feat: Add ResourceScope #1052

lmondada · 2025-08-19T14:32:19Z

Alright, couldn't resist it. Here's a PR that moves a lot of the logic that I tried to put into portgraph into tket. I think it makes much more sense this way.

I'm interested to see what you think! (a lot of the diff is docs and snapshot tests, so don't worry :P)

codecov · 2025-08-19T14:39:31Z

Codecov Report

❌ Patch coverage is 85.71429% with 57 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.45%. Comparing base (51bc0ec) to head (b07590e).

Files with missing lines	Patch %	Lines
tket/src/resource/types.rs	73.86%	23 Missing ⚠️
tket/src/resource/scope.rs	91.47%	13 Missing and 6 partials ⚠️
tket/src/resource/flow.rs	81.25%	6 Missing and 3 partials ⚠️
tket/src/resource.rs	85.00%	3 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1052      +/-   ##
==========================================
+ Coverage   78.19%   78.45%   +0.26%     
==========================================
  Files         102      106       +4     
  Lines       12783    13175     +392     
  Branches    12504    12896     +392     
==========================================
+ Hits         9996    10337     +341     
- Misses       2130     2175      +45     
- Partials      657      663       +6

Flag	Coverage Δ
python	`81.00% <ø> (ø)`

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

lmondada · 2025-08-19T18:43:34Z

Three things on my mind

could we make this share more code/structs with the CircuitBuilder, esp. CircuitUnit?
idem, with the pytket encoder/decoder
imo this is a candidate to replace the Circuit struct if we get rid of it 😅

acl-cqc

In general I quite like this - the approach seems entirely reasonable within TKET, and perhaps a bit odd/overly-specialized in portgraph :). Haven't done a detailed code review but a few thoughts mostly on docs here.

My main question is - wtf CopyableValueID? Do you really want/use it, or did you just think you had to put something in for nonlinear ports? If the latter, would this PR work/make sense if you stripped it out?

acl-cqc · 2025-08-22T14:43:37Z

tket/src/resource.rs

+//!   the circuit. Each resource has a unique [`ResourceId`] and operations on
+//!   the same resource are ordered by a [`Position`].
+//! - **Copyable values**: Regular values that can be copied and discarded
+//!   freely. Each is identified by a unique [`CopyableValueId`].


I guess I will find out, but might be good to say that these are different/not-resource-tracked/something here.

acl-cqc · 2025-08-22T14:47:07Z

tket/src/resource/flow.rs

+pub trait ResourceFlow {
+    /// Map resource IDs from operation inputs to outputs.
+    ///
+    /// Takes an operation type and the resource IDs of the operation's inputs.


Suggested change

/// Takes an operation type and the resource IDs of the operation's inputs.

/// Takes an [`OpType`] and a resource IDs for each linear input.

/// `None` is passed in place of any nonlinear inputs.

/// Returns....

acl-cqc · 2025-08-22T14:48:00Z

tket/src/resource/flow.rs

+    /// The i-th entry is Some(resource_id) if the i-th port is a linear type,
+    /// None otherwise. Returns the resource IDs of the operation's outputs
+    /// in port order. Output resource IDs should be one of the input resource
+    /// IDs for resource-preserving operations, or None for new resources or


I'm inferring here that every None is a new resource-id exactly if (iff) it's linear

acl-cqc · 2025-08-22T14:55:27Z

tket/src/resource/flow.rs

+
+/// Default implementation of ResourceFlow.
+///
+/// This implementation considers that an operation is resource-preserving if


Maybe be a bit more explicit:

This implementation defines an operation as "resource-preserving" if: * for all port indices i, * if there is an input port of linear type, *or* an output port of linear type, then both input and output ports must exist and have that same linear type For resource-preserving operations, linear inputs are then mapped to the corresponding output. (All outputs with no corresponding input, must be classical.) For other operations, all input resources are discarded and all outputs will be given fresh resource IDs.

Can get that if-then out of the definition of resource-preserving by:

* for each port indices i, either * input i or output i does not exist * neither input i nor output i is linear * both are linear and have the same type

a bit longer but I think I like the latter more?

acl-cqc · 2025-08-22T14:57:28Z

tket/src/resource/interval.rs

@@ -0,0 +1,294 @@
+//! Subgraph representation using intervals on resource paths.


Does this relate to intervals of lines in LineConvexChecker? Not for this PR, but I have been wondering if LineConvexChecker belongs in tket rather than hugr, too

acl-cqc · 2025-08-22T15:05:33Z

tket/src/resource.rs

+//!
+//! - **Linear resources**: Non-copyable values that form resource paths through
+//!   the circuit. Each resource has a unique [`ResourceId`] and operations on
+//!   the same resource are ordered by a [`Position`].


Say something like ResourceFlow allows to determine how resources are passed through, discarded or created by an op (for linear ports)

acl-cqc · 2025-08-22T15:06:39Z

tket/src/resource/scope.rs

@@ -0,0 +1,596 @@
+//! Main ResourceScope implementation for tracking resources in HUGR subgraphs.


nit: Can you drop the first line and just run with "Provides the [ResourceScope] struct which...." or does clippy/etc. complain?

acl-cqc · 2025-08-22T15:09:38Z

tket/src/resource/scope.rs

+    ) -> impl Iterator<Item = H::Node> + '_ {
+        let mut curr_node = start_node;
+
+        iter::from_fn(move || {


nit: does iter::successors let you avoid the mut curr_node?

acl-cqc · 2025-08-22T15:10:34Z

tket/src/resource/scope.rs

+        // order.
+        for node in toposort_subgraph(&self.hugr, &self.subgraph, self.find_sources()) {
+            if self.hugr.get_optype(node).dataflow_signature().is_none() {
+                // ignore non-dataflow ops


Whilst I think this is fine, I suspect non-dataflow nodes - like "dataflow" ops with no inputs or outputs - would just end any flows (and not start any new flows!) ?

acl-cqc · 2025-08-22T15:11:15Z

tket/src/resource/scope.rs

+    }
+
+    /// All copyable values on the ports of `node` in the given direction.
+    pub fn get_copyable_values(


(I note this isn't used....)

acl-cqc · 2025-08-22T15:17:13Z

I mean, if this went into Portgraph, it'd have to do so without any mention of linearity, so probably leaving out any (useful!) implementation of the ResourceFlow trait entirely; that would then have to be done in tket...

lmondada requested a review from aborgna-q August 19, 2025 14:32

lmondada force-pushed the lm/resource-scope branch from e58de1d to 57a2f0a Compare August 19, 2025 14:34

feat: Add ResourceScope

6b94a7d

lmondada force-pushed the lm/resource-scope branch from 57a2f0a to 6b94a7d Compare August 19, 2025 14:36

lmondada marked this pull request as ready for review August 19, 2025 14:36

lmondada requested a review from a team as a code owner August 19, 2025 14:36

This was referenced Aug 20, 2025

feat!: Replace Subcircuit // take 2 #1054

Open

Remove Circuit struct #1060

Open

lmondada added 2 commits August 22, 2025 13:55

Merge branch 'main' into lm/resource-scope

3c86b95

Make position a node property

b07590e

lmondada requested review from acl-cqc and tatiana-s and removed request for aborgna-q August 22, 2025 12:14

acl-cqc reviewed Aug 22, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: Add ResourceScope #1052

feat: Add ResourceScope #1052

lmondada commented Aug 19, 2025 •

edited

Loading

Uh oh!

codecov bot commented Aug 19, 2025 •

edited

Loading

Uh oh!

lmondada commented Aug 19, 2025

Uh oh!

acl-cqc left a comment

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc Aug 22, 2025

Uh oh!

acl-cqc commented Aug 22, 2025

Uh oh!

Uh oh!

-    /// Takes an operation type and the resource IDs of the operation's inputs.
+    /// Takes an [`OpType`] and a resource IDs for each linear input.
+    /// `None` is passed in place of any nonlinear inputs.
+    /// Returns....

		@@ -0,0 +1,294 @@
		//! Subgraph representation using intervals on resource paths.

		@@ -0,0 +1,596 @@
		//! Main ResourceScope implementation for tracking resources in HUGR subgraphs.

feat: Add ResourceScope #1052

Are you sure you want to change the base?

feat: Add ResourceScope #1052

Conversation

lmondada commented Aug 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Aug 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

lmondada commented Aug 19, 2025

Uh oh!

acl-cqc left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

acl-cqc commented Aug 22, 2025

Uh oh!

Uh oh!

lmondada commented Aug 19, 2025 •

edited

Loading

codecov bot commented Aug 19, 2025 •

edited

Loading