Implementation of manual classical variables: RFC 0019

[jakelishman]

What should we add?

This is the tracking epic for RFC 0019: Variable-declaration and storage representation in Qiskit. Details of the proposal are in the linked document. Since the tasklist doesn't show linked PRs, I'm just assigning myself to issues as I open a PR that closes them to give a neater visual indication.

Required

Beta Give feedback

1. Add manual variable support to expr.Var #10923
   +0
   
2. Add Store instruction class #10939
   mod: circuit +1
   
3. Add storage locations and methods for classical variables to QuantumCircuit #10924
   mod: circuit +1
   
4. Add helper methods to ControlFlowOp for managing variables in scope #10940
   mod: circuit +1
   
5. Support manual variables in QuantumCircuit.append #10925
   mod: circuit +1
   
6. Add new wire structure for manual variables to DAGCircuit appenders #10926
   +0
   
7. Add equality comparisons for manual variables to DAGCircuit.__eq__ #10927
   +0
   
8. Support manual variables in OpenQASM 3 #10929
   mod: qasm3 +1
   
9. Support manual variables in QPY #10930
   mod: qpy +1
   
10. Add testing for manual variables to transpile #10931
    mod: transpiler type: qa +2
    
11. Add support for closing over variables to control-flow builders #10938
    mod: circuit +1
    
12. Documenting backend upgrades needed for Store in 1.1.0 #12332
    documentation +1
    
13. Add changelog entry for manual variable work (closes this epic)

Additional circuit methods

Beta Give feedback

1. Support classical variables in QuantumCircuit.copy and copy_empty_like #10959
   mod: circuit +1
   
2. Support classical variables in DAGCircuit.copy and copy_empty_like #10960
   +0
   
3. Support manual variables in QuantumCircuit.compose #10932
   mod: circuit +1
   
4. Support manual variables in DAGCircuit.compose #10933
   +0
   
5. Support manual variables in DAGCircuit.replace_block_with_op #10934
   +0
   
6. Support manual variables in DAGCircuit.substitute_node #10935
   +0
   
7. Support manual variables in DAGCircuit.substitute_node_with_dag #10936
   +0
   

Long term required, not necessary for 0.45

Beta Give feedback

1. Add support for visualising classical variables #10937
   mod: visualization +1
   

[jakelishman]

Current dependency DAG of the created PRs, where an arrow from A to B means "A depends on B".

In theory all the links are clickable, but I found it finicky unless you control/command-click to open in a new tab.

stateDiagram-v2

direction BT

classDef merged fill:#8250df,color:white,
classDef unmerged fill:#1f883d,color:white


state "<a href='https://github.com/Qiskit/qiskit/pull/10944'><span style='color:white'>Add representation to <pre>Var</pre> (#10944)
</span></a>" as 10944
state "<a href='https://github.com/Qiskit/qiskit/pull/10946'><span style='color:white'>Add <pre>Store</pre> instruction (#10946)
</span></a>" as 10946
state "<a href='https://github.com/Qiskit/qiskit/pull/10962'><span style='color:white'>Add <pre>QuantumCircuit</pre> var-handling methods (#10962)
</span></a>" as 10962
state "<a href='https://github.com/Qiskit/qiskit/pull/10963'><span style='color:white'><pre>QuantumCircuit</pre> copy methods (#10963)
</span></a>" as 10963
state "<a href='https://github.com/Qiskit/qiskit/pull/10974'><span style='color:white'><pre>QuantumCircuit</pre> append control flow (#10974)
</span></a>" as 10974
state "<a href='https://github.com/Qiskit/qiskit/pull/10975'><span style='color:white'><pre>ControlFlowOp</pre> write-detection helper (#10975)
</span></a>" as 10975
state "<a href='https://github.com/Qiskit/qiskit/pull/10977'><span style='color:white'>Control-flow builder support (#10977)
</span></a>" as 10977


class keyMerged merged
class keyUnmerged unmerged

keyUnmerged --> keyMerged

class 10944 merged
class 10946 merged
class 10962 merged
class 10963 merged
class 10974 merged
class 10975 unmerged
class 10977 unmerged

10946 --> 10944
10962 --> 10946
10963 --> 10962
10974 --> 10963
10975 --> 10974
10977 --> 10974

[jakelishman]

Pushing this from 0.45 because there are some further internal API considerations about how the transpiler should reason about classical-memory access. The system is already at the limits of special casing, and with me needing / wanting to add in a read/write memory-access split for variables into the transpiler, there's more considerations to be made about how downstream passes will work. There's no immediate dependents on this work being in 0.45, because it's all internal stepping stones for Qiskit 1.0 work, not stuff that's supported immediately on hardware.

[jakelishman]

All merged PRs are reverted by #11667 for 1.0.