CLASH's SQL Dialect

We support queries of type

SELECT (list of attributes)
FROM (list of streamNames)
WHERE (list of join predicates)

Currently, there is no support for aggregation. Attribute lists as well as predicates have to be qualified, e.g.

SELECT a.foo, b.bar
FROM topic a, othertopic b
WHERE a.foo < 500
  AND a.key = b.key

This is due to the ad-hoc nature of the schema. When this query is posed, a-tuples are assumed to have a ‘foo’ and a ‘key’ attribute, b-tuples have ‘bar’ and ‘key’. If these attributes are not found, they are set to NULL. Nested queries are not allowed. Finally, windows are specified using a table function syntax:

SELECT a.foo
FROM topic a('sliding', '4', 'hours')

Connecting Relations to Stream Providers

If you query a relation in the FROM clause and assign an alias to it, like here:

SELECT *
FROM topic a

Then in the rest of the query you need to address data from that relation as a, so using the alias. The stream provider is connected by specifying it for topic, so for the base relation name.

Parsing and Interpreting

Clash uses JSqlParser for parsing plain text queries. Ultimately it produces a Query object as described here. The parser code lives in package de.unikl.dbis.clash.query.

JSql ASTNodes

You can find a little more details in the docs.