From d4fa3152e92ef72d9ee050000b1fd4952203e383 Mon Sep 17 00:00:00 2001
From: Griffin Smith <grfn@gws.fyi>
Date: Wed, 12 Oct 2022 22:47:23 -0400
Subject: feat(tvix/eval): Implement builtins.deepSeq

This is done via a new `deepForce` function on Value. Since values can
be cyclical (for example, see the test-case), we need to do some extra
work to avoid RefCell borrow errors if we ever hit a graph cycle:

While deep-forcing values, we keep a set of thunks that we have
already seen and avoid doing any work on the same thunk twice. The set
is encapsulated in a separate type to stop potentially invalid
pointers from leaking out.

Finally, since deep_force is conceptually similar to
`VM::force_for_output` (but more suited to usage in eval since it
doesn't clone the values) this removes the latter, replacing it with
the former.

Co-Authored-By: Vincent Ambo <tazjin@tvl.su>
Change-Id: Iefddefcf09fae3b6a4d161a5873febcff54b9157
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7000
Tested-by: BuildkiteCI
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
---
 tvix/eval/src/value/list.rs  | 10 ++++++++++
 tvix/eval/src/value/mod.rs   | 45 ++++++++++++++++++++++++++++++++++++++++++++
 tvix/eval/src/value/thunk.rs | 29 ++++++++++++++++++++++------
 3 files changed, 78 insertions(+), 6 deletions(-)

(limited to 'tvix/eval/src/value')

diff --git a/tvix/eval/src/value/list.rs b/tvix/eval/src/value/list.rs
index 3be5d41457..42d91b6b26 100644
--- a/tvix/eval/src/value/list.rs
+++ b/tvix/eval/src/value/list.rs
@@ -102,3 +102,13 @@ impl IntoIterator for NixList {
         self.0.into_iter()
     }
 }
+
+impl<'a> IntoIterator for &'a NixList {
+    type Item = &'a Value;
+
+    type IntoIter = std::slice::Iter<'a, Value>;
+
+    fn into_iter(self) -> Self::IntoIter {
+        self.0.iter()
+    }
+}
diff --git a/tvix/eval/src/value/mod.rs b/tvix/eval/src/value/mod.rs
index 78f4f5de67..1dc39d6832 100644
--- a/tvix/eval/src/value/mod.rs
+++ b/tvix/eval/src/value/mod.rs
@@ -27,6 +27,8 @@ pub use path::canon_path;
 pub use string::NixString;
 pub use thunk::Thunk;
 
+use self::thunk::ThunkSet;
+
 #[warn(variant_size_differences)]
 #[derive(Clone, Debug, PartialEq)]
 pub enum Value {
@@ -341,6 +343,49 @@ impl Value {
             _ => Ok(ForceResult::Immediate(self)),
         }
     }
+
+    /// Ensure `self` is *deeply* forced, including all recursive sub-values
+    pub(crate) fn deep_force(
+        &self,
+        vm: &mut VM,
+        thunk_set: &mut ThunkSet,
+    ) -> Result<(), ErrorKind> {
+        match self {
+            Value::Null
+            | Value::Bool(_)
+            | Value::Integer(_)
+            | Value::Float(_)
+            | Value::String(_)
+            | Value::Path(_)
+            | Value::Closure(_)
+            | Value::Builtin(_)
+            | Value::AttrNotFound
+            | Value::Blueprint(_)
+            | Value::DeferredUpvalue(_)
+            | Value::UnresolvedPath(_) => Ok(()),
+            Value::Attrs(a) => {
+                for (_, v) in a.iter() {
+                    v.deep_force(vm, thunk_set)?;
+                }
+                Ok(())
+            }
+            Value::List(l) => {
+                for val in l {
+                    val.deep_force(vm, thunk_set)?;
+                }
+                Ok(())
+            }
+            Value::Thunk(thunk) => {
+                if !thunk_set.insert(thunk) {
+                    return Ok(());
+                }
+
+                thunk.force(vm)?;
+                let value = thunk.value().clone();
+                value.deep_force(vm, thunk_set)
+            }
+        }
+    }
 }
 
 impl Display for Value {
diff --git a/tvix/eval/src/value/thunk.rs b/tvix/eval/src/value/thunk.rs
index 818ec0f58a..7bad7e8777 100644
--- a/tvix/eval/src/value/thunk.rs
+++ b/tvix/eval/src/value/thunk.rs
@@ -20,6 +20,7 @@
 
 use std::{
     cell::{Ref, RefCell, RefMut},
+    collections::HashSet,
     fmt::Display,
     rc::Rc,
 };
@@ -86,13 +87,12 @@ impl Thunk {
         })))
     }
 
-    /// Evaluate the content of a thunk, potentially repeatedly, until
-    /// a non-thunk value is returned.
+    /// Evaluate the content of a thunk, potentially repeatedly, until a
+    /// non-thunk value is returned.
     ///
-    /// This will change the existing thunk (and thus all references
-    /// to it, providing memoization) through interior mutability. In
-    /// case of nested thunks, the intermediate thunk representations
-    /// are replaced.
+    /// This will change the existing thunk (and thus all references to it,
+    /// providing memoization) through interior mutability. In case of nested
+    /// thunks, the intermediate thunk representations are replaced.
     pub fn force(&self, vm: &mut VM) -> Result<(), ErrorKind> {
         loop {
             let mut thunk_mut = self.0.borrow_mut();
@@ -200,3 +200,20 @@ impl Display for Thunk {
         }
     }
 }
+
+/// A wrapper type for tracking which thunks have already been seen in a
+/// context. This is necessary for cycle detection.
+///
+/// The inner `HashSet` is not available on the outside, as it would be
+/// potentially unsafe to interact with the pointers in the set.
+#[derive(Default)]
+pub struct ThunkSet(HashSet<*mut ThunkRepr>);
+
+impl ThunkSet {
+    /// Check whether the given thunk has already been seen. Will mark the thunk
+    /// as seen otherwise.
+    pub fn insert(&mut self, thunk: &Thunk) -> bool {
+        let ptr: *mut ThunkRepr = thunk.0.as_ptr();
+        self.0.insert(ptr)
+    }
+}
-- 
cgit 1.4.1