DNM [repr types] testing feature flag for reprtypecheck

misc/python/materialize/mzcompose/__init__.py

@@ -107,6 +107,7 @@ def get_minimal_system_parameters(
         "enable_rbac_checks": "true",
         "enable_reduce_mfp_fusion": "true",
         "enable_refresh_every_mvs": "true",
+        "enable_repr_typecheck": "false",
         "enable_cluster_schedule_refresh": "true",
         "enable_sql_server_source": "true",
         "enable_statement_lifecycle_logging": "true",

src/adapter/src/optimize/copy_to.rs

@@ -28,6 +28,9 @@ use mz_storage_types::connections::Connection;
 use mz_storage_types::sinks::S3UploadInfo;
 use mz_transform::dataflow::DataflowMetainfo;
 use mz_transform::normalize_lets::normalize_lets;
+use mz_transform::reprtypecheck::{
+    SharedContext as ReprTypecheckContext, empty_context as empty_repr_context,
+};
 use mz_transform::typecheck::{SharedContext as TypecheckContext, empty_context};
 use mz_transform::{StatisticsOracle, TransformCtx};
 use timely::progress::Antichain;
@@ -48,6 +51,8 @@ use crate::optimize::{
 pub struct Optimizer {
     /// A typechecking context to use throughout the optimizer pipeline.
     typecheck_ctx: TypecheckContext,
+    /// A representation typechecking context to use throughout the optimizer pipeline.
+    repr_typecheck_ctx: ReprTypecheckContext,
     /// A snapshot of the catalog state.
     catalog: Arc<Catalog>,
     /// A snapshot of the cluster that will run the dataflows.
@@ -75,6 +80,7 @@ impl Optimizer {
     ) -> Self {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             catalog,
             compute_instance,
             select_id,
@@ -167,6 +173,7 @@ impl Optimize<HirRelationExpr> for Optimizer {
         let mut transform_ctx = TransformCtx::local(
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
             Some(self.select_id),
@@ -344,6 +351,7 @@ impl<'s> Optimize<LocalMirPlan<Resolved<'s>>> for Optimizer {
             &*stats,
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
         );

src/adapter/src/optimize/index.rs

@@ -38,6 +38,9 @@ use mz_transform::TransformCtx;
 use mz_transform::dataflow::DataflowMetainfo;
 use mz_transform::normalize_lets::normalize_lets;
 use mz_transform::notice::{IndexAlreadyExists, IndexKeyEmpty};
+use mz_transform::reprtypecheck::{
+    SharedContext as ReprTypecheckContext, empty_context as empty_repr_context,
+};
 use mz_transform::typecheck::{SharedContext as TypecheckContext, empty_context};
 
 use crate::optimize::dataflows::{
@@ -51,6 +54,8 @@ use crate::optimize::{
 pub struct Optimizer {
     /// A typechecking context to use throughout the optimizer pipeline.
     typecheck_ctx: TypecheckContext,
+    /// A representation typechecking context to use throughout the optimizer pipeline.
+    repr_typecheck_ctx: ReprTypecheckContext,
     /// A snapshot of the catalog state.
     catalog: Arc<dyn OptimizerCatalog>,
     /// A snapshot of the cluster that will run the dataflows.
@@ -75,6 +80,7 @@ impl Optimizer {
     ) -> Self {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             catalog,
             compute_instance,
             exported_index_id,
@@ -177,6 +183,7 @@ impl Optimize<Index> for Optimizer {
             &mz_transform::EmptyStatisticsOracle, // TODO: wire proper stats
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
         );

src/adapter/src/optimize/materialized_view.rs

@@ -42,6 +42,9 @@ use mz_sql::plan::HirRelationExpr;
 use mz_transform::TransformCtx;
 use mz_transform::dataflow::DataflowMetainfo;
 use mz_transform::normalize_lets::normalize_lets;
+use mz_transform::reprtypecheck::{
+    SharedContext as ReprTypecheckContext, empty_context as empty_repr_context,
+};
 use mz_transform::typecheck::{SharedContext as TypecheckContext, empty_context};
 use timely::progress::Antichain;
 
@@ -56,6 +59,8 @@ use crate::optimize::{
 pub struct Optimizer {
     /// A typechecking context to use throughout the optimizer pipeline.
     typecheck_ctx: TypecheckContext,
+    /// A representation typechecking context to use throughout the optimizer pipeline.
+    repr_typecheck_ctx: ReprTypecheckContext,
     /// A snapshot of the catalog state.
     catalog: Arc<dyn OptimizerCatalog>,
     /// A snapshot of the cluster that will run the dataflows.
@@ -115,6 +120,7 @@ impl Optimizer {
     ) -> Self {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             catalog,
             compute_instance,
             sink_id,
@@ -197,6 +203,7 @@ impl Optimize<HirRelationExpr> for Optimizer {
         let mut transform_ctx = TransformCtx::local(
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
             Some(self.view_id),
@@ -286,6 +293,7 @@ impl Optimize<LocalMirPlan> for Optimizer {
             &mz_transform::EmptyStatisticsOracle, // TODO: wire proper stats
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
         );

src/adapter/src/optimize/peek.rs

@@ -25,6 +25,9 @@ use mz_sql::plan::HirRelationExpr;
 use mz_sql::session::metadata::SessionMetadata;
 use mz_transform::dataflow::DataflowMetainfo;
 use mz_transform::normalize_lets::normalize_lets;
+use mz_transform::reprtypecheck::{
+    SharedContext as ReprTypecheckContext, empty_context as empty_repr_context,
+};
 use mz_transform::typecheck::{SharedContext as TypecheckContext, empty_context};
 use mz_transform::{StatisticsOracle, TransformCtx};
 use timely::progress::Antichain;
@@ -45,6 +48,8 @@ use crate::optimize::{
 pub struct Optimizer {
     /// A typechecking context to use throughout the optimizer pipeline.
     typecheck_ctx: TypecheckContext,
+    /// A representation typechecking context to use throughout the optimizer pipeline.
+    repr_typecheck_ctx: ReprTypecheckContext,
     /// A snapshot of the catalog state.
     catalog: Arc<Catalog>,
     /// A snapshot of the cluster that will run the dataflows.
@@ -75,6 +80,7 @@ impl Optimizer {
     ) -> Self {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             catalog,
             compute_instance,
             finishing,
@@ -181,6 +187,7 @@ impl Optimize<HirRelationExpr> for Optimizer {
         let mut transform_ctx = TransformCtx::local(
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
             Some(self.select_id),
@@ -336,6 +343,7 @@ impl<'s> Optimize<LocalMirPlan<Resolved<'s>>> for Optimizer {
             &*stats,
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
         );

src/adapter/src/optimize/subscribe.rs

@@ -26,6 +26,9 @@ use mz_sql::plan::SubscribeFrom;
 use mz_transform::TransformCtx;
 use mz_transform::dataflow::DataflowMetainfo;
 use mz_transform::normalize_lets::normalize_lets;
+use mz_transform::reprtypecheck::{
+    SharedContext as ReprTypecheckContext, empty_context as empty_repr_context,
+};
 use mz_transform::typecheck::{SharedContext as TypecheckContext, empty_context};
 use timely::progress::Antichain;
 
@@ -42,6 +45,8 @@ use crate::optimize::{
 pub struct Optimizer {
     /// A typechecking context to use throughout the optimizer pipeline.
     typecheck_ctx: TypecheckContext,
+    /// A representation typechecking context to use throughout the optimizer pipeline.
+    repr_typecheck_ctx: ReprTypecheckContext,
     /// A snapshot of the catalog state.
     catalog: Arc<dyn OptimizerCatalog>,
     /// A snapshot of the cluster that will run the dataflows.
@@ -95,6 +100,7 @@ impl Optimizer {
     ) -> Self {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             catalog,
             compute_instance,
             view_id,
@@ -229,6 +235,7 @@ impl Optimize<SubscribeFrom> for Optimizer {
                 let mut transform_ctx = TransformCtx::local(
                     &self.config.features,
                     &self.typecheck_ctx,
+                    &self.repr_typecheck_ctx,
                     &mut df_meta,
                     Some(&self.metrics),
                     Some(self.view_id),
@@ -272,6 +279,7 @@ impl Optimize<SubscribeFrom> for Optimizer {
             &mz_transform::EmptyStatisticsOracle, // TODO: wire proper stats
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             Some(&self.metrics),
         );

src/adapter/src/optimize/view.rs

@@ -26,6 +26,9 @@ use mz_sql::optimizer_metrics::OptimizerMetrics;
 use mz_sql::plan::HirRelationExpr;
 use mz_transform::TransformCtx;
 use mz_transform::dataflow::DataflowMetainfo;
+use mz_transform::reprtypecheck::{
+    SharedContext as ReprTypecheckContext, empty_context as empty_repr_context,
+};
 use mz_transform::typecheck::{SharedContext as TypecheckContext, empty_context};
 
 use crate::optimize::dataflows::{ExprPrepStyle, prep_relation_expr};
@@ -37,6 +40,8 @@ use crate::optimize::{
 pub struct Optimizer<'a> {
     /// A typechecking context to use throughout the optimizer pipeline.
     typecheck_ctx: TypecheckContext,
+    /// A representation typechecking context to use throughout the optimizer pipeline.
+    repr_typecheck_ctx: ReprTypecheckContext,
     /// Optimizer config.
     config: OptimizerConfig,
     /// Optimizer metrics.
@@ -54,6 +59,7 @@ impl<'a> Optimizer<'a> {
     pub fn new(config: OptimizerConfig, metrics: Option<OptimizerMetrics>) -> Self {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             config,
             metrics,
             expr_prep_style: None,
@@ -71,6 +77,7 @@ impl<'a> Optimizer<'a> {
     ) -> Optimizer<'a> {
         Self {
             typecheck_ctx: empty_context(),
+            repr_typecheck_ctx: empty_repr_context(),
             config,
             metrics,
             expr_prep_style: Some(expr_prep_style),
@@ -95,6 +102,7 @@ impl Optimize<HirRelationExpr> for Optimizer<'_> {
         let mut transform_ctx = TransformCtx::local(
             &self.config.features,
             &self.typecheck_ctx,
+            &self.repr_typecheck_ctx,
             &mut df_meta,
             self.metrics.as_ref(),
             None,

src/expr/src/scalar/func/variadic.rs

@@ -29,7 +29,7 @@ use mz_repr::adt::range::{InvalidRangeError, Range, RangeBound, parse_range_boun
 use mz_repr::adt::system::Oid;
 use mz_repr::adt::timestamp::CheckedTimestamp;
 use mz_repr::role_id::RoleId;
-use mz_repr::{ColumnName, Datum, Row, RowArena, SqlColumnType, SqlScalarType};
+use mz_repr::{ColumnName, Datum, ReprScalarType, Row, RowArena, SqlColumnType, SqlScalarType};
 use serde::{Deserialize, Serialize};
 use sha1::Sha1;
 use sha2::{Sha224, Sha256, Sha384, Sha512};
@@ -1321,8 +1321,11 @@ impl VariadicFunc {
             .nullable(true),
             ArrayCreate { elem_type } => {
                 debug_assert!(
-                    input_types.iter().all(|t| t.scalar_type.base_eq(elem_type)),
-                    "Args to ArrayCreate should have types that are compatible with the elem_type"
+                    input_types
+                        .iter()
+                        .all(|t| ReprScalarType::from(&t.scalar_type)
+                            == ReprScalarType::from(elem_type)),
+                    "Args to ArrayCreate should have types that are repr-compatible with the elem_type"
                 );
                 match elem_type {
                     SqlScalarType::Array(_) => elem_type.clone().nullable(false),

src/repr/src/explain.rs

@@ -45,7 +45,7 @@ use crate::explain::dot::{DisplayDot, dot_string};
 use crate::explain::json::{DisplayJson, json_string};
 use crate::explain::text::{DisplayText, text_string};
 use crate::optimize::OptimizerFeatureOverrides;
-use crate::{GlobalId, SqlColumnType, SqlScalarType};
+use crate::{GlobalId, ReprColumnType, ReprScalarType, SqlColumnType, SqlScalarType};
 
 pub mod dot;
 pub mod json;
@@ -444,6 +444,12 @@ pub trait ExprHumanizer: fmt::Debug {
     /// compatibility is more important.
     fn humanize_scalar_type(&self, ty: &SqlScalarType, postgres_compat: bool) -> String;
 
+    /// Returns a human-readable name for the specified scalar type.
+    /// Calls `humanize_scalar_type` with the `SqlScalarType` representation of the specified type.
+    fn humanize_scalar_type_repr(&self, typ: &ReprScalarType, postgres_compat: bool) -> String {
+        self.humanize_scalar_type(&SqlScalarType::from_repr(typ), postgres_compat)
+    }
+
     /// Returns a human-readable name for the specified column type.
     /// Used in, e.g., EXPLAIN and error msgs, in which case exact Postgres compatibility is less
     /// important than showing as much detail as possible. Also used in `pg_typeof`, where Postgres
@@ -456,6 +462,12 @@ pub trait ExprHumanizer: fmt::Debug {
         )
     }
 
+    /// Returns a human-readable name for the specified column type.
+    /// Calls `humanize_column_type` with the `SqlColumnType` representation of the specified type.
+    fn humanize_column_type_repr(&self, typ: &ReprColumnType, postgres_compat: bool) -> String {
+        self.humanize_column_type(&SqlColumnType::from_repr(typ), postgres_compat)
+    }
+
     /// Returns a vector of column names for the relation identified by `id`.
     fn column_names_for_id(&self, id: GlobalId) -> Option<Vec<String>>;
 

src/repr/src/lib.rs

@@ -54,8 +54,8 @@ pub use crate::global_id::GlobalId;
 pub use crate::relation::{
     ColumnIndex, ColumnName, NotNullViolation, PropRelationDescDiff, ProtoColumnName,
     ProtoColumnType, ProtoRelationDesc, ProtoRelationType, RelationDesc, RelationDescBuilder,
-    RelationVersion, RelationVersionSelector, ReprColumnType, SqlColumnType, SqlRelationType,
-    UNKNOWN_COLUMN_NAME, VersionedRelationDesc, arb_relation_desc_diff,
+    RelationVersion, RelationVersionSelector, ReprColumnType, ReprRelationType, SqlColumnType,
+    SqlRelationType, UNKNOWN_COLUMN_NAME, VersionedRelationDesc, arb_relation_desc_diff,
     arb_relation_desc_projection, arb_row_for_relation,
 };
 pub use crate::row::encode::{RowColumnarDecoder, RowColumnarEncoder, preserves_order};
@@ -66,7 +66,7 @@ pub use crate::row::{
 };
 pub use crate::scalar::{
     ArrayRustType, AsColumnType, Datum, DatumType, PropArray, PropDatum, PropDict, PropList,
-    ProtoScalarType, ReprScalarType, ScalarBaseType, SqlScalarType, arb_datum,
-    arb_datum_for_column, arb_datum_for_scalar, arb_range_type,
+    ProtoScalarType, ReprScalarBaseType, ReprScalarType, SqlScalarBaseType, SqlScalarType,
+    arb_datum, arb_datum_for_column, arb_datum_for_scalar, arb_range_type,
 };
 pub use crate::timestamp::{Timestamp, TimestampManipulation};

src/repr/src/optimize.rs

@@ -131,6 +131,7 @@ optimizer_feature_flags!({
     // See the feature flag of the same name.
     enable_dequadratic_eqprop_map: bool,
     enable_fast_path_plan_insights: bool,
+    enable_repr_typecheck: bool,
 });
 
 /// A trait used to implement layered config construction.