chore: Update hugr dependency #51

aborgna-q · 2023-08-01T09:48:02Z

No description provided.

acl-cqc

Well who knows maybe we'll need another hugr release to get there....but I hope this helps.
(First TKET2 PR I've looked at so I'm not sure I've earnt "Approve" button rights yet ;-) )

acl-cqc · 2023-08-02T14:25:08Z

src/json/decoder.rs

-        let sig =
-            Signature::new_linear([vec![QB; num_qubits], vec![LINEAR_BIT; num_bits]].concat());
+        let sig = AbstractSignature::new_linear(
+            [vec![QB; num_qubits], vec![LINEAR_BIT.clone(); num_bits]].concat(),


A bit puzzled as to why you now need to clone() LINEAR_BIT

LINEAR_BIT is now initialized with lazy_static instead of defined as a const. That requires some boilerplate uses like this.

Note that vec! was doing some cloning anyways.

acl-cqc · 2023-08-02T14:26:36Z

src/json/decoder.rs

+            Some(c) => {
+                let const_type = hugr::types::ClassicType::F64;
+                assert!(c.check_type(&const_type).is_ok());
+                let const_op = Const::new(c, const_type).unwrap();


This makes the check in the previous line for you (that's the failure that the unwrap() is about), so you might simplify this

acl-cqc · 2023-08-02T14:31:00Z

src/resource.rs

@@ -25,20 +25,35 @@ pub const LINEAR_BIT_NAME: SmolStr = SmolStr::new_inline("LBit");
 pub const JSON_OP_NAME: SmolStr = SmolStr::new_inline("TKET1 Json Op");

 lazy_static! {
+
+    /// The type for linear bits. Part of the TKET1 resource.
+    pub static ref LINEAR_BIT: SimpleType = SimpleType::Qontainer(Container::Opaque(CustomType::new(


You can now (following CustomType's are Simple not necessarily Classic, and report their TypeTag hugr#328) "just" do CustomType::new(...).into() here

I guess there is no chance of changing that static ref into const and then avoiding all the .clone(). Hmmm. CustomType::new really doesn't do anything much at all, so I'd be happy to see a change on the Hugr side to enable such....?

If (2.) doesn't work, then the other way would be to do TKET1_RESOURCE.types().get(LINEAR_BIT_NAME).unwrap().instantiate_concrete(&[]) which would reduce the chance of getting anything inconsistent and is more "best practice"

Yeah, before we had a const CustomType::new_simple that didn't take arguments, but it was removed in Interactions between resource definitions and instances hugr#315

acl-cqc · 2023-08-02T14:39:00Z

src/resource.rs


+        let json_op_param = TypeParam::Value(HashableType::Container(Container::Opaque(
+            CustomType::new("TKET1 Json Op", vec![], TKET1_RESOURCE_ID, TypeTag::Simple),


Really this type should be defined in the resource too, just like LINEAR_BIT

Linear bit is a TypeDef (with a corresponding SimpleType.

This is a build-time operation TypeParam (with a corresponding TypeArg). How can I add it to a resource?

acl-cqc · 2023-08-02T14:42:34Z

src/resource.rs

@@ -56,23 +71,21 @@ pub(crate) fn wrap_json_op(op: &JsonOp) -> ExternalOp {
        TKET1_RESOURCE_ID,
        JSON_OP_NAME,
        "".into(),
-        vec![TypeArg::Value(op)],
-        Some(sig),
+        vec![TypeArg::CustomValue(op)],


The OpaqueOp containing this should now be produced by TKET1_RESOURCE.operations().get(JSON_OP_NAME).unwrap().instantiate_concrete(vec![TypeArg::CustomValue(op)]) and I think that should even calculate the signature for you (assuming json_op_signature does what I think it should)

acl-cqc · 2023-08-02T14:43:52Z

src/resource.rs

+fn json_op_signature(
+    args: &[TypeArg],
+) -> Result<(TypeRow<SimpleType>, TypeRow<SimpleType>, ResourceSet), SignatureError> {
+    let [TypeArg::CustomValue(arg)] = args else {


Yes, panicking here seems fine, the correctness of the &[TypeArg] against the declared TypeParam's should already have been checked

acl-cqc · 2023-08-02T14:45:43Z

src/resource.rs

        panic!("Wrong number of arguments");
        // TODO: Add more Signature Errors
        //return Err(SignatureError::WrongNumArgs(1, args.len()));
    };
    let op: JsonOp = serde_yaml::from_value(arg.clone()).unwrap(); // TODO Errors!
-    let sig = op.signature();
-    Ok((sig.input, sig.output, sig.output_resources))
+    let Signature {


Hmmmm, this should probably become an AbstractSignature and then you can use the resource_reqs there, input_resources is not really right here

Changed this along with JsonOp::signature

acl-cqc · 2023-08-02T14:48:22Z

src/json/op.rs

+        let signature = AbstractSignature::new_df([linear.clone(), params].concat(), linear);
+        Signature {
+            signature,
+            input_resources: ResourceSet::singleton(&TKET1_RESOURCE_ID),


This may become restricting. Basically you are saying that the inputs to whatever operation this is, will have needed TKET1_RESOURCE_ID to compute them. Hmmm. This is in impl JsonOp right? I think you should return an AbstractSignature instead, with resource_reqs being this singleton-set. Does that work?

I added the resource to the output instead using AbstractSignature::with_resource_delta.
Now the tests throw an error:

InvalidHUGR( SrcExceedsTgtResources { from: Node { index: NodeIndex(3) }, from_offset: Port { offset: Outgoing(0) }, from_resources: ResourceSet({"TKET1"}), to: Node { index: NodeIndex(2) }, to_offset: Port { offset: Incoming(0) }, to_resources: ResourceSet({}) } )

(rendered before connecting the edges to the output node)

I'll try to debug it.

acl-cqc · 2023-08-02T14:50:45Z

pyrs/Cargo.toml

Hmmmm, is this a prerequisite, a driveby, a follow-up? (All the changes I see are hugr-related AFAICT.)

Ah, it would be a small drive-by.
This makes sure the dependency here is the same as in the main Cargo.toml.

…ion)

aborgna-q · 2023-08-24T13:58:01Z

Subsumed by #53

Update hugr dep

ba92b1d

aborgna-q requested a review from acl-cqc August 1, 2023 09:48

More hugr updates

cc50402

acl-cqc reviewed Aug 2, 2023

View reviewed changes

aborgna-q added 6 commits August 2, 2023 17:05

Address comments (WIP failing test)

e547394

More dep updates: rename hugr region views

ea0ae20

Updated DfBuilder signatures, drop int constants

17c3a95

Simplified type params

5acb135

Add resource delta to DataflowNodes (Errors out on TypeArg instantiat…

9f1df22

…ion)

is still broken

7acaba4

aborgna-q marked this pull request as draft August 8, 2023 10:38

aborgna-q closed this Aug 24, 2023

ss2165 deleted the chore/update-hugr branch August 31, 2023 16:51

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

chore: Update hugr dependency #51

chore: Update hugr dependency #51

aborgna-q commented Aug 1, 2023

acl-cqc left a comment

acl-cqc Aug 2, 2023

aborgna-q Aug 2, 2023

acl-cqc Aug 2, 2023

acl-cqc Aug 2, 2023

aborgna-q Aug 2, 2023

acl-cqc Aug 2, 2023

aborgna-q Aug 2, 2023

acl-cqc Aug 2, 2023

acl-cqc Aug 2, 2023

acl-cqc Aug 2, 2023

aborgna-q Aug 2, 2023

acl-cqc Aug 2, 2023

aborgna-q Aug 2, 2023 •

edited

Loading

acl-cqc Aug 2, 2023

aborgna-q Aug 2, 2023

aborgna-q commented Aug 24, 2023


		let json_op_param = TypeParam::Value(HashableType::Container(Container::Opaque(
		CustomType::new("TKET1 Json Op", vec![], TKET1_RESOURCE_ID, TypeTag::Simple),

chore: Update hugr dependency #51

chore: Update hugr dependency #51

Conversation

aborgna-q commented Aug 1, 2023

acl-cqc left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aborgna-q Aug 2, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aborgna-q commented Aug 24, 2023

aborgna-q Aug 2, 2023 •

edited

Loading