Make RouteCQC errorout on intermediate measurements on 3+ qubits

[ghost]

router mishandling mid-circuit measurements #issue6293

Description of the issue

cirq.RouteCQC commutes gates in undefined ways due to intermediate measurements with 3 or more qubits not being handled correctly.

This PR aims to fix this issue by:

1. Raising an error and provide a descriptive message on what's occurred

Next Steps

1. Break intermediate measurements on 3+ qubits into single qubit measurements if -and only if- the result isn't stored (i.e. key is None)
2. Check if it's possible to create a ClassicallyControlledOperation controlled by the result of a multiqubit measurement

• if it's possible then we stop here until there is a need to support this case because then we will know how this should be handled.
• if it's not possible then break all intermediate multiqubit measurements into single measurements even those that have a key (because they are not using it anyway) while logging a warning to the user that the key is being dropped.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[NoureldinYosri]

thanks for helping with this issue. please check https://github.com/quantumlib/Cirq/blob/master/docs/dev/development.md#continuous-integration-and-local-testing for CI instructions. currently your PR will fail format and pylint tests and maybe others.

In the describtion section of your PR please explain what your change does. also the title should be informative.

---

As it stands this PR doesn't fix the problem. There are two ways to go about this

1. Modify the algorithm so that it handles intermidiate measurements correctly.
2. treat intermediate measurements has seperators and split the original circuits into multiple smaller circuits and then combine the results.

The first harder conceptually, the second is more of a coding excersize. I think we prefer the first option unless the routing algorithm https://arxiv.org/pdf/1902.08091.pdf can't support this case.

[ghost]

Thanks for the help, will review the routing algorthm to see if it supports intermidiate measurements case. I should be able to implment the second implemention just to get a quick solution out but will also look for example of handling intermidiate measurements correctly in the source code and on online.

[NoureldinYosri]

I think, it's better to not support intermediate multiqubit measurements. I'm not aware of any usage of intermdiate multiqubit measurements that can't be done with single qubit measurements.

[NoureldinYosri]

this seems like a lot of changes for a simple check. The check is just

if any(protocols.num_qubits(op) > 2 and protocols.is_measurement(op) for op in itertools.chain(*circuit.moments[:-1])):
   # There is at least one non-terminal measurement on 3+ qubits

[ghost]

Yes, apologies, I pulled the changes from upstream so that my branch would be up-to-date before doing the commit. It added changes I wasn't expecting.

I did 2 separate commits Errorout for mid-circuit measurements on 3+ qubits with just my changes and Merge commit 'e81919ef8e48ae3efa7f866d6d3e98bd480e7c10' into shef4/is… to fix the divergence.

Was able to revert to the intended commit. The change should only be class, unit test, and cleaned-up example notebook files.

[NoureldinYosri]

almost there. a few more changes. after you are done please run

./check/format-incremental --apply

and commit its effect

[NoureldinYosri]

a few more things.

& please clear you changes to the notebook

[ghost]

Added the changes you recommended. Cleaning up the notebook was done in removed debug code commit. Will try and replace the debug code with an empty cell in the meantime just in case.

[NoureldinYosri]

@shef4 check the changed files then

[ghost]

I see what the changes are, will update now.

[ghost]

Just updated the notebook. Changed files made it way easier, thanks!

[NoureldinYosri]

@shef4 please fix failing CI checks and update the branch to head

[ghost]

currently testing change locally

[ghost]

I ran CI and everything is passing.
Got a weird bug with code coverage saying my test cases need test cases to cover them but I think that was a mistake

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (aa312bc) 97.83% compared to head (59d8e21) 97.83%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #6307   +/-   ##
=======================================
  Coverage   97.83%   97.83%           
=======================================
  Files        1108     1108           
  Lines       96352    96372   +20     
=======================================
+ Hits        94270    94290   +20     
  Misses       2082     2082           

Files	Coverage Δ
...ore/cirq/transformers/routing/route_circuit_cqc.py	99.22% <100.00%> (+0.01%)	⬆️
...irq/transformers/routing/route_circuit_cqc_test.py	100.00% <100.00%> (ø)

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