implement string context

fix-abstract-vm/src/host.rs

@@ -4,8 +4,7 @@ use fix_error::Source;
 use hashbrown::HashSet;
 
 use crate::{
-    AttrSet, Closure, ExtraScope, List, NixString, NixType, Null, Path, PrimOp, PrimOpApp,
-    StaticValue, StrictValue, Thunk, ThunkState, Value,
+    AttrSet, Closure, ExtraScope, List, NixString, NixType, Null, Path, PrimOp, PrimOpApp, StaticValue, StrictValue, StringContext, Thunk, ThunkState, Value
 };
 
 pub trait VmContext {
@@ -36,6 +35,9 @@ pub trait VmRuntimeCtxExt: VmRuntimeCtx {
         &'a mut self,
         val: StrictValue<'gc>,
     ) -> std::result::Result<StringId, NixType>;
+    /// Returns the string context attached to `val`, or `&[]` if `val` is
+    /// either a non-string or a string without context.
+    fn get_string_context<'gc>(&self, val: StrictValue<'gc>) -> &'gc StringContext;
     fn convert_value(&self, val: Value) -> fix_common::Value;
 }
 
@@ -73,6 +75,14 @@ impl<T: VmRuntimeCtx> VmRuntimeCtxExt for T {
         }
     }
 
+    fn get_string_context<'gc>(&self, val: StrictValue<'gc>) -> &'gc StringContext {
+        if let Some(ns) = val.as_gc::<NixString>() {
+            ns.as_ref().context()
+        } else {
+            StringContext::empty()
+        }
+    }
+
     fn convert_value(&self, val: Value) -> fix_common::Value {
         self.convert_value_with_seen(val, &mut HashSet::new())
     }

fix-abstract-vm/src/lib.rs

@@ -6,6 +6,7 @@ mod machine;
 mod path_util;
 mod resolve;
 mod state;
+mod string_context;
 mod value;
 
 pub use bytecode_reader::*;
@@ -15,4 +16,5 @@ pub use machine::*;
 pub use path_util::*;
 pub use resolve::*;
 pub use state::*;
+pub use string_context::*;
 pub use value::*;

fix-abstract-vm/src/string_context.rs

@@ -0,0 +1,161 @@
+use std::cmp::Ordering;
+
+use smallvec::SmallVec;
+
+/// A string context element
+#[derive(Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub enum StringContextElem {
+    // Plain store path reference
+    Opaque {
+        path: Box<str>,
+    },
+    // All outputs of a derivation
+    // encoded `=<drvPath>`
+    DrvDeep {
+        drv_path: Box<str>,
+    },
+    // A specific output of a derivation
+    // encoded `!<output>!<drvPath>`
+    Built {
+        drv_path: Box<str>,
+        output: Box<str>,
+    },
+}
+
+impl StringContextElem {
+    /// Decode the CppNix wire form (`!out!/p`, `=/p`, `/p`). Falls back to
+    /// `Opaque` for malformed `!`-prefixed inputs (matching nix-js).
+    pub fn decode(encoded: &str) -> Self {
+        if let Some(drv_path) = encoded.strip_prefix('=') {
+            Self::DrvDeep {
+                drv_path: drv_path.into(),
+            }
+        } else if let Some(rest) = encoded.strip_prefix('!') {
+            if let Some(second_bang) = rest.find('!') {
+                Self::Built {
+                    output: rest[..second_bang].into(),
+                    drv_path: rest[second_bang + 1..].into(),
+                }
+            } else {
+                Self::Opaque {
+                    path: encoded.into(),
+                }
+            }
+        } else {
+            Self::Opaque {
+                path: encoded.into(),
+            }
+        }
+    }
+
+    pub fn encode(&self) -> String {
+        match self {
+            Self::Opaque { path } => path.to_string(),
+            Self::DrvDeep { drv_path } => format!("={drv_path}"),
+            Self::Built { drv_path, output } => format!("!{output}!{drv_path}"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct StringContext {
+    data: SmallVec<[StringContextElem; 1]>,
+}
+
+impl IntoIterator for StringContext {
+    type Item = StringContextElem;
+    type IntoIter = <SmallVec<[StringContextElem; 1]> as IntoIterator>::IntoIter;
+    fn into_iter(self) -> Self::IntoIter {
+        self.data.into_iter()
+    }
+}
+
+impl<'a> IntoIterator for &'a StringContext {
+    type Item = &'a StringContextElem;
+    type IntoIter = <&'a SmallVec<[StringContextElem; 1]> as IntoIterator>::IntoIter;
+    fn into_iter(self) -> Self::IntoIter {
+        self.data.iter()
+    }
+}
+
+impl<'a> IntoIterator for &'a mut StringContext {
+    type Item = &'a mut StringContextElem;
+    type IntoIter = <&'a mut SmallVec<[StringContextElem; 1]> as IntoIterator>::IntoIter;
+    fn into_iter(self) -> Self::IntoIter {
+        self.data.iter_mut()
+    }
+}
+
+impl FromIterator<StringContextElem> for StringContext {
+    fn from_iter<T: IntoIterator<Item = StringContextElem>>(iter: T) -> Self {
+        Self {
+            data: iter.into_iter().collect()
+        }
+    }
+}
+
+impl StringContext {
+    pub fn empty() -> &'static Self {
+        static EMPTY: StringContext = StringContext {
+            data: SmallVec::new_const(),
+        };
+        &EMPTY
+    }
+
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.data.is_empty()
+    }
+
+    pub fn insert(&mut self, elem: StringContextElem) {
+        match self.data.binary_search(&elem) {
+            Ok(_) => {}
+            Err(pos) => self.data.insert(pos, elem),
+        }
+    }
+
+    pub fn merge(&self, other: &Self) -> Self {
+        if self.data.is_empty() {
+            return other.clone();
+        }
+        if other.data.is_empty() {
+            return self.clone();
+        }
+
+        let a = &self.data;
+        let b = &other.data;
+        let mut out = SmallVec::with_capacity(a.len() + b.len());
+        let (mut i, mut j) = (0, 0);
+        while i < a.len() && j < b.len() {
+            match a[i].cmp(&b[j]) {
+                Ordering::Less => {
+                    out.push(a[i].clone());
+                    i += 1;
+                }
+                Ordering::Greater => {
+                    out.push(b[j].clone());
+                    j += 1;
+                }
+                Ordering::Equal => {
+                    out.push(a[i].clone());
+                    i += 1;
+                    j += 1;
+                }
+            }
+        }
+        out.extend(a[i..].iter().cloned());
+        out.extend(b[j..].iter().cloned());
+        Self { data: out }
+    }
+
+    pub fn iter(&self) -> <&Self as IntoIterator>::IntoIter {
+        self.into_iter()
+    }
+
+    pub fn iter_mut(&mut self) -> <&mut Self as IntoIterator>::IntoIter {
+        self.into_iter()
+    }
+}

fix-abstract-vm/src/value.rs

@@ -17,6 +17,7 @@ use string_interner::Symbol;
 use string_interner::symbol::SymbolU32;
 
 use crate::boxing::{RawBox, RawStore, RawTag, Value as RawValue};
+use crate::string_context::StringContext;
 
 mod private {
     pub trait Cealed {}
@@ -420,26 +421,41 @@ impl RawStore for Path {
     }
 }
 
-/// Heap-allocated Nix string.
-///
-/// Stored on the GC heap via `Gc<'gc, NixString>`. The string data itself
-/// lives in a standard `Box<str>` owned by this struct; the GC only manages
-/// the outer allocation.
 #[derive(Collect)]
 #[collect(require_static)]
 pub struct NixString {
     data: Box<str>,
-    // TODO: string context for derivation dependency tracking
+    context: StringContext,
 }
 
 impl NixString {
     pub fn new(s: impl Into<Box<str>>) -> Self {
-        Self { data: s.into() }
+        Self {
+            data: s.into(),
+            context: StringContext::new(),
+        }
+    }
+
+    /// Construct a `NixString` whose `context` is already sorted+deduped.
+    /// The caller is responsible for invariant maintenance.
+    pub fn with_context(s: impl Into<Box<str>>, context: StringContext) -> Self {
+        Self {
+            data: s.into(),
+            context,
+        }
     }
 
     pub fn as_str(&self) -> &str {
         &self.data
     }
+
+    pub fn context(&self) -> &StringContext {
+        &self.context
+    }
+
+    pub fn has_context(&self) -> bool {
+        !self.context.is_empty()
+    }
 }
 
 impl fmt::Debug for NixString {