Slicing on classical registers

[vmartinv]

I'm trying to write code in QASM 3 and this is the only compiler I found. I'm having problems with slicing classical registers as shown below:

OPENQASM 3;

bit c[8]="01010101";
print(c[0:4]);

gives me:

bash-5.1$ qcor -qpu aer -shots 1  test.qasm -o test

[OPENQASM3 MLIRGen] cannot parse the source.
error: no viable alternative at input 'print(c[0:'
occurred at test:3:9, offending symbol = :

[qcor-exec] fatal error.

Am I doing something wrong?

Using let it seems that it somehow thinks is a qubit array:

OPENQASM 3;

bit c[8]="01010101";
let slice = c[0:4];
print(slice);

gives me:

bash-5.1$ qcor -qpu aer -shots 1  test.qasm -o test

[OPENQASM3 MLIRGen] Error
Could not infer qubit[] size from block argument. No size attribute for variable in symbol table.

[qcor-exec] fatal error.

I don't know if I'm making something wrong or if there's an error in qcor.

[ausbin]

Are either of these supported by OpenQASM 3?

Looking at the official grammar, it seems the slicing with : is only present in rangeDefinition which cannot be produced from an expression, and arguments to subroutines must be expressions. So the last line of the following doesn't seem to be syntactically valid:

OPENQASM 3;
bit c[4] = "0101";
print(c[1:2]);

Regarding using aliasing with a classical register, it seems like valid syntax according to the grammar, but I noticed the OpenQASM 3 spec currently says (emphasis mine):

The let keyword allows quantum bits and registers to be referred to by another name as long as the alias is in scope.

Notice that it doesn't say anything about classical registers.

I'm not sure if I'm taking the (living) spec too literally, and I'm not trying to argue against supporting this, just my thoughts

[vmartinv]

@ausbin

Looking at the official grammar, it seems the slicing with : is only present in rangeDefinition which cannot be produced from an expression, and arguments to subroutines must be expressions. So the last line of the following doesn't seem to be syntactically valid:

c[1:2] it's a valid expression, the expansion is as follows:
expression -> unaryExpression -> powerExpression -> indexExpression -> indexExpression indexOperator ->* Identifier LBRACKET rangeDefinition RBRACKET ->* etc

The let keyword allows quantum bits and registers to be referred to by another name as long as the alias is in scope.

Notice that it doesn't say anything about classical registers.

Exactly, it doesn't say anything about classical registers so the grammar becomes our only source of truth, thus the expression is valid if we're being logical. Specifically, it says "let keyword allows X" and not "let keyword only allows X". On top of that I think the correct reading is: "(quantum bits) and registers". From the common sense side, this should be supported.

[ausbin]

You're right, thanks for correcting my misunderstanding of the official grammar. However, I don't think the grammar in the qcor parser supports that:

qcor/mlir/parsers/qasm3/antlr/qasm3.g4

Line 345 in 6f86835

: unaryOperator expressionTerminator

(compare with https://github.com/Qiskit/openqasm/blob/3bba89ab69c440934663af55d139890a4569311f/source/grammar/qasm3.g4#L391)

So it seems to me like the grammar in this repository needs to be updated (which I assume will require updating the visitor that walks the parse tree, oh boy)