Warning instead of `ValidationError` for creg size = 0

[ryanhill1]

Currently, a program that includes a creg declaration with size 0 raises a ValidationError. The creg of size 0 is a convention used by pyqpanda3, which we are currently trying to integrate into the qBraid-SDK transpiler (qBraid/qBraid#963). However, the ValidationError demonstrated below is causing errors in testing, see comment.

import pyqpanda3

program = pyqpanda3.core.QProg()
program << pyqpanda3.core.RX(0, 3.1415926) << pyqpanda3.core.CNOT(0, 2)

qasm = pyqpanda3.intermediate_compiler.convert_qprog_to_qasm(program)

print(qasm)

OPENQASM 2.0;
include "qelib1.inc";
qreg q[3];
creg c[0];
rx(3.14159260) q[0];
cx q[0],q[2];

import pyqasm

module = pyqasm.loads(qasm)

ValidationError: Failed to parse OpenQASM string: L4:C6: creg size must be positive

I think we should instead just emit a warning if creg size = 0, but still raise a ValidationError if creg size < 0. If creg size is set to 0, and no further creg operations exist in the program, I think the program should still be considered "valid", but we should emit a warning and recommend that instead of declaring a creg of size 0, they should just omit the creg declaration.

Obviously, if creg size is set to 0 and the program also contains a measurement operation, for example, that should still warrant a ValidationError.

[TheGupta2012]

@ryanhill1 the error traceback which I see is from openqasm3 -

openqasm3.parser.QASM3ParsingError: L5:C6: creg size must be positive

This means that creg c[0]; is a syntactically invalid program, in which case semantic validation does not make sense.

If we see the OpenQASM 2.0 grammar on Page 21 of the Open Quantum Assembly Language paper, the grammar rule for the declaration is of the following form -

This clearly states that a nninteger (non - negative integer) is to be present as the size for the declarations.

I believe the correct way forward would be to contribute this change upstream to pyqpanda3 so that it generates a syntactically valid program

[ryanhill1]

But 0 is, in fact, a "non-negative integer", so doesn't that mean that the program is semantically valid based on that grammar rule?

[TheGupta2012]

Ah yes, that was a silly deduction , apologies!
I'll see if we can capture this error inside pyqasm or contribute this change to upstream openqasm3

[ryanhill1]

Yeah if it's an openqasm3.parser.QASM3ParsingError, we should raise an issue in the openqasm3 repo. Could maybe even contribute the PR as well, because it seems like it would be a simple fix.

[TheGupta2012]

Opened issue #592 in the openqasm3 repo, will track the updates and move ahead once a decision is made by the qasm maintainers