feat(tvix/eval): Record formals on lambda r/5153

In preparation for both implementing the `functionArgs` builtin and
adding support for validating closed formals, record information about
the formal arguments to a function *on the Lambda itself*. This may seem
a little odd for the purposes of just closed formal checking, but is
something we have to have anyway for builtins.functionArgs so I figured
I'd do it this way to kill both birds with one stone.

Change-Id: Ie3770a607bf352a1eb395c79ca29bb25d5978cd8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7001
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI

4 files changed, 42 insertions, 13 deletions

diff --git a/tvix/eval/src/compiler/mod.rs b/tvix/eval/src/compiler/mod.rs
index b7d2d88530de..259431842956 100644
--- a/tvix/eval/src/compiler/mod.rs
+++ b/tvix/eval/src/compiler/mod.rs
@@ -30,7 +30,7 @@ use crate::errors::{Error, ErrorKind, EvalResult};
 use crate::observer::CompilerObserver;
 use crate::opcode::{CodeIdx, Count, JumpOffset, OpCode, UpvalueIdx};
 use crate::spans::ToSpan;
-use crate::value::{Closure, Lambda, Thunk, Value};
+use crate::value::{Closure, Formals, Lambda, Thunk, Value};
 use crate::warnings::{EvalWarning, WarningKind};
 
 use self::scope::{LocalIdx, LocalPosition, Scope, Upvalue, UpvalueKind};
@@ -775,7 +775,7 @@ impl Compiler<'_> {
     /// many arguments are provided. This is done by emitting a
     /// special instruction that checks the set of keys from a
     /// constant containing the expected keys.
-    fn compile_param_pattern(&mut self, pattern: &ast::Pattern) {
+    fn compile_param_pattern(&mut self, pattern: &ast::Pattern) -> Formals {
         let span = self.span_for(pattern);
         let set_idx = match pattern.pat_bind() {
             Some(name) => self.declare_local(&name, name.ident().unwrap().to_string()),
@@ -792,12 +792,15 @@ impl Compiler<'_> {
         // then finalise any necessary recursion into the scope.
         let mut entries: Vec<(LocalIdx, ast::PatEntry)> = vec![];
         let mut indices: Vec<LocalIdx> = vec![];
+        let mut arguments = HashMap::default();
 
         for entry in pattern.pat_entries() {
             let ident = entry.ident().unwrap();
             let idx = self.declare_local(&ident, ident.to_string());
+            let has_default = entry.default().is_some();
             entries.push((idx, entry));
             indices.push(idx);
+            arguments.insert(ident.into(), has_default);
         }
 
         // For each of the bindings, push the set on the stack and
@@ -836,15 +839,22 @@ impl Compiler<'_> {
 
         // TODO: strictly check if all keys have been consumed if
         // there is no ellipsis.
-        if pattern.ellipsis_token().is_none() {
+        let ellipsis = pattern.ellipsis_token().is_some();
+        if !ellipsis {
             self.emit_warning(pattern, WarningKind::NotImplemented("closed formals"));
         }
+
+        Formals {
+            arguments,
+            ellipsis,
+        }
     }
 
     fn compile_lambda(&mut self, slot: LocalIdx, node: &ast::Lambda) {
-        // Compile the function itself
-        match node.param().unwrap() {
-            ast::Param::Pattern(pat) => self.compile_param_pattern(&pat),
+        // Compile the function itself, recording its formal arguments (if any)
+        // for later use
+        let formals = match node.param().unwrap() {
+            ast::Param::Pattern(pat) => Some(self.compile_param_pattern(&pat)),
 
             ast::Param::IdentParam(param) => {
                 let name = param
@@ -857,10 +867,12 @@ impl Compiler<'_> {
 
                 let idx = self.declare_local(&param, &name);
                 self.scope_mut().mark_initialised(idx);
+                None
             }
-        }
+        };
 
         self.compile(slot, &node.body().unwrap());
+        self.context_mut().lambda.formals = formals;
     }
 
     fn thunk<N, F>(&mut self, outer_slot: LocalIdx, node: &N, content: F)
@@ -1005,10 +1017,7 @@ impl Compiler<'_> {
     /// several operations related to attribute sets, where
     /// identifiers are used as string keys.
     fn emit_literal_ident(&mut self, ident: &ast::Ident) {
-        self.emit_constant(
-            Value::String(ident.ident_token().unwrap().text().into()),
-            ident,
-        );
+        self.emit_constant(Value::String(ident.clone().into()), ident);
     }
 
     /// Patch the jump instruction at the given index, setting its
diff --git a/tvix/eval/src/value/function.rs b/tvix/eval/src/value/function.rs
index 0923e1b1cba9..e58aab19c0e7 100644
--- a/tvix/eval/src/value/function.rs
+++ b/tvix/eval/src/value/function.rs
@@ -1,6 +1,7 @@
 //! This module implements the runtime representation of functions.
 use std::{
     cell::{Ref, RefCell, RefMut},
+    collections::HashMap,
     rc::Rc,
 };
 
@@ -9,6 +10,17 @@ use crate::{
     upvalues::{UpvalueCarrier, Upvalues},
 };
 
+use super::NixString;
+
+#[derive(Clone, Debug, PartialEq)]
+pub(crate) struct Formals {
+    /// Map from argument name, to whether that argument is required
+    pub(crate) arguments: HashMap<NixString, bool>,
+
+    /// Do the formals of this function accept extra arguments
+    pub(crate) ellipsis: bool,
+}
+
 /// The opcodes for a thunk or closure, plus the number of
 /// non-executable opcodes which are allowed after an OpClosure or
 /// OpThunk referencing it.  At runtime `Lambda` is usually wrapped
@@ -16,7 +28,6 @@ use crate::{
 /// quite large).
 #[derive(Debug, PartialEq)]
 pub struct Lambda {
-    // name: Option<NixString>,
     pub(crate) chunk: Chunk,
 
     /// Number of upvalues which the code in this Lambda closes
@@ -24,14 +35,15 @@ pub struct Lambda {
     /// runtime.  Information about the variables is emitted using
     /// data-carrying opcodes (see [`OpCode::DataLocalIdx`]).
     pub(crate) upvalue_count: usize,
+    pub(crate) formals: Option<Formals>,
 }
 
 impl Lambda {
     pub fn new_anonymous() -> Self {
         Lambda {
-            // name: None,
             chunk: Default::default(),
             upvalue_count: 0,
+            formals: None,
         }
     }
 
diff --git a/tvix/eval/src/value/mod.rs b/tvix/eval/src/value/mod.rs
index a1216452fbb5..78f4f5de67a4 100644
--- a/tvix/eval/src/value/mod.rs
+++ b/tvix/eval/src/value/mod.rs
@@ -20,6 +20,7 @@ use crate::opcode::StackIdx;
 use crate::vm::VM;
 pub use attrs::NixAttrs;
 pub use builtin::Builtin;
+pub(crate) use function::Formals;
 pub use function::{Closure, Lambda};
 pub use list::NixList;
 pub use path::canon_path;
diff --git a/tvix/eval/src/value/string.rs b/tvix/eval/src/value/string.rs
index d146ee4cec95..4caef653f2cc 100644
--- a/tvix/eval/src/value/string.rs
+++ b/tvix/eval/src/value/string.rs
@@ -1,5 +1,6 @@
 //! This module implements Nix language strings and their different
 //! backing implementations.
+use rnix::ast;
 use smol_str::SmolStr;
 use std::ffi::OsStr;
 use std::hash::Hash;
@@ -55,6 +56,12 @@ impl From<SmolStr> for NixString {
     }
 }
 
+impl From<ast::Ident> for NixString {
+    fn from(ident: ast::Ident) -> Self {
+        ident.ident_token().unwrap().text().into()
+    }
+}
+
 impl Hash for NixString {
     fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
         self.as_str().hash(state)