[FEAT] connect: add drop support

src/daft-connect/src/translation/logical_plan.rs

@@ -8,11 +8,12 @@ use spark_connect::{relation::RelType, Limit, Relation};
 use tracing::warn;
 
 use crate::translation::logical_plan::{
-    aggregate::aggregate, local_relation::local_relation, project::project, range::range,
-    to_df::to_df,
+    aggregate::aggregate, drop::drop, local_relation::local_relation, project::project,
+    range::range, to_df::to_df,
 };
 
 mod aggregate;
+mod drop;
 mod local_relation;
 mod project;
 mod range;
@@ -53,6 +54,7 @@ pub fn to_logical_plan(relation: Relation) -> eyre::Result<Plan> {
         RelType::LocalRelation(l) => {
             local_relation(l).wrap_err("Failed to apply local_relation to logical plan")
         }
+        RelType::Drop(d) => drop(*d).wrap_err("Failed to apply drop to logical plan"),
         plan => bail!("Unsupported relation type: {plan:?}"),
     }
 }

src/daft-connect/src/translation/logical_plan/drop.rs

@@ -0,0 +1,39 @@
+use eyre::bail;
+
+use crate::translation::{to_logical_plan, Plan};
+
+pub fn drop(drop: spark_connect::Drop) -> eyre::Result<Plan> {
+    let spark_connect::Drop {
+        input,
+        columns,
+        column_names,
+    } = drop;
+
+    let Some(input) = input else {
+        bail!("input is required");
+    };
+
+    if !columns.is_empty() {
+        bail!("columns is not supported; use column_names instead");
+    }
+
+    let mut plan = to_logical_plan(*input)?;
+
+    // Get all column names from the schema
+    let all_columns = plan.builder.schema().names();
+
+    // Create a set of columns to drop for efficient lookup
+    let columns_to_drop: std::collections::HashSet<_> = column_names.iter().collect();
+
+    // Create expressions for all columns except the ones being dropped
+    let to_select = all_columns
+        .iter()
+        .filter(|col_name| !columns_to_drop.contains(*col_name))
+        .map(|col_name| daft_dsl::col(col_name.clone()))
+        .collect();
+
+    // Use select to keep only the columns we want
+    plan.builder = plan.builder.select(to_select)?;
+
+    Ok(plan)
+}

tests/connect/test_drop.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+
+def test_drop(spark_session):
+    # Create DataFrame from range(10)
+    df = spark_session.range(10)
+
+    # Drop the 'id' column
+    df_dropped = df.drop("id")
+
+    # Verify the drop was successful
+    assert "id" not in df_dropped.columns, "Column 'id' should be dropped"
+    assert len(df_dropped.columns) == len(df.columns) - 1, "Should have one less column after drop"
+
+    # Verify the DataFrame has no columns after dropping all columns"
+    assert len(df_dropped.toPandas().columns) == 0, "DataFrame should have no columns after dropping 'id'"
+    assert df_dropped.count() == df.count(), "Row count should be unchanged after drop"