Remove varchar_to_text from MIR

src/expr/src/relation.rs

@@ -1707,7 +1707,7 @@ impl MirRelationExpr {
                 .column_types
                 .iter()
                 .zip_eq(typ.column_types.iter())
-                .all(|(t1, t2)| t1.scalar_type.base_eq(&t2.scalar_type)));
+                .all(|(t1, t2)| t1.scalar_type.physical_eq(&t2.scalar_type)));
         }
         let mut typ = typ.unwrap_or_else(|| self.typ());
         typ.keys = vec![vec![]];

src/expr/src/scalar/func.rs

@@ -33,8 +33,8 @@ use mz_ore::fmt::FormatBuffer;
 use mz_ore::lex::LexBuf;
 use mz_ore::option::OptionExt;
 use mz_ore::result::ResultExt;
-use mz_ore::soft_assert_eq_or_log;
 use mz_ore::str::StrExt;
+use mz_ore::{soft_assert_eq_or_log, soft_assert_or_log};
 use mz_pgrepr::Type;
 use mz_pgtz::timezone::{Timezone, TimezoneSpec};
 use mz_proto::chrono::any_naive_datetime;
@@ -8033,8 +8033,10 @@ impl VariadicFunc {
             }
             .nullable(true),
             ArrayCreate { elem_type } => {
-                debug_assert!(
-                    input_types.iter().all(|t| t.scalar_type.base_eq(elem_type)),
+                soft_assert_or_log!(
+                    input_types
+                        .iter()
+                        .all(|t| t.scalar_type.physical_eq(elem_type)),
                     "Args to ArrayCreate should have types that are compatible with the elem_type"
                 );
                 match elem_type {

src/expr/src/scalar/func/impls/string.rs

@@ -579,7 +579,14 @@ impl<'a> EagerUnaryFunc<'a> for CastStringToChar {
 
 impl fmt::Display for CastStringToChar {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        f.write_str("text_to_char")
+        match self.length {
+            Some(length) => {
+                write!(f, "text_to_char[len={}, fail_on_len={}]", length.into_u32(), self.fail_on_len)
+            },
+            None => {
+                f.write_str("text_to_char[len=None]")
+            },
+        }
     }
 }
 
@@ -707,7 +714,14 @@ impl<'a> EagerUnaryFunc<'a> for CastStringToVarChar {
 
 impl fmt::Display for CastStringToVarChar {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        f.write_str("text_to_varchar")
+        match self.length {
+            Some(length) => {
+                write!(f, "text_to_varchar[len={}, fail_on_len={}]", length.into_u32(), self.fail_on_len)
+            },
+            None => {
+                f.write_str("text_to_varchar[len=None]")
+            },
+        }
     }
 }
 

src/repr/src/relation.rs

@@ -69,7 +69,7 @@ impl ColumnType {
                     nullable: self.nullable || other.nullable,
                 })
             }
-            (scalar_type, other_scalar_type) if scalar_type.base_eq(&other_scalar_type) => {
+            (scalar_type, other_scalar_type) if scalar_type.physical_eq(&other_scalar_type) => {
                 Ok(ColumnType {
                     scalar_type: scalar_type.without_modifiers(),
                     nullable: self.nullable || other.nullable,

src/repr/src/scalar.rs

@@ -2908,12 +2908,25 @@ impl ScalarType {
         self.eq_inner(other, false)
     }
 
-    // Determines equality among scalar types that ignores any custom OIDs or
-    // embedded values.
+    /// Determines equality among scalar types that ignores any custom OIDs or
+    /// embedded values.
     pub fn structural_eq(&self, other: &ScalarType) -> bool {
         self.eq_inner(other, true)
     }
 
+    /// //////////// todo: comment
+    pub fn physical_eq(&self, other: &ScalarType) -> bool {
+        use ScalarType::*;
+        if self.base_eq(other) {
+            return true;
+        }
+        match (self, other) {
+            (String, VarChar { max_length: None }) => true,
+            (VarChar { max_length: None }, String) => true,
+            _ => false,
+        }
+    }
+
     pub fn eq_inner(&self, other: &ScalarType, structure_only: bool) -> bool {
         use ScalarType::*;
         match (self, other) {

src/sql/src/plan/lowering.rs

@@ -41,7 +41,7 @@ use std::iter::repeat;
 
 use itertools::Itertools;
 use mz_expr::visit::Visit;
-use mz_expr::{AccessStrategy, AggregateFunc, MirRelationExpr, MirScalarExpr};
+use mz_expr::{AccessStrategy, AggregateFunc, MirRelationExpr, MirScalarExpr, UnaryFunc};
 use mz_ore::collections::CollectionExt;
 use mz_ore::stack::maybe_grow;
 use mz_repr::*;
@@ -955,17 +955,23 @@ impl HirScalarExpr {
                 Literal(row, typ) => SS::Literal(Ok(row), typ),
                 Parameter(_) => panic!("cannot decorrelate expression with unbound parameters"),
                 CallUnmaterializable(func) => SS::CallUnmaterializable(func),
-                CallUnary { func, expr } => SS::CallUnary {
-                    func,
-                    expr: Box::new(expr.applied_to(
-                        id_gen,
-                        col_map,
-                        cte_map,
-                        inner,
-                        subquery_map,
-                        context,
-                    )?),
-                },
+                CallUnary { func, expr } => {
+                    if !matches!(func, UnaryFunc::CastVarCharToString(..)) {
+                        SS::CallUnary {
+                            func,
+                            expr: Box::new(expr.applied_to(
+                                id_gen,
+                                col_map,
+                                cte_map,
+                                inner,
+                                subquery_map,
+                                context,
+                            )?),
+                        }
+                    } else {
+                        expr.applied_to(id_gen, col_map, cte_map, inner, subquery_map, context)?
+                    }
+                }
                 CallBinary { func, expr1, expr2 } => SS::CallBinary {
                     func,
                     expr1: Box::new(expr1.applied_to(
@@ -1621,10 +1627,16 @@ impl HirScalarExpr {
             Column(ColumnRef { level: 0, column }) => SS::Column(column),
             Literal(datum, typ) => SS::Literal(Ok(datum), typ),
             CallUnmaterializable(func) => SS::CallUnmaterializable(func),
-            CallUnary { func, expr } => SS::CallUnary {
-                func,
-                expr: Box::new(expr.lower_uncorrelated()?),
-            },
+            CallUnary { func, expr } => {
+                if !matches!(func, UnaryFunc::CastVarCharToString(..)) {
+                    SS::CallUnary {
+                        func,
+                        expr: Box::new(expr.lower_uncorrelated()?),
+                    }
+                } else {
+                    expr.lower_uncorrelated()?
+                }
+            }
             CallBinary { func, expr1, expr2 } => SS::CallBinary {
                 func,
                 expr1: Box::new(expr1.lower_uncorrelated()?),

src/transform/src/column_knowledge.rs

@@ -635,7 +635,7 @@ impl DatumKnowledge {
                 unreachable!();
             };
 
-            if !s_typ.base_eq(o_typ) {
+            if !s_typ.physical_eq(o_typ) {
                 ::tracing::error!("Undefined join of non-equal base types {s_typ:?} != {o_typ:?}");
                 *self = Self::top();
             } else if s_val != o_val {
@@ -735,7 +735,7 @@ impl DatumKnowledge {
                 unreachable!();
             };
 
-            if !s_typ.base_eq(o_typ) {
+            if !s_typ.physical_eq(o_typ) {
                 soft_panic_or_log!("Undefined meet of non-equal base types {s_typ:?} != {o_typ:?}");
                 *self = Self::top(); // this really should be Nothing
             } else if s_val != o_val {

src/transform/src/normalize_lets.rs

@@ -318,7 +318,7 @@ mod support {
                         .column_types
                         .iter()
                         .zip(typ.column_types.iter())
-                        .all(|(t1, t2)| t1.scalar_type.base_eq(&t2.scalar_type))
+                        .all(|(t1, t2)| t1.scalar_type.physical_eq(&t2.scalar_type))
                     {
                         Err(crate::TransformError::Internal(format!(
                             "scalar types do not match: {:?} v {:?}",

src/transform/src/typecheck.rs

@@ -397,8 +397,7 @@ pub fn scalar_subtype_difference(sub: &ScalarType, sup: &ScalarType) -> Vec<Colu
             }
         }
         (_, _) => {
-            // TODO(mgree) confirm that we don't want to allow numeric subtyping
-            if ScalarBaseType::from(sub) != ScalarBaseType::from(sup) {
+            if !sub.physical_eq(sup) {
                 diffs.push(ColumnTypeDifference::NotSubtype {
                     sub: sub.clone(),
                     sup: sup.clone(),
@@ -410,20 +409,6 @@ pub fn scalar_subtype_difference(sub: &ScalarType, sup: &ScalarType) -> Vec<Colu
     diffs
 }
 
-/// Returns true when it is safe to treat a `sub` row as an `sup` row
-///
-/// In particular, the core types must be equal, and if a column in `sup` is nullable, that column should also be nullable in `sub`
-/// Conversely, it is okay to treat a known non-nullable column as nullable: `sub` may be nullable when `sup` is not
-pub fn is_subtype_of(sub: &[ColumnType], sup: &[ColumnType]) -> bool {
-    if sub.len() != sup.len() {
-        return false;
-    }
-
-    sub.iter().zip_eq(sup.iter()).all(|(got, known)| {
-        (!known.nullable || got.nullable) && got.scalar_type.base_eq(&known.scalar_type)
-    })
-}
-
 /// Check that the visible type of each query has not been changed
 #[derive(Debug)]
 pub struct Typecheck {
@@ -459,7 +444,7 @@ impl Typecheck {
 
     /// New non-transient global IDs will be treated as an error
     ///
-    /// Only turn this on after the context has been appropraitely populated by, e.g., an earlier run
+    /// Only turn this on after the context has been appropriately populated by, e.g., an earlier run
     pub fn disallow_new_globals(mut self) -> Self {
         self.disallow_new_globals = true;
         self
@@ -1362,7 +1347,7 @@ impl ColumnTypeDifference {
                 let sub = h.humanize_scalar_type(sub);
                 let sup = h.humanize_scalar_type(sup);
 
-                writeln!(f, "{sub} is a not a subtype of {sup}")
+                writeln!(f, "{sub} is not a subtype of {sup}")
             }
             Nullability { sub, sup } => {
                 let sub = h.humanize_column_type(sub);

src/transform/tests/test_transforms/typecheck.spec

@@ -225,7 +225,7 @@ Return
 mismatched column types: couldn't compute union of column types in let rec: Can't union types: Bool and Int64
       got Int64
 expected Bool?
-  Bool is a not a subtype of Int64
+  Bool is not a subtype of Int64
   Bool? is nullable but Int64 is not
 ----
 ----
@@ -344,7 +344,7 @@ Filter #2
 mismatched column types: expected boolean condition
       got String
 expected Bool?
-  String is a not a subtype of Bool
+  String is not a subtype of Bool
 ----
 ----
 

test/sqllogictest/freshmart.slt

@@ -291,7 +291,7 @@ Explained Query:
           Get l6
   Return
     Project (#0, #13, #11)
-      Map (case when (#8) IS NULL then null else #7 end, (((((#2 * case when (#3 <= 3) then 1.2 else case when ((#3 <= 10) AND (#3 >= 4)) then 1.1 else 0.9 end end) * coalesce((1 - (#5 / 100)), 1)) * case when (#12 <= 3) then 1.1 else case when ((#12 <= 10) AND (#12 >= 4)) then 1.05 else 1 end end) * case when (#2 > #4) then (1 + ((#2 - #4) / #4)) else (1 - ((#4 - #2) / #4)) end) * case when ilike["%cheap%"](varchar_to_text(#1)) then 0.8 else 1 end))
+      Map (case when (#8) IS NULL then null else #7 end, (((((#2 * case when (#3 <= 3) then 1.2 else case when ((#3 <= 10) AND (#3 >= 4)) then 1.1 else 0.9 end end) * coalesce((1 - (#5 / 100)), 1)) * case when (#12 <= 3) then 1.1 else case when ((#12 <= 10) AND (#12 >= 4)) then 1.05 else 1 end end) * case when (#2 > #4) then (1 + ((#2 - #4) / #4)) else (1 - ((#4 - #2) / #4)) end) * case when ilike["%cheap%"](#1) then 0.8 else 1 end))
         Join on=(#0 = #6 = #9 = #10) type=delta
           ArrangeBy keys=[[#0]]
             Get l6
@@ -639,7 +639,7 @@ Explained Query:
           Get l7
     cte l11 =
       Project (#0, #1, #3, #5, #18)
-        Map (case when (#14) IS NULL then null else #13 end, (((((#8 * case when (#9 <= 3) then 1.2 else case when ((#9 <= 10) AND (#9 >= 4)) then 1.1 else 0.9 end end) * coalesce((1 - (#11 / 100)), 1)) * case when (#17 <= 3) then 1.1 else case when ((#17 <= 10) AND (#17 >= 4)) then 1.05 else 1 end end) * case when (#8 > #10) then (1 + ((#8 - #10) / #10)) else (1 - ((#10 - #8) / #10)) end) * case when ilike["%cheap%"](varchar_to_text(#7)) then 0.8 else 1 end))
+        Map (case when (#14) IS NULL then null else #13 end, (((((#8 * case when (#9 <= 3) then 1.2 else case when ((#9 <= 10) AND (#9 >= 4)) then 1.1 else 0.9 end end) * coalesce((1 - (#11 / 100)), 1)) * case when (#17 <= 3) then 1.1 else case when ((#17 <= 10) AND (#17 >= 4)) then 1.05 else 1 end end) * case when (#8 > #10) then (1 + ((#8 - #10) / #10)) else (1 - ((#10 - #8) / #10)) end) * case when ilike["%cheap%"](#7) then 0.8 else 1 end))
           Join on=(#0 = #2 = #6 = #12 = #15 = #16 AND #3 = #4) type=delta
             ArrangeBy keys=[[#0]]
               Project (#0, #1)
@@ -895,7 +895,7 @@ Explained Query:
           Get l7
     cte l11 =
       Project (#0, #1, #3, #5, #18)
-        Map (case when (#14) IS NULL then null else #13 end, (((((#8 * case when (#9 <= 3) then 1.2 else case when ((#9 <= 10) AND (#9 >= 4)) then 1.1 else 0.9 end end) * coalesce((1 - (#11 / 100)), 1)) * case when (#17 <= 3) then 1.1 else case when ((#17 <= 10) AND (#17 >= 4)) then 1.05 else 1 end end) * case when (#8 > #10) then (1 + ((#8 - #10) / #10)) else (1 - ((#10 - #8) / #10)) end) * case when ilike["%cheap%"](varchar_to_text(#7)) then 0.8 else 1 end))
+        Map (case when (#14) IS NULL then null else #13 end, (((((#8 * case when (#9 <= 3) then 1.2 else case when ((#9 <= 10) AND (#9 >= 4)) then 1.1 else 0.9 end end) * coalesce((1 - (#11 / 100)), 1)) * case when (#17 <= 3) then 1.1 else case when ((#17 <= 10) AND (#17 >= 4)) then 1.05 else 1 end end) * case when (#8 > #10) then (1 + ((#8 - #10) / #10)) else (1 - ((#10 - #8) / #10)) end) * case when ilike["%cheap%"](#7) then 0.8 else 1 end))
           Join on=(#0 = #2 = #6 = #12 = #15 = #16 AND #3 = #4) type=delta
             ArrangeBy keys=[[#0]]
               Project (#0, #1)

test/sqllogictest/github-5536.slt

@@ -44,14 +44,14 @@ Explained Query:
               Filter (#0 <= #0) AND (#0 >= #0)
                 ReadStorage materialize.public.t5
           ArrangeBy keys=[[]]
-            Filter (#1 = text_to_char(text_to_varchar(boolean_to_text(like["0.31161855206970124"](padchar(#1))))))
+            Filter (#1 = text_to_char[len=None](text_to_varchar[len=None](boolean_to_text(like["0.31161855206970124"](padchar(#1))))))
               ReadStorage materialize.public.t3
           ArrangeBy keys=[[]]
             ReadStorage materialize.public.t5
 
 Source materialize.public.t0
 Source materialize.public.t3
-  filter=((#1 = text_to_char(text_to_varchar(boolean_to_text(like["0.31161855206970124"](padchar(#1)))))))
+  filter=((#1 = text_to_char[len=None](text_to_varchar[len=None](boolean_to_text(like["0.31161855206970124"](padchar(#1)))))))
 Source materialize.public.t5
 
 Target cluster: quickstart

test/sqllogictest/ldbc_bi.slt

@@ -2017,7 +2017,7 @@ Explained Query:
             ArrangeBy keys=[[#0{rootpostlanguage}]] // { arity: 1 }
               Distinct project=[#0{rootpostlanguage}] // { arity: 1 }
                 Project (#0{rootpostlanguage}) // { arity: 1 }
-                  Filter (#0{rootpostlanguage} = varchar_to_text(#1)) // { arity: 2 }
+                  Filter (#0{rootpostlanguage} = #1) // { arity: 2 }
                     FlatMap unnest_array({"es", "ta", "pt"}) // { arity: 2 }
                       Distinct project=[#0{rootpostlanguage}] // { arity: 1 }
                         Project (#13{rootpostlanguage}) // { arity: 1 }

test/sqllogictest/ldbc_bi_eager.slt

@@ -2024,7 +2024,7 @@ Explained Query:
             ArrangeBy keys=[[#0{rootpostlanguage}]] // { arity: 1 }
               Distinct project=[#0{rootpostlanguage}] // { arity: 1 }
                 Project (#0{rootpostlanguage}) // { arity: 1 }
-                  Filter (#0{rootpostlanguage} = varchar_to_text(#1)) // { arity: 2 }
+                  Filter (#0{rootpostlanguage} = #1) // { arity: 2 }
                     FlatMap unnest_array({"es", "ta", "pt"}) // { arity: 2 }
                       Distinct project=[#0{rootpostlanguage}] // { arity: 1 }
                         Project (#13{rootpostlanguage}) // { arity: 1 }