this Friday, nov 25...
> > you'll need a video connection to Google, and I think we need to become
> > acquaintences on Google + to make it happen.
>
> > It seems like connecting to Smart would be a good demo of the technology
> > and ideas...
>
> > On Wed, Nov 23, 2011 at 12:15 PM, George Lilly <gli...@glilly.net> wrote:
>
> >> Tom:
>
> >> I'd love to join any Google+ group you start to talk about these things.
>
> >> Also, with regard to "Smart Enabling VistA".. we are focused so far on
> >> the incoming scenarios .. the ability to use Smart Enabled incoming data in
> >> VistA. To implement a Smart Container with VistA, we would need to work on
> >> the outgoing scenarios. If we can get far enough on the path we are on, the
> >> infrastructure we are building would certainly be very useful in that
> >> effort.
>
> >> So... are we working on Smart-enabling VistA? Yes and no.
>
> >> gpl
>

> >> On Wed, Nov 23, 2011 at 12:51 PM, Tom Munnecke <munne...@gmail.com>wrote:
>
> >>> I ranted a bit about semantic web at
>

> >>>http://www.osehra.org/discussion/visual-cross-reference-packages-rout...

>
> >>> I'm very interested in getting smarter in all this, and am
> >>> particularly looking for a sandbox I can either install or play

> >>> with... is anyone connectinghttp://www.smartplatforms.org/to VistA?

>
> >>> I'm still fuzzy how one extracts meaning from triple stores... for

> >>> example, inhttp://dbpedia.org/page/The_Lord_of_the_Rings, how does

[Tom Munnecke]

hmmmm... this probably exceeds my skill level/patience for figuring this all out :( 

can you describe what an RDF output of the XINDEX would be?