Remove single source requirement for DSGs

[ss2165]

from @cqc-alec :

The spec currently states that "In a data dependency subgraph, a valid ordering of operations can be achieved by topologically sorting the nodes starting from Input with respect to the Value and Order edges." I vote we drop this requirement, since it places unnecessary constraints on execution order (there's no reason why a LoadConstant operation shouldn't be executed before knowing the inputs to a DFG). If efficiency of topsort is a big concern we can find other ways to optimize it.

[cqc-alec]

I believe this was fixed in the spec by #762 , but the implementation still needs updating (to not add the unnecessary Order edges).

[acl-cqc]

Do we actually have any implementation code adding such Order edges? I check e.g.

hugr/src/builder/build_traits.rs

Lines 73 to 77 in 968c8b0

	fn add_constant(&mut self, constant: impl Into<ops::Const>) -> Result<ConstID, BuildError> {
	let const_n = self.add_child_node(NodeType::new(constant.into(), ExtensionSet::new()))?;
	Ok(const_n.into())
	}

and

hugr/src/builder/build_traits.rs

Lines 352 to 370 in 968c8b0

	fn load_const(&mut self, cid: &ConstID) -> Result<Wire, BuildError> {
	let const_node = cid.node();
	let nodetype = self.hugr().get_nodetype(const_node);
	let op: ops::Const = nodetype
	.op()
	.clone()
	.try_into()
	.expect("ConstID does not refer to Const op.");
	let load_n = self.add_dataflow_op(
	ops::LoadConstant {
	datatype: op.const_type().clone(),
	},
	// Constant wire from the constant value node
	vec![Wire::new(const_node, OutgoingPort::from(0))],
	)?;
	Ok(load_n.out_wire(0))
	}

in the builder, and don't see anything; and we haven't implemented InsertConst(Ignore) rewrites...

[cqc-alec]

I may be wrong about that, but my assumption was based on comments @peter-campora made a few weeks ago where he was surprised by the presence of some Order edges.

[cqc-alec]

I can't find anywhere in the code where we do this either.

[ss2165]

Builder used to do this but was removed in this commit:

c7bfffe#diff-7b05863e5e5858f70dcaba4fa126e7782aa9b8da6fd64b80715d821b9867bef0L634-L641

[cqc-alec]

OK! So can we close this issue?

[ss2165]

Looks like it, the validation code does not depend on it either

hugr/src/hugr/validate.rs

Lines 376 to 377 in 0c910cf

	let postorder = Topo::new(&region.as_petgraph());
	let nodes_visited = postorder