[SoC] SoC report: Geo-BI dashboards

[Etienne Dube]

Hi everybody,

1. What did I do this week:

- Implemented the serialization of OLAP CellSets. The result of a query 
is returned as a JSON object to the client as part of the response 
(along the GeoJSON data). This will allow other dashboard components 
(crosstabs, charts) to display another view of the data (synchronized 
with the map view).

- Investigated a way to implement OLAP operators. It turns out that 
there isn't a way in olap4j to manipulate and transform an existing MDX 
query. This is planned as part of the final specification, but hasn't 
been implemented yet. This is a major blocking point, and I will have to 
implement at least part of this functionality myself. I'm investigating 
the way JPivot (a Java/JSP based web OLAP client) is managing this for 
basic OLAP navigation (drill-down on a member and so on). However, 
JPivot is coupled direcly to Mondrian (it hasn't been updated to use the 
olap4j API) and isn't stateless (the server maintains a state with the 
current execution context of the client, i.e. current MDX query, result 
axes, etc. and sets a cookie on the client with the session ID). I would 
prefer a stateless way of doing this, even if the client has to submit 
some information about its current context on each invocation of the 
servlet. Moreover, I want to keep using the olap4j API because 
Mondrian's native API is on the way to be deprecated. This task will 
require serious efforts to overcome, and will be my sole focus for the 
next two weeks (and possibly more). This is an absolute requirement to 
get a working application which is generic and usable with any data cube.

I have discussed this problem with my mentor and we have decided to bite 
the bullet and work on solving this, as it will enable the development 
of new OLAP operators in a faster and easier way. It could also be an 
important and welcome contribution to the olap4j API (www.olap4j.org). 
However, the time needed for this task means that other planned features 
(more mapping styles, legends, ...) will probably not be fully 
implemented before the end of SoC.

2. What do I plan on doing next week:

- Continue working on a way to implement MDX query transformations and 
OLAP operators.

3. Am I blocked on anything:

- The problem with OLAP operators/query transformations was unexpected, 
but it is not really blocking as I'm working on a solution. See above.

Etienne