Tagging noise operations clutters the circuit output

[mpharrigan]

The extra [] are from the tag, I think

xref #2719

0: ───X───D(0.005)[]───H────────D(0.005)[]───

1: ───X───D(0.005)[]───H────────D(0.005)[]───

2: ───────D(0.005)[]───H────────D(0.005)[]───

3: ───────D(0.005)[]───X^-0.5───D(0.005)[]───

[dabacon]

This is when you add the noise to the circuit? I think that should remove the tags.

[mpharrigan]

Interestingly, this is more broken than I first thought. In the terminal it displays as

(0, 0): ───I───D(0.001)[<virtual>]───I───D(0.001)[<virtual>]───I───D(0.001)[<virtual>]───

(0, 1): ───I───D(0.001)[<virtual>]───I───D(0.001)[<virtual>]───I───D(0.001)[<virtual>]───

(0, 2): ───I───D(0.001)[<virtual>]───I───D(0.001)[<virtual>]───I───D(0.001)[<virtual>]───

but the <virtual> tags are interpreted as badly formed html tags in e.g. Jupyter notebook and are not displayed. In the SVGCircuit implementation, the virtual tag hides the closing ]

[dabacon]

For physical versus virtual Z gates, clearly seeing the tag is important. But for noice, I'm not sure this is correct, and as I say in #2984 I don't think we should reuse virtual in this form for noise. It is re-using a tag for no real reason.

[mpharrigan]

Based on the above, the "virtual" specifier probably gets eaten in jupyter notebook for virtual Z gates then

[mpharrigan]

The behavior has changed. It now does D(0.001)[cirq.VirtualTag()]

[mpharrigan]

I'm contributing a hack in https://github.com/quantumlib/Cirq/pull/3076/files#diff-d949b9f3211c62c5defeb9f7d1439416R16 which should be reverted after this is fixed