We're trying to pass in schema data manually to process queries for BigQuery, but that causes problems in the QuerySqlStatisticProvider because it requires a JdbcSchema in order to get the dialect and dataSource, as well as using a Jdbc Builder.
I propose refactoring QuerySqlStatisticProvider to grab the dialect and dataSource using unwrap on the schema instead, so that any schema that implements unwrap methods for these types will work with QuerySqlStatisticProvider. As well as using a generic RelBuilder, instead of a JDBC_BUILDER.
cc Julian Hyde since we have been discussing this
I've opened a PR at https://github.com/apache/calcite/pull/1467