Skip useless pruning predicates

Author: alamb

datafusion/core/src/physical_optimizer/pruning.rs

@@ -50,6 +50,7 @@ use datafusion_expr::expr::{BinaryExpr, Cast};
 use datafusion_expr::expr_rewriter::{ExprRewritable, ExprRewriter};
 use datafusion_expr::{binary_expr, cast, try_cast, ExprSchemable};
 use datafusion_physical_expr::create_physical_expr;
+use datafusion_physical_expr::expressions::Literal;
 use log::trace;
 
 /// Interface to pass statistics information to [`PruningPredicate`]
@@ -224,6 +225,15 @@ impl PruningPredicate {
         &self.predicate_expr
     }
 
+    /// Returns true if this pruning predicate is "always true" (aka will not prune anything)
+    pub fn allways_true(&self) -> bool {
+        self.predicate_expr
+            .as_any()
+            .downcast_ref::<Literal>()
+            .map(|l| matches!(l.value(), ScalarValue::Boolean(Some(true))))
+            .unwrap_or_default()
+    }
+
     /// Returns all need column indexes to evaluate this pruning predicate
     pub(crate) fn need_input_columns_ids(&self) -> HashSet<usize> {
         let mut set = HashSet::new();

datafusion/core/src/physical_plan/file_format/parquet.rs

@@ -97,19 +97,28 @@ impl ParquetExec {
         let predicate_creation_errors =
             MetricBuilder::new(&metrics).global_counter("num_predicate_creation_errors");
 
-        let pruning_predicate = predicate.and_then(|predicate_expr| {
-            match PruningPredicate::try_new(
-                predicate_expr,
-                base_config.file_schema.clone(),
-            ) {
-                Ok(pruning_predicate) => Some(pruning_predicate),
-                Err(e) => {
-                    debug!("Could not create pruning predicate for: {}", e);
-                    predicate_creation_errors.add(1);
+        let pruning_predicate = predicate
+            .and_then(|predicate_expr| {
+                match PruningPredicate::try_new(
+                    predicate_expr,
+                    base_config.file_schema.clone(),
+                ) {
+                    Ok(pruning_predicate) => Some(pruning_predicate),
+                    Err(e) => {
+                        debug!("Could not create pruning predicate for: {}", e);
+                        predicate_creation_errors.add(1);
+                        None
+                    }
+                }
+            })
+            .and_then(|pruning_predicate| {
+                // If the pruning predicate can't prune anything, don't try
+                if pruning_predicate.allways_true() {
                     None
+                } else {
+                    Some(pruning_predicate)
                 }
-            }
-        });
+            });
 
         let (projected_schema, projected_statistics) = base_config.project();
 
@@ -680,7 +689,7 @@ mod tests {
     use chrono::{TimeZone, Utc};
     use datafusion_common::assert_contains;
     use datafusion_common::ScalarValue;
-    use datafusion_expr::{col, lit};
+    use datafusion_expr::{col, lit, when};
     use futures::StreamExt;
     use object_store::local::LocalFileSystem;
     use object_store::path::Path;
@@ -1544,6 +1553,10 @@ mod tests {
 
         let rt = round_trip(vec![batch1], None, None, Some(filter), true, false).await;
 
+        // should have a pruning predicate
+        let pruning_predicate = &rt.parquet_exec.pruning_predicate;
+        assert!(pruning_predicate.is_some());
+
         // convert to explain plan form
         let display = displayable(rt.parquet_exec.as_ref()).indent().to_string();
 
@@ -1554,6 +1567,37 @@ mod tests {
         assert_contains!(&display, "projection=[c1]");
     }
 
+    #[tokio::test]
+    async fn parquet_exec_skip_empty_pruning() {
+        let c1: ArrayRef = Arc::new(StringArray::from(vec![
+            Some("Foo"),
+            None,
+            Some("bar"),
+            Some("bar"),
+            Some("bar"),
+            Some("bar"),
+            Some("zzz"),
+        ]));
+
+        // batch1: c1(string)
+        let batch1 = create_batch(vec![("c1", c1.clone())]);
+
+        // filter is too complicated for pruning
+        let filter = when(col("c1").not_eq(lit("bar")), lit(true))
+            .otherwise(lit(false))
+            .unwrap();
+
+        let rt = round_trip(vec![batch1], None, None, Some(filter), true, false).await;
+
+        // Should not contain a pruning predicate
+        let pruning_predicate = &rt.parquet_exec.pruning_predicate;
+        assert!(
+            pruning_predicate.is_none(),
+            "Still had pruning  predicate: {:?}",
+            pruning_predicate
+        )
+    }
+
     /// returns the sum of all the metrics with the specified name
     /// the returned set.
     ///