Add partiql-types and literals typing

This is the first commit to introduce types to `partiql-lang-rust`.
With the changes in this commit, we're introducing `partiql-types`
create which includes the initial model for our built-in types.
We also introduce an `AstTyper` in `partiql-ast-passes` which
provides an API for receiving `partiql-ast` and outputting a
mapping from AST Node Ids to their corresponding types.

In the first version, we're loosely typing the literal with the
expection that this goes through multiple iterations.

Here is the high-level plan:
[ ] Type literals
[ ] Type variable references in SFW
[ ] Type operators
[ ] Type using schemas in typing environment
[ ] Add type coercions (using a Type Lattice?)
[ ] Update CHANGELOG.md and adh documentation

CHANGELOG.md

@@ -6,6 +6,12 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
+### Changed
+- *BREAKING:* partiql-logical-planner: moves `NameResolver` to `partiql-ast-passes`
+
+### Added
+- Adds `partiql-types` crate that includes data models for PartiQL Types.
+- Adds `partiql_ast_passes::typer` for type annotating the AST.
 
 ## [0.5.0] - 2023-06-06
 ### Changed

partiql-ast-passes/Cargo.toml

@@ -21,8 +21,11 @@ bench = false
 
 [dependencies]
 partiql-ast = { path = "../partiql-ast", version = "0.5.*" }
+partiql-parser = { path = "../partiql-parser", version = "0.5.*" }
 partiql-catalog = { path = "../partiql-catalog", version = "0.5.*" }
+partiql-types = { path = "../partiql-types", version = "0.5.*" }
 
+assert_matches = "1.5.*"
 fnv = "1"
 indexmap = "1.9"
 thiserror = "1.0"

partiql-ast-passes/src/lib.rs

@@ -6,3 +6,4 @@
 
 pub mod error;
 pub mod name_resolver;
+pub mod typer;

partiql-ast-passes/src/typer.rs

@@ -0,0 +1,270 @@
+use crate::error::{AstTransformError, AstTransformationError};
+use indexmap::IndexMap;
+use partiql_ast::ast;
+use partiql_ast::ast::{Bag, Expr, List, Lit, NodeId, Query, QuerySet};
+use partiql_ast::visit::{Traverse, Visit, Visitor};
+use partiql_catalog::Catalog;
+use partiql_types::{ArrayType, BagType, StaticType, StaticTypeKind, StructType};
+
+pub type AstTypeMap<T> = IndexMap<ast::NodeId, T>;
+
+#[derive(Debug, Clone)]
+#[allow(dead_code)]
+pub struct AstTyper<'c, T> {
+    id_stack: Vec<NodeId>,
+    container_stack: Vec<Vec<StaticType>>,
+    errors: Vec<AstTransformError>,
+    type_map: AstTypeMap<T>,
+    catalog: &'c dyn Catalog,
+}
+
+impl<'c> AstTyper<'c, StaticType> {
+    pub fn new(catalog: &'c dyn Catalog) -> Self {
+        AstTyper {
+            id_stack: Default::default(),
+            container_stack: Default::default(),
+            errors: Default::default(),
+            type_map: Default::default(),
+            catalog,
+        }
+    }
+
+    pub fn map_types(
+        mut self,
+        query: &ast::AstNode<ast::Query>,
+    ) -> Result<AstTypeMap<StaticType>, AstTransformationError> {
+        query.visit(&mut self);
+        if !self.errors.is_empty() {
+            return Err(AstTransformationError {
+                errors: self.errors,
+            });
+        }
+        Ok(self.type_map)
+    }
+
+    #[inline]
+    fn current_node(&self) -> &NodeId {
+        self.id_stack.last().unwrap()
+    }
+}
+
+impl<'c, 'ast> Visitor<'ast> for AstTyper<'c, StaticType> {
+    fn enter_ast_node(&mut self, id: NodeId) -> Traverse {
+        self.id_stack.push(id);
+        Traverse::Continue
+    }
+    fn exit_ast_node(&mut self, id: NodeId) -> Traverse {
+        assert_eq!(self.id_stack.pop(), Some(id));
+        Traverse::Continue
+    }
+
+    fn enter_query(&mut self, _query: &'ast Query) -> Traverse {
+        Traverse::Continue
+    }
+
+    fn exit_query(&mut self, _query: &'ast Query) -> Traverse {
+        Traverse::Continue
+    }
+
+    fn enter_query_set(&mut self, _query_set: &'ast QuerySet) -> Traverse {
+        match _query_set {
+            QuerySet::SetOp(_) => {
+                todo!()
+            }
+            QuerySet::Select(_) => {}
+            QuerySet::Expr(_) => {}
+            QuerySet::Values(_) => {
+                todo!()
+            }
+            QuerySet::Table(_) => {
+                todo!()
+            }
+        }
+        Traverse::Continue
+    }
+
+    fn exit_query_set(&mut self, _query_set: &'ast QuerySet) -> Traverse {
+        Traverse::Continue
+    }
+
+    fn enter_expr(&mut self, _expr: &'ast Expr) -> Traverse {
+        Traverse::Continue
+    }
+
+    fn exit_expr(&mut self, _expr: &'ast Expr) -> Traverse {
+        Traverse::Continue
+    }
+
+    fn enter_lit(&mut self, _lit: &'ast Lit) -> Traverse {
+        let kind = match _lit {
+            Lit::Null => StaticTypeKind::Null,
+            Lit::Missing => StaticTypeKind::Missing,
+            Lit::Int8Lit(_) => StaticTypeKind::Int,
+            Lit::Int16Lit(_) => StaticTypeKind::Int,
+            Lit::Int32Lit(_) => StaticTypeKind::Int,
+            Lit::Int64Lit(_) => StaticTypeKind::Int,
+            Lit::DecimalLit(_) => StaticTypeKind::Decimal,
+            Lit::NumericLit(_) => StaticTypeKind::Decimal,
+            Lit::RealLit(_) => StaticTypeKind::Float32,
+            Lit::FloatLit(_) => StaticTypeKind::Float32,
+            Lit::DoubleLit(_) => StaticTypeKind::Float64,
+            Lit::BoolLit(_) => StaticTypeKind::Bool,
+            Lit::IonStringLit(_) => todo!(),
+            Lit::CharStringLit(_) => StaticTypeKind::String,
+            Lit::NationalCharStringLit(_) => StaticTypeKind::String,
+            Lit::BitStringLit(_) => todo!(),
+            Lit::HexStringLit(_) => todo!(),
+            Lit::StructLit(_) => StaticTypeKind::Struct(StructType::unconstrained()),
+            Lit::ListLit(_) => StaticTypeKind::Array(ArrayType::array()),
+            Lit::BagLit(_) => StaticTypeKind::Bag(BagType::bag()),
+            Lit::TypedLit(_, _) => todo!(),
+        };
+
+        let ty = StaticType::new(kind);
+
+        let id = *self.current_node();
+        if let Some(c) = self.container_stack.last_mut() {
+            c.push(ty.clone())
+        }
+        self.type_map.insert(id, ty);
+        Traverse::Continue
+    }
+
+    fn enter_struct(&mut self, _struct: &'ast ast::Struct) -> Traverse {
+        self.container_stack.push(vec![]);
+        Traverse::Continue
+    }
+
+    fn exit_struct(&mut self, _struct: &'ast ast::Struct) -> Traverse {
+        let id = *self.current_node();
+        let fields = self.container_stack.pop();
+        // TODO Check if all keys are StaticType::String
+        match fields {
+            Some(f) => {
+                f.chunks(2).for_each(|w| {
+                    match w[0].is_string() {
+                        true => {}
+                        // TODO remove the panic with Traverse::Continue and capturing the error
+                        _ => {
+                            self.errors.push(AstTransformError::IllegalState(
+                                "Struct keys can only be of String type".to_string(),
+                            ));
+                        }
+                    }
+                });
+
+                let ty = StaticType::new_struct(StructType::unconstrained());
+                self.type_map.insert(id, ty.clone());
+                if let Some(c) = self.container_stack.last_mut() {
+                    c.push(ty)
+                }
+            }
+            None => {
+                let ty = StaticType::new_struct(StructType::unconstrained());
+                self.type_map.insert(id, ty.clone());
+                if let Some(c) = self.container_stack.last_mut() {
+                    c.push(ty)
+                }
+            }
+        }
+
+        Traverse::Continue
+    }
+
+    fn enter_bag(&mut self, _bag: &'ast Bag) -> Traverse {
+        self.container_stack.push(vec![]);
+        Traverse::Continue
+    }
+
+    fn exit_bag(&mut self, _bag: &'ast Bag) -> Traverse {
+        // TODO add schema validation of BAG elements, e.g. for Schema Bag<Int> if there is at least
+        // one element that isn't INT there is a type checking error.
+
+        // TODO clarify if we need to record the internal types of bag literal or stick w/Schema?
+        self.container_stack.pop();
+
+        let id = *self.current_node();
+        let ty = StaticType::new_bag(BagType::bag());
+
+        self.type_map.insert(id, ty.clone());
+        if let Some(s) = self.container_stack.last_mut() {
+            s.push(ty)
+        }
+        Traverse::Continue
+    }
+
+    fn enter_list(&mut self, _list: &'ast List) -> Traverse {
+        self.container_stack.push(vec![]);
+        Traverse::Continue
+    }
+
+    fn exit_list(&mut self, _list: &'ast List) -> Traverse {
+        // TODO clarify if we need to record the internal types of bag literal or stick w/Schema?
+        // one element that isn't INT there is a type checking error.
+
+        // We don't care about internal elements for now
+        self.container_stack.pop();
+
+        let id = *self.current_node();
+        let ty = StaticType::new_array(ArrayType::array());
+
+        self.type_map.insert(id, ty.clone());
+        if let Some(s) = self.container_stack.last_mut() {
+            s.push(ty)
+        }
+        Traverse::Continue
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use assert_matches::assert_matches;
+    use partiql_ast::ast;
+    use partiql_catalog::PartiqlCatalog;
+    use partiql_types::{StaticType, StaticTypeKind};
+
+    #[test]
+    fn simple_test() {
+        assert_matches!(run_literal_test("NULL"), StaticTypeKind::Null);
+        assert_matches!(run_literal_test("MISSING"), StaticTypeKind::Missing);
+        assert_matches!(run_literal_test("Missing"), StaticTypeKind::Missing);
+        assert_matches!(run_literal_test("true"), StaticTypeKind::Bool);
+        assert_matches!(run_literal_test("false"), StaticTypeKind::Bool);
+        assert_matches!(run_literal_test("1"), StaticTypeKind::Int);
+        assert_matches!(run_literal_test("1.5"), StaticTypeKind::Decimal);
+        assert_matches!(run_literal_test("'hello world!'"), StaticTypeKind::String);
+        assert_matches!(
+            run_literal_test("[1, 2 , {'a': b}]"),
+            StaticTypeKind::Array(_)
+        );
+        assert_matches!(
+            run_literal_test("<<'1', {'a': b}>>"),
+            StaticTypeKind::Bag(_)
+        );
+        assert_matches!(
+            run_literal_test("{'a': 1, 'b': 3, 'c': [1, 2]}"),
+            StaticTypeKind::Struct(_)
+        );
+    }
+
+    fn run_literal_test(q: &str) -> StaticTypeKind {
+        let out = type_statement(q);
+        let values: Vec<&StaticType> = out.values().collect();
+        values.last().unwrap().kind().clone()
+    }
+
+    fn type_statement(q: &str) -> AstTypeMap<StaticType> {
+        let parsed = partiql_parser::Parser::default()
+            .parse(q)
+            .expect("Expect successful parse");
+
+        let catalog = PartiqlCatalog::default();
+        let typer = AstTyper::new(&catalog);
+        if let ast::Expr::Query(q) = parsed.ast.as_ref() {
+            typer.map_types(&q).expect("type map")
+        } else {
+            panic!("Typing statement other than `Query` are unsupported")
+        }
+    }
+}

partiql-ast/Cargo.toml

@@ -21,13 +21,10 @@ bench = false
 
 [dependencies]
 rust_decimal = { version = "1.25.0", default-features = false, features = ["std"] }
-
 serde = { version = "1.*", features = ["derive"], optional = true }
 
-
 [dev-dependencies]
 
-
 [features]
 default = []
 serde = [

partiql-ast/src/ast.rs

@@ -25,6 +25,7 @@ pub struct NodeId(pub u32);
 #[derive(Clone, Debug, Eq, PartialEq)]
 #[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
 pub struct AstNode<T> {
+    // pub typ: Option<Box<dyn partiql_types::Typ>>,
     pub id: NodeId,
     pub node: T,
 }

partiql-catalog/Cargo.toml

@@ -24,6 +24,7 @@ bench = false
 partiql-value = { path = "../partiql-value", version = "0.5.*" }
 partiql-parser = { path = "../partiql-parser", version = "0.5.*" }
 partiql-logical = { path = "../partiql-logical", version = "0.5.*" }
+
 thiserror = "1.0"
 ordered-float = "3.*"
 itertools = "0.10.*"

partiql-logical-planner/Cargo.toml

@@ -28,6 +28,7 @@ partiql-ast = { path = "../partiql-ast", version = "0.5.*" }
 partiql-parser = { path = "../partiql-parser", version = "0.5.*" }
 partiql-catalog = { path = "../partiql-catalog", version = "0.5.*" }
 partiql-ast-passes = { path = "../partiql-ast-passes", version = "0.5.*" }
+
 ion-rs = "0.17"
 ordered-float = "3.*"
 itertools = "0.10.*"

partiql-types/Cargo.toml

@@ -21,6 +21,7 @@ edition.workspace = true
 bench = false
 
 [dependencies]
+
 ordered-float = "3.*"
 itertools = "0.10.*"
 unicase = "2.6"