Treat single quote literals as ranged numbers for inference purposes

crates/compiler/can/src/copy.rs

@@ -259,7 +259,7 @@ fn deep_copy_expr_help<C: CopyEnv>(env: &mut C, copied: &mut Vec<Variable>, expr
         Int(v1, v2, str, val, bound) => Int(sub!(*v1), sub!(*v2), str.clone(), *val, *bound),
         Float(v1, v2, str, val, bound) => Float(sub!(*v1), sub!(*v2), str.clone(), *val, *bound),
         Str(str) => Str(str.clone()),
-        SingleQuote(char) => SingleQuote(*char),
+        SingleQuote(v1, v2, char, bound) => SingleQuote(sub!(*v1), sub!(*v2), *char, *bound),
         List {
             elem_var,
             loc_elems,

crates/compiler/can/src/expr.rs

@@ -22,6 +22,7 @@ use roc_parse::ast::{self, Defs, EscapedChar, StrLiteral};
 use roc_parse::pattern::PatternType::*;
 use roc_problem::can::{PrecedenceProblem, Problem, RuntimeError};
 use roc_region::all::{Loc, Region};
+use roc_types::num::SingleQuoteBound;
 use roc_types::subs::{ExhaustiveMark, IllegalCycleMark, RedundantMark, VarStore, Variable};
 use roc_types::types::{Alias, Category, LambdaSet, OptAbleVar, Type};
 use std::fmt::{Debug, Display};
@@ -91,7 +92,8 @@ pub enum Expr {
     Int(Variable, Variable, Box<str>, IntValue, IntBound),
     Float(Variable, Variable, Box<str>, f64, FloatBound),
     Str(Box<str>),
-    SingleQuote(char),
+    // Number variable, precision variable, value, bound
+    SingleQuote(Variable, Variable, char, SingleQuoteBound),
     List {
         elem_var: Variable,
         loc_elems: Vec<Loc<Expr>>,
@@ -637,7 +639,15 @@ pub fn canonicalize_expr<'a>(
             let mut it = string.chars().peekable();
             if let Some(char) = it.next() {
                 if it.peek().is_none() {
-                    (Expr::SingleQuote(char), Output::default())
+                    (
+                        Expr::SingleQuote(
+                            var_store.fresh(),
+                            var_store.fresh(),
+                            char,
+                            SingleQuoteBound::from_char(char),
+                        ),
+                        Output::default(),
+                    )
                 } else {
                     // multiple chars is found
                     let error = roc_problem::can::RuntimeError::MultipleCharsInSingleQuote(region);
@@ -1642,7 +1652,7 @@ pub fn inline_calls(var_store: &mut VarStore, scope: &mut Scope, expr: Expr) ->
         | other @ Int(..)
         | other @ Float(..)
         | other @ Str { .. }
-        | other @ SingleQuote(_)
+        | other @ SingleQuote(..)
         | other @ RuntimeError(_)
         | other @ EmptyRecord
         | other @ Accessor { .. }
@@ -2703,7 +2713,7 @@ fn get_lookup_symbols(expr: &Expr, var_store: &mut VarStore) -> Vec<(Symbol, Var
             | Expr::Str(_)
             | Expr::ZeroArgumentTag { .. }
             | Expr::Accessor(_)
-            | Expr::SingleQuote(_)
+            | Expr::SingleQuote(..)
             | Expr::EmptyRecord
             | Expr::TypedHole(_)
             | Expr::RuntimeError(_)

crates/compiler/can/src/module.rs

@@ -1038,7 +1038,7 @@ fn fix_values_captured_in_closure_expr(
         | Int(..)
         | Float(..)
         | Str(_)
-        | SingleQuote(_)
+        | SingleQuote(..)
         | Var(_)
         | AbilityMember(..)
         | EmptyRecord

crates/compiler/constrain/src/builtins.rs

@@ -4,7 +4,7 @@ use roc_can::expected::Expected::{self, *};
 use roc_can::num::{FloatBound, FloatWidth, IntBound, IntLitWidth, NumBound, SignDemand};
 use roc_module::symbol::Symbol;
 use roc_region::all::Region;
-use roc_types::num::NumericRange;
+use roc_types::num::{NumericRange, SingleQuoteBound};
 use roc_types::subs::Variable;
 use roc_types::types::Type::{self, *};
 use roc_types::types::{AliasKind, Category};
@@ -99,6 +99,42 @@ pub fn int_literal(
     constraints.exists([num_var], and_constraint)
 }
 
+pub fn single_quote_literal(
+    constraints: &mut Constraints,
+    num_var: Variable,
+    precision_var: Variable,
+    expected: Expected<Type>,
+    region: Region,
+    bound: SingleQuoteBound,
+) -> Constraint {
+    let reason = Reason::IntLiteral;
+
+    // Always add the bound first; this improves the resolved type quality in case it's an alias like "U8".
+    let mut constrs = ArrayVec::<_, 3>::new();
+    let num_type = add_numeric_bound_constr(
+        constraints,
+        &mut constrs,
+        num_var,
+        precision_var,
+        bound,
+        region,
+        Category::Character,
+    );
+
+    constrs.extend([
+        constraints.equal_types(
+            num_type.clone(),
+            ForReason(reason, num_int(Type::Variable(precision_var)), region),
+            Category::Character,
+            region,
+        ),
+        constraints.equal_types(num_type, expected, Category::Character, region),
+    ]);
+
+    let and_constraint = constraints.and_constraint(constrs);
+    constraints.exists([num_var], and_constraint)
+}
+
 #[inline(always)]
 pub fn float_literal(
     constraints: &mut Constraints,
@@ -332,6 +368,16 @@ impl TypedNumericBound for NumBound {
     }
 }
 
+impl TypedNumericBound for SingleQuoteBound {
+    fn numeric_bound(&self) -> NumericBound {
+        match self {
+            &SingleQuoteBound::AtLeast { width } => {
+                NumericBound::Range(NumericRange::IntAtLeastEitherSign(width))
+            }
+        }
+    }
+}
+
 /// A bound placed on a number because of its literal value.
 /// e.g. `-5` cannot be unsigned, and 300 does not fit in a U8
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]

crates/compiler/constrain/src/expr.rs

@@ -1,7 +1,8 @@
 use std::ops::Range;
 
 use crate::builtins::{
-    empty_list_type, float_literal, int_literal, list_type, num_literal, num_u32, str_type,
+    empty_list_type, float_literal, int_literal, list_type, num_literal, single_quote_literal,
+    str_type,
 };
 use crate::pattern::{constrain_pattern, PatternState};
 use roc_can::annotation::IntroducedVariables;
@@ -292,7 +293,14 @@ pub fn constrain_expr(
             constraints.exists(vars, and_constraint)
         }
         Str(_) => constraints.equal_types(str_type(), expected, Category::Str, region),
-        SingleQuote(_) => constraints.equal_types(num_u32(), expected, Category::Character, region),
+        SingleQuote(num_var, precision_var, _, bound) => single_quote_literal(
+            constraints,
+            *num_var,
+            *precision_var,
+            expected,
+            region,
+            *bound,
+        ),
         List {
             elem_var,
             loc_elems,

crates/compiler/mono/src/ir.rs

@@ -4027,12 +4027,18 @@ pub fn with_hole<'a>(
             hole,
         ),
 
-        SingleQuote(character) => Stmt::Let(
+        SingleQuote(_, _, character, _) => {
-            assigned,
+            let layout = layout_cache
-            Expr::Literal(Literal::Int((character as i128).to_ne_bytes())),
+                .from_var(env.arena, variable, env.subs)
-            Layout::int_width(IntWidth::I32),
+                .unwrap();
-            hole,
+
-        ),
+            Stmt::Let(
+                assigned,
+                Expr::Literal(Literal::Int((character as i128).to_ne_bytes())),
+                layout,
+                hole,
+            )
+        }
         LetNonRec(def, cont) => from_can_let(
             env,
             procs,

crates/compiler/solve/tests/solve_expr.rs

@@ -7841,4 +7841,49 @@ mod solve_expr {
             "hasher -> hasher | hasher has Hasher",
         );
     }
+
+    #[test]
+    fn check_char_as_u8() {
+        infer_eq_without_problem(
+            indoc!(
+                r#"
+                x : U8
+                x = '.'
+
+                x
+                "#
+            ),
+            "U8",
+        );
+    }
+
+    #[test]
+    fn check_char_as_u16() {
+        infer_eq_without_problem(
+            indoc!(
+                r#"
+                x : U16
+                x = '.'
+
+                x
+                "#
+            ),
+            "U16",
+        );
+    }
+
+    #[test]
+    fn check_char_as_u32() {
+        infer_eq_without_problem(
+            indoc!(
+                r#"
+                x : U32
+                x = '.'
+
+                x
+                "#
+            ),
+            "U32",
+        );
+    }
 }

crates/compiler/test_derive/src/pretty_print.rs

@@ -58,7 +58,7 @@ fn expr<'a>(c: &Ctx, p: EPrec, f: &'a Arena<'a>, e: &'a Expr) -> DocBuilder<'a,
     match e {
         Num(_, n, _, _) | Int(_, _, n, _, _) | Float(_, _, n, _, _) => f.text(&**n),
         Str(s) => f.text(format!(r#""{}""#, s)),
-        SingleQuote(c) => f.text(format!("'{}'", c)),
+        SingleQuote(_, _, c, _) => f.text(format!("'{}'", c)),
         List {
             elem_var: _,
             loc_elems,

crates/compiler/types/src/num.rs

@@ -328,6 +328,26 @@ pub enum NumBound {
     },
 }
 
+#[derive(Clone, Copy, PartialEq, Eq, Debug)]
+pub enum SingleQuoteBound {
+    AtLeast { width: IntLitWidth },
+}
+
+impl SingleQuoteBound {
+    pub fn from_char(c: char) -> Self {
+        let n = c as u32;
+        let width = if n > u16::MAX as _ {
+            IntLitWidth::U32
+        } else if n > u8::MAX as _ {
+            IntLitWidth::U16
+        } else {
+            IntLitWidth::U8
+        };
+
+        Self::AtLeast { width }
+    }
+}
+
 pub const fn int_lit_width_to_variable(w: IntLitWidth) -> Variable {
     match w {
         IntLitWidth::U8 => Variable::U8,

crates/reporting/tests/test_reporting.rs

@@ -11072,4 +11072,32 @@ All branches in an `if` must have the same type!
         U8
     "###
     );
+
+    test_report!(
+        big_char_does_not_fit_in_u8,
+        indoc!(
+            r#"
+            digits : List U8
+            digits = List.range '0' '9'
+
+            List.contains digits '☃'
+            "#
+        ),
+    @r###"
+    ── TYPE MISMATCH ───────────────────────────────────────── /code/proj/Main.roc ─
+
+    This 2nd argument to `contains` has an unexpected type:
+
+    7│      List.contains digits '☃'
+                                 ^^^^^
+
+    The argument is a Unicode scalar value of type:
+
+        U16, I32, U32, I64, Nat, U64, I128, or U128
+
+    But `contains` needs its 2nd argument to be:
+
+        U8
+    "###
+    );
 }