feat: Open lists and tuples in Term (#2360)

zrho · web-flow · commit 292af8010dba · 2025-06-26T10:44:56.000Z
This PR introduces open lists and tuples in `core` terms. The `Term` type is extended with `Term::ListConcat` and `Term::TupleConcat` variants and helper methods are added to deal with them. Closes #2294.
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/hugr-core/Cargo.toml b/hugr-core/Cargo.toml
@@ -63,6 +63,9 @@ semver = { workspace = true, features = ["serde"] }
 zstd = { workspace = true, optional = true }
 ordered-float = { workspace = true, features = ["serde"] }
 base64.workspace = true
+relrc = { workspace = true, features = ["petgraph", "serde"] }
+smallvec = "1.15.0"
+tracing = "0.1.41"
 
 [dev-dependencies]
 rstest = { workspace = true }
diff --git a/hugr-core/src/export.rs b/hugr-core/src/export.rs
@@ -945,22 +945,17 @@ impl<'a> Context<'a> {
 
                 self.make_term_apply(model::CORE_TYPE, &[])
             }
-            Term::BoundedNatType { .. } => self.make_term_apply(model::CORE_NAT_TYPE, &[]),
+            Term::BoundedNatType(_) => self.make_term_apply(model::CORE_NAT_TYPE, &[]),
             Term::StringType => self.make_term_apply(model::CORE_STR_TYPE, &[]),
             Term::BytesType => self.make_term_apply(model::CORE_BYTES_TYPE, &[]),
             Term::FloatType => self.make_term_apply(model::CORE_FLOAT_TYPE, &[]),
             Term::ListType(item_type) => {
                 let item_type = self.export_term(item_type, None);
                 self.make_term_apply(model::CORE_LIST_TYPE, &[item_type])
             }
-            Term::TupleType(params) => {
-                let item_types = self.bump.alloc_slice_fill_iter(
-                    params
-                        .iter()
-                        .map(|param| table::SeqPart::Item(self.export_term(param, None))),
-                );
-                let types = self.make_term(table::Term::List(item_types));
-                self.make_term_apply(model::CORE_TUPLE_TYPE, &[types])
+            Term::TupleType(item_types) => {
+                let item_types = self.export_term(item_types, None);
+                self.make_term_apply(model::CORE_TUPLE_TYPE, &[item_types])
             }
             Term::Runtime(ty) => self.export_type(ty),
             Term::BoundedNat(value) => self.make_term(model::Literal::Nat(*value).into()),
@@ -975,6 +970,14 @@ impl<'a> Context<'a> {
                 );
                 self.make_term(table::Term::List(parts))
             }
+            Term::ListConcat(lists) => {
+                let parts = self.bump.alloc_slice_fill_iter(
+                    lists
+                        .iter()
+                        .map(|elem| table::SeqPart::Splice(self.export_term(elem, None))),
+                );
+                self.make_term(table::Term::List(parts))
+            }
             Term::Tuple(elems) => {
                 let parts = self.bump.alloc_slice_fill_iter(
                     elems
@@ -983,6 +986,14 @@ impl<'a> Context<'a> {
                 );
                 self.make_term(table::Term::Tuple(parts))
             }
+            Term::TupleConcat(tuples) => {
+                let parts = self.bump.alloc_slice_fill_iter(
+                    tuples
+                        .iter()
+                        .map(|elem| table::SeqPart::Splice(self.export_term(elem, None))),
+                );
+                self.make_term(table::Term::Tuple(parts))
+            }
             Term::Variable(v) => self.export_type_arg_var(v),
             Term::StaticType => self.make_term_apply(model::CORE_STATIC, &[]),
         }
diff --git a/hugr-core/src/extension/resolution/types.rs b/hugr-core/src/extension/resolution/types.rs
@@ -231,8 +231,16 @@ pub(super) fn collect_term_exts(
             collect_term_exts(item_type, used_extensions, missing_extensions)
         }
         Term::TupleType(item_types) => {
-            for item_type in item_types {
-                collect_term_exts(item_type, used_extensions, missing_extensions);
+            collect_term_exts(item_types, used_extensions, missing_extensions)
+        }
+        Term::ListConcat(lists) => {
+            for list in lists {
+                collect_term_exts(list, used_extensions, missing_extensions);
+            }
+        }
+        Term::TupleConcat(tuples) => {
+            for tuple in tuples {
+                collect_term_exts(tuple, used_extensions, missing_extensions);
             }
         }
         Term::Variable(_)
diff --git a/hugr-core/src/extension/resolution/types_mut.rs b/hugr-core/src/extension/resolution/types_mut.rs
@@ -222,23 +222,19 @@ pub(super) fn resolve_term_exts(
 ) -> Result<(), ExtensionResolutionError> {
     match term {
         Term::Runtime(ty) => resolve_type_exts(node, ty, extensions, used_extensions)?,
-        Term::List(elems) => {
-            for elem in elems.iter_mut() {
-                resolve_term_exts(node, elem, extensions, used_extensions)?;
-            }
-        }
-        Term::Tuple(elems) => {
-            for elem in elems.iter_mut() {
-                resolve_term_exts(node, elem, extensions, used_extensions)?;
+        Term::List(children)
+        | Term::ListConcat(children)
+        | Term::Tuple(children)
+        | Term::TupleConcat(children) => {
+            for child in children.iter_mut() {
+                resolve_term_exts(node, child, extensions, used_extensions)?;
             }
         }
         Term::ListType(item_type) => {
-            resolve_term_exts(node, item_type, extensions, used_extensions)?;
+            resolve_term_exts(node, item_type.as_mut(), extensions, used_extensions)?;
         }
         Term::TupleType(item_types) => {
-            for item_type in item_types.iter_mut() {
-                resolve_term_exts(node, item_type, extensions, used_extensions)?;
-            }
+            resolve_term_exts(node, item_types.as_mut(), extensions, used_extensions)?;
         }
         Term::Variable(_)
         | Term::RuntimeType(_)
diff --git a/hugr-core/src/hugr/serialize/test.rs b/hugr-core/src/hugr/serialize/test.rs
@@ -482,7 +482,7 @@ fn polyfunctype2() -> PolyFuncTypeRV {
 #[case(PolyFuncType::new([TypeParam::StringType], Signature::new_endo(vec![Type::new_var_use(0, TypeBound::Copyable)])))]
 #[case(PolyFuncType::new([TypeBound::Copyable.into()], Signature::new_endo(vec![Type::new_var_use(0, TypeBound::Copyable)])))]
 #[case(PolyFuncType::new([TypeParam::new_list_type(TypeBound::Any)], Signature::new_endo(type_row![])))]
-#[case(PolyFuncType::new([TypeParam::TupleType([TypeBound::Any.into(), TypeParam::bounded_nat_type(2.try_into().unwrap())].into())], Signature::new_endo(type_row![])))]
+#[case(PolyFuncType::new([TypeParam::new_tuple_type([TypeBound::Any.into(), TypeParam::bounded_nat_type(2.try_into().unwrap())])], Signature::new_endo(type_row![])))]
 #[case(PolyFuncType::new(
     [TypeParam::new_list_type(TypeBound::Any)],
     Signature::new_endo(Type::new_tuple(TypeRV::new_row_var_use(0, TypeBound::Any)))))]
@@ -495,7 +495,7 @@ fn roundtrip_polyfunctype_fixedlen(#[case] poly_func_type: PolyFuncType) {
 #[case(PolyFuncTypeRV::new([TypeParam::StringType], FuncValueType::new_endo(vec![Type::new_var_use(0, TypeBound::Copyable)])))]
 #[case(PolyFuncTypeRV::new([TypeBound::Copyable.into()], FuncValueType::new_endo(vec![Type::new_var_use(0, TypeBound::Copyable)])))]
 #[case(PolyFuncTypeRV::new([TypeParam::new_list_type(TypeBound::Any)], FuncValueType::new_endo(type_row![])))]
-#[case(PolyFuncTypeRV::new([TypeParam::TupleType([TypeBound::Any.into(), TypeParam::bounded_nat_type(2.try_into().unwrap())].into())], FuncValueType::new_endo(type_row![])))]
+#[case(PolyFuncTypeRV::new([TypeParam::new_tuple_type([TypeBound::Any.into(), TypeParam::bounded_nat_type(2.try_into().unwrap())])], FuncValueType::new_endo(type_row![])))]
 #[case(PolyFuncTypeRV::new(
     [TypeParam::new_list_type(TypeBound::Any)],
     FuncValueType::new_endo(TypeRV::new_row_var_use(0, TypeBound::Any))))]
diff --git a/hugr-core/src/import.rs b/hugr-core/src/import.rs
@@ -25,13 +25,14 @@ use crate::{
     types::{
         CustomType, FuncTypeBase, MaybeRV, PolyFuncType, PolyFuncTypeBase, RowVariable, Signature,
         Term, Type, TypeArg, TypeBase, TypeBound, TypeEnum, TypeName, TypeRow,
-        type_param::TypeParam, type_row::TypeRowBase,
+        type_param::{SeqPart, TypeParam},
+        type_row::TypeRowBase,
     },
 };
 use fxhash::FxHashMap;
 use hugr_model::v0 as model;
 use hugr_model::v0::table;
-use itertools::Either;
+use itertools::{Either, Itertools};
 use smol_str::{SmolStr, ToSmolStr};
 use thiserror::Error;
 
@@ -1255,14 +1256,10 @@ impl<'a> Context<'a> {
             if let Some([item_types]) = self.match_symbol(term_id, model::CORE_TUPLE_TYPE)? {
                 // At present `hugr-model` has no way to express that the item
                 // types of a tuple must be copyable. Therefore we import it as `Any`.
-                let item_types = (|| {
-                    self.import_closed_list(item_types)?
-                        .into_iter()
-                        .map(|param| self.import_term(param))
-                        .collect::<Result<_, _>>()
-                })()
-                .map_err(|err| error_context!(err, "item types of tuple type"))?;
-                return Ok(TypeParam::TupleType(item_types));
+                let item_types = self
+                    .import_term(item_types)
+                    .map_err(|err| error_context!(err, "item types of tuple type"))?;
+                return Ok(TypeParam::new_tuple_type(item_types));
             }
 
             match self.get_term(term_id)? {
@@ -1277,28 +1274,24 @@ impl<'a> Context<'a> {
                     Ok(Term::new_var_use(var.1 as _, decl))
                 }
 
-                table::Term::List { .. } => {
-                    let elems = (|| {
-                        self.import_closed_list(term_id)?
-                            .iter()
-                            .map(|item| self.import_term(*item))
-                            .collect::<Result<_, _>>()
-                    })()
-                    .map_err(|err| error_context!(err, "list items"))?;
-
-                    Ok(Term::List(elems))
+                table::Term::List(parts) => {
+                    // PERFORMANCE: Can we do this without the additional allocation?
+                    let parts: Vec<_> = parts
+                        .iter()
+                        .map(|part| self.import_seq_part(part))
+                        .collect::<Result<_, _>>()
+                        .map_err(|err| error_context!(err, "list parts"))?;
+                    Ok(TypeArg::new_list_from_parts(parts))
                 }
 
-                table::Term::Tuple { .. } => {
-                    let elems = (|| {
-                        self.import_closed_list(term_id)?
-                            .iter()
-                            .map(|item| self.import_term(*item))
-                            .collect::<Result<_, _>>()
-                    })()
-                    .map_err(|err| error_context!(err, "tuple items"))?;
-
-                    Ok(Term::Tuple(elems))
+                table::Term::Tuple(parts) => {
+                    // PERFORMANCE: Can we do this without the additional allocation?
+                    let parts: Vec<_> = parts
+                        .iter()
+                        .map(|part| self.import_seq_part(part))
+                        .try_collect()
+                        .map_err(|err| error_context!(err, "tuple parts"))?;
+                    Ok(TypeArg::new_tuple_from_parts(parts))
                 }
 
                 table::Term::Literal(model::Literal::Str(value)) => {
@@ -1322,6 +1315,16 @@ impl<'a> Context<'a> {
         .map_err(|err| error_context!(err, "term {}", term_id))
     }
 
+    fn import_seq_part(
+        &mut self,
+        seq_part: &'a table::SeqPart,
+    ) -> Result<SeqPart<TypeArg>, ImportError> {
+        Ok(match seq_part {
+            table::SeqPart::Item(term_id) => SeqPart::Item(self.import_term(*term_id)?),
+            table::SeqPart::Splice(term_id) => SeqPart::Splice(self.import_term(*term_id)?),
+        })
+    }
+
     /// Import a `Type` from a term that represents a runtime type.
     fn import_type<RV: MaybeRV>(
         &mut self,
diff --git a/hugr-core/src/types/poly_func.rs b/hugr-core/src/types/poly_func.rs
@@ -271,7 +271,7 @@ pub(crate) mod test {
         for decl in [
             Term::new_list_type(Term::max_nat_type()),
             Term::StringType,
-            Term::TupleType(vec![TypeBound::Any.into(), Term::max_nat_type()]),
+            Term::new_tuple_type([TypeBound::Any.into(), Term::max_nat_type()]),
         ] {
             let invalid_ts = PolyFuncTypeBase::new_validated([decl.clone()], body_type.clone());
             assert_eq!(
diff --git a/hugr-core/src/types/serialize.rs b/hugr-core/src/types/serialize.rs
@@ -78,7 +78,7 @@ pub(super) enum TypeParamSer {
     Float,
     StaticType,
     List { param: Box<Term> },
-    Tuple { params: Vec<Term> },
+    Tuple { params: ArrayOrTermSer },
 }
 
 #[derive(Clone, Debug, serde::Deserialize, serde::Serialize)]
@@ -104,9 +104,15 @@ pub(super) enum TypeArgSer {
     List {
         elems: Vec<Term>,
     },
+    ListConcat {
+        lists: Vec<Term>,
+    },
     Tuple {
         elems: Vec<Term>,
     },
+    TupleConcat {
+        tuples: Vec<Term>,
+    },
     Variable {
         #[serde(flatten)]
         v: TermVar,
@@ -130,15 +136,19 @@ impl From<Term> for TermSer {
             Term::BytesType => TermSer::TypeParam(TypeParamSer::Bytes),
             Term::FloatType => TermSer::TypeParam(TypeParamSer::Float),
             Term::ListType(param) => TermSer::TypeParam(TypeParamSer::List { param }),
-            Term::TupleType(params) => TermSer::TypeParam(TypeParamSer::Tuple { params }),
             Term::Runtime(ty) => TermSer::TypeArg(TypeArgSer::Type { ty }),
+            Term::TupleType(params) => TermSer::TypeParam(TypeParamSer::Tuple {
+                params: (*params).into(),
+            }),
             Term::BoundedNat(n) => TermSer::TypeArg(TypeArgSer::BoundedNat { n }),
             Term::String(arg) => TermSer::TypeArg(TypeArgSer::String { arg }),
             Term::Bytes(value) => TermSer::TypeArg(TypeArgSer::Bytes { value }),
             Term::Float(value) => TermSer::TypeArg(TypeArgSer::Float { value }),
             Term::List(elems) => TermSer::TypeArg(TypeArgSer::List { elems }),
             Term::Tuple(elems) => TermSer::TypeArg(TypeArgSer::Tuple { elems }),
             Term::Variable(v) => TermSer::TypeArg(TypeArgSer::Variable { v }),
+            Term::ListConcat(lists) => TermSer::TypeArg(TypeArgSer::ListConcat { lists }),
+            Term::TupleConcat(tuples) => TermSer::TypeArg(TypeArgSer::TupleConcat { tuples }),
         }
     }
 }
@@ -154,7 +164,7 @@ impl From<TermSer> for Term {
                 TypeParamSer::Bytes => Term::BytesType,
                 TypeParamSer::Float => Term::FloatType,
                 TypeParamSer::List { param } => Term::ListType(param),
-                TypeParamSer::Tuple { params } => Term::TupleType(params),
+                TypeParamSer::Tuple { params } => Term::TupleType(Box::new(params.into())),
             },
             TermSer::TypeArg(arg) => match arg {
                 TypeArgSer::Type { ty } => Term::Runtime(ty),
@@ -165,11 +175,39 @@ impl From<TermSer> for Term {
                 TypeArgSer::List { elems } => Term::List(elems),
                 TypeArgSer::Tuple { elems } => Term::Tuple(elems),
                 TypeArgSer::Variable { v } => Term::Variable(v),
+                TypeArgSer::ListConcat { lists } => Term::ListConcat(lists),
+                TypeArgSer::TupleConcat { tuples } => Term::TupleConcat(tuples),
             },
         }
     }
 }
 
+/// Helper type that serialises lists as JSON arrays for compatibility.
+#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
+#[serde(untagged)]
+pub(super) enum ArrayOrTermSer {
+    Array(Vec<Term>),
+    Term(Box<Term>),
+}
+
+impl From<ArrayOrTermSer> for Term {
+    fn from(value: ArrayOrTermSer) -> Self {
+        match value {
+            ArrayOrTermSer::Array(terms) => Term::new_list(terms),
+            ArrayOrTermSer::Term(term) => *term,
+        }
+    }
+}
+
+impl From<Term> for ArrayOrTermSer {
+    fn from(term: Term) -> Self {
+        match term {
+            Term::List(terms) => ArrayOrTermSer::Array(terms),
+            term => ArrayOrTermSer::Term(Box::new(term)),
+        }
+    }
+}
+
 /// Helper for to serialize and deserialize the byte string in [`TypeArg::Bytes`] via base64.
 mod base64 {
     use std::sync::Arc;
diff --git a/hugr-core/src/types/type_param.rs b/hugr-core/src/types/type_param.rs
diff --git a/hugr-py/src/hugr/_serialization/tys.py b/hugr-py/src/hugr/_serialization/tys.py
diff --git a/hugr-py/src/hugr/tys.py b/hugr-py/src/hugr/tys.py

Original file line number	Diff line number	Diff line change
`@@ -231,8 +231,16 @@ pub(super) fn collect_term_exts(`
`231`	`231`	`collect_term_exts(item_type, used_extensions, missing_extensions)`
`232`	`232`	`}`
`233`	`233`	`Term::TupleType(item_types) => {`
`234`		`- for item_type in item_types {`
`235`		`- collect_term_exts(item_type, used_extensions, missing_extensions);`
	`234`	`+ collect_term_exts(item_types, used_extensions, missing_extensions)`
	`235`	`+ }`
	`236`	`+ Term::ListConcat(lists) => {`
	`237`	`+ for list in lists {`
	`238`	`+ collect_term_exts(list, used_extensions, missing_extensions);`
	`239`	`+ }`
	`240`	`+ }`
	`241`	`+ Term::TupleConcat(tuples) => {`
	`242`	`+ for tuple in tuples {`
	`243`	`+ collect_term_exts(tuple, used_extensions, missing_extensions);`
`236`	`244`	`}`
`237`	`245`	`}`
`238`	`246`	`Term::Variable(_)`