alamb commented on issue #533:
URL:
https://github.com/apache/arrow-datafusion/issues/533#issuecomment-1221293210
An update here:
There are several examples of people who have implemented their own custom
query language / domain specific language using DataFusion. For example vega
fusion with the vega language, and InfluxDB IOx with the language that
underlies flux and influxql.
Typically the approach is to use the `LogialPlanBuilder` and skip the
datafusion SQLPLanner entirely
One way we might be able to accomplish this extension point would be like:
```rust
/// Default SQL planner
pub trait SQLPlanner :
fn plan_query(&mut self, query: sqlparser::ast::Query) -> LogicalPlan {
// call free function
plan_query(self, query)
}
fn plan_expr(&mut self, expr: sqlparser::ast::Expr) -> Expr {
// existing logic work
plan_expr(self, expr)
}
}
// Free functions that do the work (as you can't call default trait impls)
//
https://stackoverflow.com/questions/43849041/reuse-default-method-implementations-in-trait-impls
fn plan_query<P: SQLPlanner>(planner: &mut P, query: sqlparser::ast::Query)
-> LogicalPlan {
// existing logic
}
fn plan_expr<P: SQLPlanner>(planner: &mut P, expr: sqlparser::ast::Expr) ->
Expr {
// existing logic work
}
```
I was able to make this type of pattern work for "sql rewriting" where the
user is allowed to rewrite an SQL query prior to planning with the existing
datafusion planner. You can see a sketch of such a solution in
https://github.com/influxdata/influxdb_iox/pull/5438.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscr...@arrow.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
