What is Piet?

Piet is an open source  system  that integrates  Geographic Information Systems (GIS) and OLAP systems, named after the Ducth painter Piet Mondrian. The implemented model consists of application data contained in a data warehouse, and geographic data (a map with geometric elements stored in different layers). Geometric information that composes the map is organized in a set of hierarchies that conforms a GIS dimension. Additionally, elements from the geometric section are related to externally stored application data (usually, a data warehouse). Based on this model, we can answer queries aggregating both kinds of data. These queries are expressed over the geometric and application parts, allowing to navigate the query results using OLAP tools. A Technical Report on the data model of Piet can be found here

 

A simple SQL-like query language has been developed, based on a first-order formal query language. This language is called GISOLAP-QL, and  it is of the form:

GIS-query | MDX query.

The  GIS  part of the query  returns a list of object identifiers which are   passed to a Mondrian OLAP engine. The parser merges the objects returned by the GIS part of the query with an MDX expression in the OLAP part, and generates a query. The result of this query can be navigated in the usual OLAP fashion.

 

Piet supports four types of queries: (a) GISOLAP-QL  queries, like "Give me the total sales per product in all stores located in cities crossed by a river". This query involves different types of data, and its result can be further exploited using traditional OLAP operations. (b) Geometric aggregation queries, like "number of stores in states with more than three airports", that summarizes values associated to geometries. (c) Typical GIS queries. (d) Typical OLAP queries.

 

An off-line executed preprocessing stage was implemented for solving geometric aggregation queries. This process generates the so-called "subpolygonization" of the map, creating a set of subgeometries used for reorganizing the GIS dimension and for splitting original geometric components and their associated fact values into smaller objects (subgeometries) which can be shared between different components. The overlay of map layers that contains geometric components is also computed during this preprocessing stage. The idea of this method is borrowed from view materialization techniques traditionally used in relational data bases and   OLAP query optimization. Our tests showed that overlay precomputation is overcomes R-trees indexing techniques  when the query is performed over the entire map. However, when a query region is specified, R-trees perform better than precomputation. Our implementation (although currently based on precomputation) can support both kinds of techniques.

 

Piet was developed using Java programming language and open source technology for OLAP and GIS (such as Mondrian, JUMP and PostGIS). The solution includes three interfaces: a web interface, a JUMP interface and a file-configured interface.

 

 

Who has developed Piet?

Piet was first started as Ariel Escribano's undergrad thesis, supervised by Alejandro Vaisman. The current version of Piet was developed by

Leticia Gomez, who is a PhD candidate at Instituto Tecnologico de Buenos Aires and the University of Buenos Aires.   

 

The Piet project is a collaboration between the University of Buenos Aires and the University of Hasselt, funded by a bilateral agreement between FWO in Flandes and SECyT in Argentina (Project #FW/PA05-EXI/005, "GISOLAP: An Approach for Spatial Data Aggregation  based in  OLAP"), and a PICT project in Argentina, #21350, "Using OLAP techniques in GEographic Information Systems".

 

Do you want to see Piet Demo?

a) To run queries from the Web: Click here.

 

b) To install on your computer a pre-configured and limited Piet standalone demo: Follow these instructions.

 

 

Do you want to see experimental evaluation results?

a) With Belgium maps: Click here.

 

b) With maps of the northern hemisphere: Click here.

 

How can you get Piet?

Send an email to avaisman@dc.uba.ar