[BUG] Identical blocks might break dependecy_graph

[gwwatkin]

Issue Description

This line suggests it:

lattice-surgery-compiler/src/lsqecc/logical_lattice_ops/dependency_graph.py

Line 134 in faad034

graph[curr.op] = list()

If two blocks are identical they will be interpreted as the same key in this map which doesn't sounds like the desired behavior.

Proposed Solution

Need to investigate if this is the case and find a solution. Also the code in dependency_graph.py is very hard to read because it's missing type annotations. Especially because it's generic/templated code it should be annotated.

Additional References

The equivalence of rotations

lattice-surgery-compiler/src/lsqecc/pauli_rotations/rotation.py

Lines 172 to 177 in faad034

	def __eq__(self, other) -> bool:
	return (
	isinstance(other, PauliRotation)
	and self.rotation_amount == other.rotation_amount
	and self.ops_list == other.ops_list
	)

and measurements

lattice-surgery-compiler/src/lsqecc/pauli_rotations/rotation.py

Lines 211 to 216 in faad034

	def __eq__(self, other) -> bool:
	return (
	isinstance(other, Measurement)
	and self.isNegative == other.isNegative
	and self.ops_list == other.ops_list
	)

[alexnguyenn]

A quick solution I can think of is to remove that method since DependencyGraph.Node has parents and children list, which is enough for us to figure out the adjacency list anyway.

[gwwatkin]

It should, as we are only ever using it in a debug file to view the graph. I think working on this class is lower priority than anything in the UF epic, so I think it's fine to leave this for now, but good to have the issue logged if we do start to use it.

[gwwatkin]

If we go back to the plan of parallelizing operations

[alexnguyenn]

I think it will also break this:

lattice-surgery-compiler/src/lsqecc/logical_lattice_ops/dependency_graph.py

Lines 113 to 116 in faad034

	for parent in s.parents:
	if parent.op not in visited:
	visited.add(parent.op)
	queue.append(parent)

Which is part of the DAG generating method, with visited being a set. Looking through the codecov report seems like it is not covered by our test case which is why I missed it. I might need to look more into it but I think we can just use the node there instead of op (by visisted.add(parent))?

[alexnguyenn]

Yeah I will fix the type annotations in there as well. They are a mess 😓

[alexnguyenn]

lattice-surgery-compiler/src/lsqecc/logical_lattice_ops/dependency_graph.py

Lines 41 to 45 in faad034

	def __eq__(self, other) -> bool:
	if isinstance(other, DependencyGraph.Node):
	return self.op == other.op
	else:
	return False

lol I override __eq__ for Node as well. Might need to get rid of it.

[gwwatkin]

It's good we have all these notes. Don't think we need to act on it yet.