(datafusion) 01/02: fix: custom_datasource example ignores projection pushdown in execute() (#22417)

[gi...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

github-merge-queue[bot] pushed a commit to branch gh-readonly-queue/main/pr-21976-b62b0cb4c4eff993d3cce9c6465ff49cb35d98a5
in repository https://gitbox.apache.org/repos/asf/datafusion.git

commit b62b0cb4c4eff993d3cce9c6465ff49cb35d98a5
Author: Kumar Ujjawal <uj...@gmail.com>
AuthorDate: Sat May 23 17:48:56 2026 +0530

    fix: custom_datasource example ignores projection pushdown in execute() (#22417)
    
    ## Which issue does this PR close?
    
    <!--
    We generally require a GitHub issue to be filed for all bug fixes and
    enhancements and this helps us generate change logs for our releases.
    You can link an issue to this PR using the GitHub syntax. For example
    `Closes #123` indicates that this PR will close issue #123.
    -->
    
    - Closes #18578.
    
    ## Rationale for this change
    
    The `custom_data_source` example computes a `projected_schema` from the
    projection it receives in `scan`, advertising projection-pushdown
    support but its `CustomExec::execute` then always emits both source
    columns. As soon as
    the planner pushes a non-identity projection (e.g. `SELECT 1 FROM t` →
    `Some([])`, `SELECT COUNT(id) FROM t` → `Some([0])`), the RecordBatch's
    column count diverges from `projected_schema` and the query fails.
    
    
    
    <!--
    Why are you proposing this change? If this is already explained clearly
    in the issue then this section is not needed.
    Explaining clearly why changes are proposed helps reviewers understand
    your changes and offer better suggestions for fixes.
    -->
    
    ## What changes are included in this PR?
    
    - Store the projection on `CustomExec` so `execute` can apply it.
    - In `execute`, build a `RecordBatch` over the full source schema and
    use `RecordBatch::project` to drop the columns the planner didn't ask
    for.`project` preserves row count, which matters when the projection
    selects zero columns (`SELECT 1 FROM t`).
    - Extend `custom_datasource()` to register the table with a
    `SessionContext` and run `SELECT 1 AS a FROM accounts` and `SELECT
    COUNT(id) FROM accounts`, which are the two shapes reported in the
    issue.
    
    <!--
    There is no need to duplicate the description in the issue here but it
    is sometimes worth providing a summary of the individual changes in this
    PR.
    -->
    
    ## Are these changes tested?
    
    Yes
    
    <!--
    We typically require tests for all PRs in order to:
    1. Prevent the code from being accidentally broken by subsequent changes
    2. Serve as another way to document the expected behavior of the code
    
    If tests are not included in your PR, please explain why (for example,
    are they covered by existing tests)?
    -->
    
    ## Are there any user-facing changes?
    
    No
    
    <!--
    If there are user-facing changes then we may require documentation to be
    updated before approving the PR.
    -->
    
    <!--
    If there are any breaking changes to public APIs, please add the `api
    change` label.
    -->
---
 .../custom_data_source/custom_datasource.rs        | 56 ++++++++++++++++++----
 1 file changed, 48 insertions(+), 8 deletions(-)

diff --git a/datafusion-examples/examples/custom_data_source/custom_datasource.rs b/datafusion-examples/examples/custom_data_source/custom_datasource.rs
index 937452a286..a67738520b 100644
--- a/datafusion-examples/examples/custom_data_source/custom_datasource.rs
+++ b/datafusion-examples/examples/custom_data_source/custom_datasource.rs
@@ -26,6 +26,7 @@ use async_trait::async_trait;
 use datafusion::arrow::array::{UInt8Builder, UInt64Builder};
 use datafusion::arrow::datatypes::{DataType, Field, Schema, SchemaRef};
 use datafusion::arrow::record_batch::RecordBatch;
+use datafusion::common::assert_batches_eq;
 use datafusion::datasource::{TableProvider, TableType, provider_as_source};
 use datafusion::error::Result;
 use datafusion::execution::context::TaskContext;
@@ -52,6 +53,33 @@ pub async fn custom_datasource() -> Result<()> {
     search_accounts(db.clone(), Some(col("bank_account").gt(lit(8000u64))), 1).await?;
     search_accounts(db.clone(), Some(col("bank_account").gt(lit(200u64))), 2).await?;
 
+    // exercise SQL paths that push down non-trivial projections:
+    // - `SELECT 1 ...` requests no source columns (projection: Some([]))
+    // - `SELECT COUNT(id) ...` requests a single column (projection: Some([0]))
+    let ctx = SessionContext::new();
+    ctx.register_table("accounts", Arc::new(db))?;
+    let constant_batches = ctx
+        .sql("SELECT 1 AS a FROM accounts")
+        .await?
+        .collect()
+        .await?;
+    assert_batches_eq!(
+        [
+            "+---+", "| a |", "+---+", "| 1 |", "| 1 |", "| 1 |", "+---+",
+        ],
+        &constant_batches
+    );
+
+    let count_batches = ctx
+        .sql("SELECT COUNT(id) AS cnt FROM accounts")
+        .await?
+        .collect()
+        .await?;
+    assert_batches_eq!(
+        ["+-----+", "| cnt |", "+-----+", "| 3   |", "+-----+",],
+        &count_batches
+    );
+
     Ok(())
 }
 
@@ -186,6 +214,7 @@ impl TableProvider for CustomDataSource {
 #[derive(Debug, Clone)]
 struct CustomExec {
     db: CustomDataSource,
+    projection: Option<Vec<usize>>,
     projected_schema: SchemaRef,
     cache: Arc<PlanProperties>,
 }
@@ -201,6 +230,7 @@ impl CustomExec {
         let cache = Self::compute_properties(projected_schema.clone());
         Self {
             db,
+            projection: projections.cloned(),
             projected_schema,
             cache: Arc::new(cache),
         }
@@ -262,15 +292,25 @@ impl ExecutionPlan for CustomExec {
             account_array.append_value(user.bank_account);
         }
 
+        // Build a batch holding every column the table can produce, then let
+        // Arrow drop the columns the query didn't ask for. `RecordBatch::project`
+        // preserves the row count, which matters when the projection selects
+        // zero columns (e.g. `SELECT 1 FROM t`).
+        let full_batch = RecordBatch::try_new(
+            self.db.schema(),
+            vec![
+                Arc::new(id_array.finish()),
+                Arc::new(account_array.finish()),
+            ],
+        )?;
+        let batch = match &self.projection {
+            Some(indices) => full_batch.project(indices)?,
+            None => full_batch,
+        };
+
         Ok(Box::pin(MemoryStream::try_new(
-            vec![RecordBatch::try_new(
-                self.projected_schema.clone(),
-                vec![
-                    Arc::new(id_array.finish()),
-                    Arc::new(account_array.finish()),
-                ],
-            )?],
-            self.schema(),
+            vec![batch],
+            self.projected_schema.clone(),
             None,
         )?))
     }


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datafusion.apache.org
For additional commands, e-mail: commits-help@datafusion.apache.org