refactor(#2772): graph walks belong to jobq, not to its caller

hive-c0re hand-rolled four traversals over a graph it doesn't own, because
`Graph` exposed only `node()` and `nodes()`. They are generic — nothing in
them knows what a hyperhive DAG is — so they move to `hive-jobq` and core
delegates.

`Graph` gains `root_of`, `descendants`, `roots`, `is_settled` and
`first_error`; they reuse the private `is_descendant` the crate already had
for its dep-scope rule. `subtree` gets faster on the way: core walked every
node's whole parent chain to the root for every node in the graph, where
`is_descendant` stops as soon as it sees the ancestor.

`first_error` deliberately looks for the first `Failed` descendant that
*carries* an error rather than the first `Failed` one. A node that rolled
its failure up from a child holds no error of its own and sorts before that
child, so the simpler version reports `None` for the common case and the
dashboard loses the reason. The distinction has its own test.

`dag_is_terminal` is deleted rather than moved: it was already a plain
`state.is_terminal()` read, and its three call sites now ask the graph.
This commit is contained in:
atlas 2026-07-27 19:59:54 +02:00 committed by mara
commit 4de5a8dd7c
2 changed files with 130 additions and 35 deletions

View file

@ -480,7 +480,7 @@ impl JobQueue {
inner
.containers()
.into_iter()
.filter(|&c| !inner.dag_is_terminal(c))
.filter(|&c| !inner.sched.graph().is_settled(c))
.count()
}
}
@ -505,25 +505,17 @@ impl QueueInner {
})
}
/// The DAG container a node belongs to — walk its parent chain to the root
/// (`parent == None`), which is the container. Returns `id` itself for a
/// container node.
/// The DAG container a node belongs to. A container is exactly a group root,
/// so this is the graph's own parent-chain walk.
fn dag_of(&self, id: NodeId) -> Option<NodeId> {
let mut cur = id;
loop {
match self.sched.graph().node(cur)?.parent {
Some(p) => cur = p,
None => return Some(cur),
}
}
self.sched.graph().root_of(id)
}
/// The DAG's work nodes — its `container`'s subtree, excluding the container.
fn subtree(&self, container: NodeId) -> Vec<NodeId> {
self.sched
.graph()
.nodes()
.filter(|n| n.id != container && self.dag_of(n.id) == Some(container))
.descendants(container)
.map(|n| n.id)
.collect()
}
@ -553,15 +545,6 @@ impl QueueInner {
})
}
/// True when the DAG has settled — its container has rolled up terminal
/// (equivalent to every work node being terminal).
fn dag_is_terminal(&self, container: NodeId) -> bool {
self.sched
.graph()
.node(container)
.is_some_and(|n| n.state.is_terminal())
}
/// Whether `id` has an edge that accepts a **dropped** dependency — i.e. the
/// node exists to report on work that may never run. Used by
/// [`JobQueue::cancel`] to decide what to spare, so the decision comes from
@ -574,18 +557,12 @@ impl QueueInner {
})
}
/// First failed work node's error (read off the graph `Node`), for the
/// dashboard's DAG-level error line.
/// First failed work node's error, for the dashboard's DAG-level error line.
fn dag_first_error(&self, container: NodeId) -> Option<String> {
for id in self.subtree(container) {
if let Some(n) = self.sched.graph().node(id)
&& n.state == JobState::Failed
&& let Some(e) = n.error.clone()
{
return Some(e);
}
}
None
self.sched
.graph()
.first_error(container)
.map(ToOwned::to_owned)
}
/// Project a DAG into its wire [`DagView`]: a near-raw view of the
@ -675,7 +652,7 @@ impl QueueInner {
if !any_unsettled {
return None;
}
let is_terminal = self.dag_is_terminal(container);
let is_terminal = self.sched.graph().is_settled(container);
Some(DagView {
id: container.get(),
source: meta.source,
@ -718,7 +695,7 @@ impl QueueInner {
let mut live: Vec<NodeId> = Vec::new();
let mut terminal: Vec<(NodeId, i64)> = Vec::new();
for c in self.containers() {
if self.dag_is_terminal(c) {
if self.sched.graph().is_settled(c) {
terminal.push((c, self.dag_finished_at(c)));
} else {
live.push(c);

View file

@ -462,6 +462,56 @@ impl<N, R> Graph<N, R> {
self.nodes.iter()
}
/// Top of `id`'s [`Node::parent`] chain — the group root whose subtree `id`
/// lives in. Returns `id` itself when `id` is already a root, and `None`
/// only when `id` isn't in the graph.
#[must_use]
pub fn root_of(&self, id: NodeId) -> Option<NodeId> {
let mut cur = id;
loop {
match self.node(cur)?.parent {
Some(p) => cur = p,
None => return Some(cur),
}
}
}
/// Every node in `id`'s subtree, excluding `id` itself, in insertion order.
pub fn descendants(&self, id: NodeId) -> impl Iterator<Item = &Node<N, R>> {
self.nodes
.iter()
.filter(move |n| self.is_descendant(n.id, id))
}
/// Every group root — the nodes with no parent.
pub fn roots(&self) -> impl Iterator<Item = &Node<N, R>> {
self.nodes.iter().filter(|n| n.parent.is_none())
}
/// Whether `id` has settled. A group root's state is its subtree's roll-up,
/// so for a root this answers "is everything under it finished" — which is
/// why callers don't scan the subtree themselves. Unknown ids are not
/// settled.
#[must_use]
pub fn is_settled(&self, id: NodeId) -> bool {
self.node(id).is_some_and(|n| n.state.is_terminal())
}
/// Why `id`'s subtree failed: the error of the first `Failed` descendant
/// that carries one, in insertion order.
///
/// Skipping the ones without an error is the point, not an optimisation. A
/// node that rolled up `Failed` from a child holds no error of its own, and
/// such a node can sort before the child that actually broke — stopping at
/// the first `Failed` node would report `None` while the real reason sits
/// further down the subtree.
#[must_use]
pub fn first_error(&self, id: NodeId) -> Option<&str> {
self.descendants(id)
.filter(|n| matches!(n.state, State::Failed))
.find_map(|n| n.error.as_deref())
}
/// Whether `ancestor` lies on `node`'s [`Node::parent`] chain (i.e. `node` is
/// in `ancestor`'s subtree). `node` is not its own ancestor.
fn is_descendant(&self, node: NodeId, ancestor: NodeId) -> bool {
@ -594,6 +644,74 @@ mod tests {
));
}
/// Borrow a node mutably so a test can force its lifecycle state. Looks the
/// node up by id rather than indexing, so it doesn't quietly depend on ids
/// and positions coinciding.
fn node_mut<'g>(
g: &'g mut Graph<&'static str, String>,
id: NodeId,
) -> &'g mut Node<&'static str, String> {
g.nodes
.iter_mut()
.find(|n| n.id == id)
.expect("node in graph")
}
/// `root_of` walks to the top of the parent chain; `descendants` is its
/// inverse and excludes the node itself.
#[test]
fn root_of_and_descendants_span_the_subtree() {
let mut g: Graph<&str, String> = Graph::new();
let root = g.insert("root", vec![], None).unwrap();
let mid = g.insert("mid", vec![], Some(root)).unwrap();
let leaf = g.insert("leaf", vec![], Some(mid)).unwrap();
let other = g.insert("other-root", vec![], None).unwrap();
assert_eq!(g.root_of(leaf), Some(root), "walks the whole chain");
assert_eq!(g.root_of(root), Some(root), "a root is its own root");
assert_eq!(g.root_of(NodeId(99)), None, "unknown id");
let mut under_root: Vec<NodeId> = g.descendants(root).map(|n| n.id).collect();
under_root.sort_unstable();
assert_eq!(under_root, vec![mid, leaf], "excludes the node itself");
assert_eq!(g.descendants(other).count(), 0);
let roots: Vec<NodeId> = g.roots().map(|n| n.id).collect();
assert_eq!(roots, vec![root, other]);
}
/// The reason `first_error` looks for the first failed descendant **that
/// carries an error** rather than simply the first failed one: a node that
/// rolled its `Failed` up from a child holds no error of its own, and it
/// sorts *before* that child. Stopping at the first `Failed` node would
/// report `None` and lose the real reason.
#[test]
fn first_error_skips_a_rolled_up_failure_carrying_no_error() {
let mut g: Graph<&str, String> = Graph::new();
let container = g.insert("dag", vec![], None).unwrap();
let rolled_up = g.insert("prebuild", vec![], Some(container)).unwrap();
let broke = g.insert("swap", vec![], Some(rolled_up)).unwrap();
node_mut(&mut g, rolled_up).state = State::Failed;
let broken = node_mut(&mut g, broke);
broken.state = State::Failed;
broken.error = Some("nix build exploded".to_owned());
assert_eq!(g.first_error(container), Some("nix build exploded"));
}
#[test]
fn is_settled_tracks_node_state() {
let mut g: Graph<&str, String> = Graph::new();
let n = g.insert("n", vec![], None).unwrap();
assert!(!g.is_settled(n), "Pending is not settled");
node_mut(&mut g, n).state = State::Finishing;
assert!(!g.is_settled(n), "Finishing still has children running");
node_mut(&mut g, n).state = State::Done;
assert!(g.is_settled(n));
assert!(!g.is_settled(NodeId(99)), "unknown id is not settled");
}
#[test]
fn state_terminality() {
assert!(State::Done.is_terminal());