WIP

rs/Cargo.toml

@@ -40,7 +40,7 @@ serde_yaml = "0.9"
 # both the schema and the input, so we need to depend on that as well, even
 # though we don't actually do any JSON serialization and deserialization.
 jsonschema = { version = "=0.15.0", default-features = false }
-serde_json = "1"
+serde_json = { version = "1", features = ["preserve_order"] }
 
 # Used for checking identifier syntax (could be removed if regexes don't end up
 # being useful elsewhere too).

rs/src/output/diagnostic.rs

@@ -246,6 +246,9 @@ pub enum Classification {
     #[strum(props(HiddenDescription = "invalid compound vs. simple function name usage"))]
     LinkCompoundVsSimpleFunctionName = 3010,
 
+    #[strum(props(Description = "discouraged name"))]
+    LinkDiscouragedName = 3011,
+
     // Type-related diagnostics (group 4).
     #[strum(props(HiddenDescription = "type-related diagnostics"))]
     Type = 4000,

rs/src/output/extension/namespace.rs

@@ -531,6 +531,22 @@ impl<T> ResolutionResult<T> {
         self.expect(parse_context, if_not_applicable, |_, _| true, true, false)
     }
 
+    /// Emits an error if one or more definitions were found for this name
+    /// resolution, to be used just before defining a new item.
+    pub fn expect_not_yet_defined(&self, parse_context: &mut context::Context) {
+        if !self.visible.is_empty() {
+            traversal::push_diagnostic(
+                parse_context,
+                diagnostic::Level::Error,
+                cause!(
+                    LinkDuplicateDefinition,
+                    "{} is already defined",
+                    self.unresolved_reference
+                ),
+            );
+        }
+    }
+
     /// Silently returns the first matching item, if any. If there are none,
     /// this just returns an unresolved reference. Use
     /// filter_items().expect_one() to formulate error messages if there are
@@ -567,18 +583,4 @@ impl<T> ResolutionResult<T> {
             .next()
             .flatten()
     }
-
-    /// Return an error if one or more definitions were found for this name
-    /// resolution, to be used just before defining a new item.
-    pub fn expect_not_yet_defined(&self) -> diagnostic::Result<()> {
-        if self.visible.is_empty() {
-            Ok(())
-        } else {
-            Err(cause!(
-                LinkDuplicateDefinition,
-                "{} is already defined",
-                self.unresolved_reference
-            ))
-        }
-    }
 }

rs/src/output/extension/simple/type_class.rs

@@ -12,23 +12,27 @@ use crate::output::type_system::meta::pattern::Pattern;
 /// A definition of a user-defined type class.
 #[derive(Clone, Debug, PartialEq, Eq, Default)]
 pub struct Definition {
-    /// The underlying structure of the type.
-    pub structure: Vec<(String, data::class::Simple)>,
+    /// Name of the type class.
+    pub name: String,
 
     /// The parameters expected by the data type.
     pub parameter_slots: Vec<ParameterSlot>,
 
     /// Whether or not the last parameter slot is variadic.
     pub parameters_variadic: bool,
+
+    /// Optional underlying structure of the type.
+    /// TODO: change to Option<Vec<(String, meta::Program)>> and propagate changes
+    pub structure: Vec<(String, data::class::Simple)>,
 }
 
 /// A parameter slot for a user-defined data type.
 #[derive(Clone, Debug, PartialEq, Eq)]
 pub struct ParameterSlot {
-    /// YAML-provided name of the parameter.
+    /// Name of the parameter.
     pub name: String,
 
-    /// YAML-provided human-readable description of the parameter.
+    /// Human-readable description of the parameter.
     pub description: String,
 
     /// Pattern for type- and bounds-checking parameters bound to this slot.

rs/src/parse/extensions/simple/common.rs

@@ -0,0 +1,41 @@
+// SPDX-License-Identifier: Apache-2.0
+
+//! Module providing private helper functions that are not specific to any
+//! particular simple extension construct.
+
+use crate::output::diagnostic::Result;
+use crate::parse::context;
+
+/// Parser for names given to things.
+pub fn parse_name(x: &str, y: &mut context::Context, construct: &str) -> Result<String> {
+    static IDENTIFIER_RE: once_cell::sync::Lazy<regex::Regex> =
+        once_cell::sync::Lazy::new(|| regex::Regex::new("[a-zA-Z_][a-zA-Z0-9_\\.]*").unwrap());
+
+    if x.is_empty() {
+        diagnostic!(
+            y,
+            Info,
+            LinkDiscouragedName,
+            "using the empty string as a {construct} name is not explicitly \
+            illegal, but probably not a good idea"
+        );
+    } else if !IDENTIFIER_RE.is_match(x) {
+        diagnostic!(
+            y,
+            Info,
+            LinkDiscouragedName,
+            "it is recommended for {construct} names to case-insensitively \
+            match [a-z_][a-z0-9_]* for maximum compatibility"
+        );
+    } else if x.contains('.') {
+        diagnostic!(
+            y,
+            Info,
+            LinkDiscouragedName,
+            "using periods within a {construct} name is not explicitly \
+            illegal, but probably not a good idea, as they are also used as \
+            namespace separators for dependencies"
+        );
+    }
+    Ok(x.to_owned())
+}

rs/src/parse/extensions/simple/function_decls.rs → rs/src/parse/extensions/simple/functions.rs

@@ -6,13 +6,13 @@
 use crate::input::yaml;
 use crate::output::diagnostic::Result;
 use crate::parse::context;
-use crate::parse::extensions::simple::builder;
+use crate::parse::extensions::simple::modules;
 
 /// Parse a scalar function declaration.
 pub fn parse_scalar_function(
     _x: &yaml::Value,
     _y: &mut context::Context,
-    _z: &mut builder::Builder,
+    _z: &mut modules::Builder,
 ) -> Result<()> {
     // TODO
     Ok(())
@@ -22,7 +22,7 @@ pub fn parse_scalar_function(
 pub fn parse_aggregate_function(
     _x: &yaml::Value,
     _y: &mut context::Context,
-    _z: &mut builder::Builder,
+    _z: &mut modules::Builder,
 ) -> Result<()> {
     // TODO
     Ok(())

rs/src/parse/extensions/simple/mod.rs

@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: Apache-2.0
 
-//! Module providing parse/validation functions for advanced extensions, i.e.
+//! Module providing parse/validation functions for simple extensions, i.e.
 //! those based around YAML files.
 
 use crate::input::proto::substrait;
@@ -10,12 +10,12 @@ use crate::output::extension::simple::module::Scope;
 use crate::output::type_system::data;
 use crate::parse::context;
 
-mod builder;
+mod common;
 mod derivations;
-mod function_decls;
-mod type_decls;
-mod type_variation_decls;
-mod yaml;
+mod functions;
+mod modules;
+mod type_classes;
+mod type_variations;
 
 /// Parse a user-defined name. Note that names are matched case-insensitively
 /// because we return the name as lowercase.
@@ -51,7 +51,7 @@ fn parse_simple_extension_yaml_uri_mapping(
 ) -> Result<()> {
     // Parse the fields.
     let anchor = proto_primitive_field!(x, y, extension_uri_anchor, parse_anchor).1;
-    let yaml_data = proto_primitive_field!(x, y, uri, yaml::parse_uri)
+    let yaml_data = proto_primitive_field!(x, y, uri, modules::parse_uri)
         .1
         .unwrap();
 

rs/src/parse/extensions/simple/yaml.rs → rs/src/parse/extensions/simple/modules.rs

@@ -1,49 +1,115 @@
 // SPDX-License-Identifier: Apache-2.0
 
-//! Module providing parse/validation functions for parsing YAML extension
-//! files.
+//! Module providing a builder structure to be used while parsing a simple
+//! extension file.
 
 use crate::input::yaml;
 use crate::output::diagnostic::Result;
 use crate::output::extension;
 use crate::output::path;
 use crate::parse::context;
-use crate::parse::extensions::simple::builder;
-use crate::parse::extensions::simple::function_decls;
-use crate::parse::extensions::simple::type_decls;
-use crate::parse::extensions::simple::type_variation_decls;
+use crate::parse::extensions::simple::functions;
+use crate::parse::extensions::simple::modules;
+use crate::parse::extensions::simple::type_classes;
+use crate::parse::extensions::simple::type_variations;
 use crate::parse::traversal;
 use crate::util;
+use std::collections::HashMap;
 use std::sync::Arc;
 
+#[derive(Clone, Debug, Default)]
+pub struct Builder {
+    /// Map with references to dependencies.
+    pub dependencies: HashMap<String, extension::simple::module::Reference>,
+
+    /// Namespace used for type classes defined in this extension and its
+    /// dependencies.
+    pub type_classes: extension::simple::type_class::NamespaceDefinition,
+
+    /// Namespace used for type variations defined in this extension and its
+    /// dependencies.
+    pub type_variations: extension::simple::type_variation::NamespaceDefinition,
+
+    /// Namespace used for functions defined in this extension and its
+    /// dependencies. Both simple and compound names are registered.
+    pub function_impls: extension::simple::function::NamespaceDefinition,
+}
+
+impl From<Builder> for extension::simple::module::Definition {
+    fn from(builder: Builder) -> Self {
+        extension::simple::module::Definition {
+            dependencies: builder.dependencies,
+            type_classes: Arc::new(builder.type_classes),
+            type_variations: Arc::new(builder.type_variations),
+            function_impls: Arc::new(builder.function_impls),
+        }
+    }
+}
+
+impl extension::simple::module::Scope for Builder {
+    /// Resolves a to-be-resolved reference to a type class.
+    fn resolve_type_class<T>(&self, name: T) -> extension::simple::type_class::ResolutionResult
+    where
+        T: Into<extension::simple::type_class::UnresolvedReference>,
+    {
+        self.type_classes.resolve_local(name.into())
+    }
+
+    /// Resolves a to-be-resolved reference to a type variation.
+    fn resolve_type_variation<T>(
+        &self,
+        name: T,
+    ) -> extension::simple::type_variation::ResolutionResult
+    where
+        T: Into<extension::simple::type_variation::UnresolvedReference>,
+    {
+        self.type_variations.resolve_local(name.into())
+    }
+
+    /// Resolves a to-be-resolved reference to a function.
+    fn resolve_function<T>(&self, name: T) -> extension::simple::function::ResolutionResult
+    where
+        T: Into<extension::simple::function::UnresolvedReference>,
+    {
+        self.function_impls.resolve_local(name.into())
+    }
+}
+
 /// Toplevel parse function for a simple extension YAML file.
 fn parse_root(
     x: &yaml::Value,
     y: &mut context::Context,
 ) -> Result<extension::simple::module::Definition> {
-    let mut builder = builder::Builder::default();
-    yaml_repeated_field!(x, y, "types", type_decls::parse_type, 0, &mut builder)?;
+    let mut builder = modules::Builder::default();
+    yaml_repeated_field!(
+        x,
+        y,
+        "types",
+        type_classes::parse_type_class,
+        0,
+        &mut builder
+    )?;
     yaml_repeated_field!(
         x,
         y,
         "type_variations",
-        type_variation_decls::parse_type_variation,
+        type_variations::parse_type_variation,
         0,
         &mut builder
     )?;
     yaml_repeated_field!(
         x,
         y,
         "scalar_functions",
-        function_decls::parse_scalar_function,
+        functions::parse_scalar_function,
         0,
         &mut builder
     )?;
     yaml_repeated_field!(
         x,
         y,
         "aggregate_functions",
-        function_decls::parse_aggregate_function,
+        functions::parse_aggregate_function,
         0,
         &mut builder
     )?;