Re: [Gnumed-devel] user contributed report

[Tim Churches]

Karsten Hilbert wrote:
> On Wed, Apr 11, 2007 at 05:37:39AM +1000, Tim Churches wrote:
> 
>>>>> For that we'd have to add post-processing to query results
>>>>> and import an R binding. Which can be done but needs work to
>>>>> get done.
>>>> That's not too hard. I did this for my thesis. Just use rpy bindings and 
>>>> you 
>>>> can manipulate data all you want. Just have a script run the queries and 
>>>> work 
>>>> with the data. used it to calculate some statistics and plot the data with 
>>>> gnuplot.  
>> Surely if you transfer the query results to R you would plot the results
>> with R? It has much better statistical features than gnuplot.
> Oh, the whole thing was scripted. What Sebastian did was
> pipe data from PostgreSQL into R for doing some statistics.
> The *results* were piped into gnuplot for plotting only.
> 
>> I should also point out NetEpi analysis, which uses R via Rpy for these
>> things. Perhaps an automated GNUmed importer can be developed - we are
>> working on that for other data sources (currently you have to write an
>> import script by hand).
> That's of course the real thing to do. *My* plan is not to go
> much beyond what we currently have in terms of analysis
> features. The report plugin will stay what it is - a tool to
> do "fairly simple" reports. If a clever user augments that
> by pimping PostgreSQL with pl/R, fine, but for anything
> sophisticated I suggest using any number of third party
> report generators -- such as NetEpi ;-)
> 
> Writing another one makes as much sense as writing another,
> say, Plone clone.

There is a role for built-in often-used pre-defined reports, as well as
export of data to a periodically-refreshed "data warehouse" for more
serious ad-hoc reporting and analysis work.

Tim C