[GitHub] [arrow-datafusion] jackwener commented on a change in pull request #2054: add some pre-check for LogicalPlanBuilder(WIP)

[GitBox <gi...@apache.org>]

jackwener commented on a change in pull request #2054:
URL: https://github.com/apache/arrow-datafusion/pull/2054#discussion_r831847529



##########
File path: datafusion/src/logical_plan/builder.rs
##########
@@ -916,7 +916,11 @@ impl LogicalPlanBuilder {
 
     /// Build the plan
     pub fn build(&self) -> Result<LogicalPlan> {
-        Ok(self.plan.clone())
+        if let Some(err) = check_plan_invalid(&self.plan) {

Review comment:
       I recommend to return `Result` instead of `Option<Err>`.
   
   We can use like this.
   
   ```rust
   match err {
       Ok() => Ok(s),
       Err() => Err(e),
   }
   ```

##########
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()
+}
+
+/// do some checks for exprs in a logical plan
+fn check_invalid_expr(expr: &Expr, schema: &DFSchemaRef) -> Option<DataFusionError> {

Review comment:
       ditto




-- 
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-unsubscribe@arrow.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org