doki23 commented on a change in pull request #2054:
URL: https://github.com/apache/arrow-datafusion/pull/2054#discussion_r832080740
##########
File path: datafusion/src/logical_plan/builder.rs
##########
@@ -1090,6 +1094,101 @@ pub(crate) fn expand_wildcard(
}
}
+/// check whether the logical plan we are building is valid
+fn check_plan_invalid(plan: &LogicalPlan) -> Option<DataFusionError> {
+ match plan {
+ LogicalPlan::Projection(Projection { expr, input, .. })
+ | LogicalPlan::Sort(Sort { expr, input })
+ | LogicalPlan::Window(Window {
+ window_expr: expr,
+ input,
+ ..
+ }) => check_plan_invalid(input)
+ .or_else(|| check_any_invalid_expr(expr, input.schema())),
+
+ LogicalPlan::Filter(Filter {
+ predicate: expr,
+ input,
+ }) => {
+ check_plan_invalid(input).or_else(|| check_invalid_expr(expr,
input.schema()))
+ }
+
+ LogicalPlan::Aggregate(Aggregate {
+ input,
+ group_expr,
+ aggr_expr,
+ ..
+ }) => {
+ let schema = input.schema();
+ check_plan_invalid(input)
+ .or_else(|| check_any_invalid_expr(group_expr, schema))
+ .or_else(|| check_any_invalid_expr(aggr_expr, schema))
+ }
+
+ LogicalPlan::Join(Join { left, right, .. })
+ | LogicalPlan::CrossJoin(CrossJoin { left, right, .. }) => {
+ check_plan_invalid(left).or_else(|| check_plan_invalid(right))
+ }
+
+ LogicalPlan::Repartition(Repartition { input, .. })
+ | LogicalPlan::Limit(Limit { input, .. })
+ | LogicalPlan::Explain(Explain { plan: input, .. })
+ | LogicalPlan::Analyze(Analyze { input, .. }) =>
check_plan_invalid(input),
+
+ LogicalPlan::Union(Union { inputs, .. }) => {
+ inputs.iter().filter_map(check_plan_invalid).next()
+ }
+
+ LogicalPlan::TableScan(TableScan {
+ table_name: _,
+ source,
+ projection,
+ projected_schema,
+ filters,
+ limit: _,
+ }) => {
+ if let Some(projection) = projection {
+ if projection.len() > projected_schema.fields().len() {
+ return Some(DataFusionError::Plan(
+ "projection contains columns that doesnt belong to
projected schema"
+ .to_owned(),
+ ));
+ }
+ }
+ let schema = &source.schema().to_dfschema_ref().ok()?;
+ check_any_invalid_expr(filters, schema)
+ }
+
+ _ => None,
+ }
+}
+
+/// find first error in the exprs
+fn check_any_invalid_expr(
+ exprs: &[Expr],
+ schema: &DFSchemaRef,
+) -> Option<DataFusionError> {
+ exprs
+ .iter()
+ .filter_map(|e| check_invalid_expr(e, schema))
+ .next()
+}
Review comment:
Hm, if use Result, code here looks like:
```
for expr in exprs {
let r = check_invalid_expr(expr, schema);
if r.is_err() {
return r;
}
}
Ok(())
```
or
```
exprs
.iter()
.map(|e| check_invalid_expr(e, schema))
.fold_ok((), |start, v| start) // I dont like it
```
I dont like the `fold_ok((), |start, v| start)`
But now, it's
```
exprs
.iter()
.filter_map(|e| check_invalid_expr(e, schema))
.next()
```
I don't know how to make it more elegant, please give more suggestions.
@jackwener @alamb
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
