This project is an example of running MDX on Druid via Mondrian and Apache Calcite.
First, it creates a JDBC connection using Calcite's Druid adapter. Next, it generates a Mondrian model (assuming that numeric columns are measures and other columns are dimensions). Last, it connects to Mondrian and executes an MDX query.
Here is the Mondrian model that is generated:
<?xml version='1.0'?>
<Schema name='wiki' metamodelVersion='4.0'>
<PhysicalSchema>
<Table name='wikiticker'/>
</PhysicalSchema>
<Cube name='wiki'>
<Dimensions>
<Dimension name='channel'>
<Attributes>
<Attribute name='channel' table='wikiticker' keyColumn='channel'/>
</Attributes>
</Dimension>
<Dimension name='cityName'>
<Attributes>
<Attribute name='cityName' table='wikiticker' keyColumn='cityName'/>
</Attributes>
</Dimension>
<Dimension name='comment'>
<Attributes>
<Attribute name='comment' table='wikiticker' keyColumn='comment'/>
</Attributes>
</Dimension>
<Dimension name='countryIsoCode'>
<Attributes>
<Attribute name='countryIsoCode' table='wikiticker' keyColumn='countryIsoCode'/>
</Attributes>
</Dimension>
<Dimension name='countryName'>
<Attributes>
<Attribute name='countryName' table='wikiticker' keyColumn='countryName'/>
</Attributes>
</Dimension>
<Dimension name='isAnonymous'>
<Attributes>
<Attribute name='isAnonymous' table='wikiticker' keyColumn='isAnonymous'/>
</Attributes>
</Dimension>
<Dimension name='isMinor'>
<Attributes>
<Attribute name='isMinor' table='wikiticker' keyColumn='isMinor'/>
</Attributes>
</Dimension>
<Dimension name='isNew'>
<Attributes>
<Attribute name='isNew' table='wikiticker' keyColumn='isNew'/>
</Attributes>
</Dimension>
<Dimension name='isRobot'>
<Attributes>
<Attribute name='isRobot' table='wikiticker' keyColumn='isRobot'/>
</Attributes>
</Dimension>
<Dimension name='isUnpatrolled'>
<Attributes>
<Attribute name='isUnpatrolled' table='wikiticker' keyColumn='isUnpatrolled'/>
</Attributes>
</Dimension>
<Dimension name='metroCode'>
<Attributes>
<Attribute name='metroCode' table='wikiticker' keyColumn='metroCode'/>
</Attributes>
</Dimension>
<Dimension name='namespace'>
<Attributes>
<Attribute name='namespace' table='wikiticker' keyColumn='namespace'/>
</Attributes>
</Dimension>
<Dimension name='page'>
<Attributes>
<Attribute name='page' table='wikiticker' keyColumn='page'/>
</Attributes>
</Dimension>
<Dimension name='regionIsoCode'>
<Attributes>
<Attribute name='regionIsoCode' table='wikiticker' keyColumn='regionIsoCode'/>
</Attributes>
</Dimension>
<Dimension name='regionName'>
<Attributes>
<Attribute name='regionName' table='wikiticker' keyColumn='regionName'/>
</Attributes>
</Dimension>
<Dimension name='user'>
<Attributes>
<Attribute name='user' table='wikiticker' keyColumn='user'/>
</Attributes>
</Dimension>
<Dimension name='user_unique'>
<Attributes>
<Attribute name='user_unique' table='wikiticker' keyColumn='user_unique'/>
</Attributes>
</Dimension>
</Dimensions>
<MeasureGroups>
<MeasureGroup table='wikiticker'>
<Measures>
<Measure name='__time' aggregator='sum' column='__time'/>
<Measure name='added' aggregator='sum' column='added'/>
<Measure name='count' aggregator='sum' column='count'/>
<Measure name='deleted' aggregator='sum' column='deleted'/>
<Measure name='delta' aggregator='sum' column='delta'/>
</Measures>
<DimensionLinks>
<FactLink dimension='channel'/>
<FactLink dimension='cityName'/>
<FactLink dimension='comment'/>
<FactLink dimension='countryIsoCode'/>
<FactLink dimension='countryName'/>
<FactLink dimension='isAnonymous'/>
<FactLink dimension='isMinor'/>
<FactLink dimension='isNew'/>
<FactLink dimension='isRobot'/>
<FactLink dimension='isUnpatrolled'/>
<FactLink dimension='metroCode'/>
<FactLink dimension='namespace'/>
<FactLink dimension='page'/>
<FactLink dimension='regionIsoCode'/>
<FactLink dimension='regionName'/>
<FactLink dimension='user'/>
<FactLink dimension='user_unique'/>
</DimensionLinks>
</MeasureGroup>
</MeasureGroups>
</Cube>
</Schema>When the query
select
[Measures].members on columns,
Order([countryName].members, [Measures].[added], DESC) on rows
from [wiki]is run, it produces the following output:
__time added count deleted delta
=============== =========================== ====================== ========= ====== ======= =========
All countryName 56,592,352,129,829,184 9,385,573 39,244 394,298 8,991,275
#null 51,113,948,353,410,816 8,761,516 35,445 346,816 8,414,700
Colombia 99,503,304,065,741 60,398 69 787 59,611
Russia 279,761,176,573,129 50,561 194 2,457 48,104
United States 761,411,157,953,310 44,433 528 5,551 38,882
Italy 369,169,251,122,223 41,073 256 1,982 39,091
France 295,623,947,063,302 39,853 205 2,572 37,281
United Kingdom 337,444,146,319,684 38,587 234 2,730 35,857
India 180,257,135,637,316 30,313 125 1,147 29,166
Germany 233,615,258,310,628 26,807 162 1,224 25,583
To build, you need Java 1.7 or 1.8, Apache Maven 3.2.1 or higher:
$ mvn installTo run, you need Druid running with the query node at
http://localhost:8082, the coordinator at http://localhost:8081,
populated with the the "wikiticker" data source.
You can run from the command line as follows:
$ mvn exec:javaor run net.hydromatic.druid.mdx.Main from any Java IDE.
Because the Mondrian model is generated on the fly, you could probably apply this example to other Druid data sources. If you want advanced features such as hierarchies, dimensions with composite keys, or calculated members, you can write the Mondrian model by hand.