Strengthened definition of gflow means Pauli presimulation does not preserve gflow.

[emlynsg]

With reference to PR #337 and PR #347, and issue #168.

Describe the bug
Before and after PR #337, example two in the docs has changed.

"No flow or gflow detected in the graph."

Additional context
As explained by Thierry and Mateo, the definition of gflow was strengthened in PR #337 to match the definition from the paper (input nodes cannot be corrected by a gflow). This strengthened definition seems to be incompatible with Pauli presimulation preserving the gflow, leading to examples where a gflow was detected before.

However, Pauli presimulation preserving the gflow is a desirable and expected property, so we should consider how to deal with this change.

Need to change Pauli presimulation to force plus states.

[mgarnier59]

Not sure I understand what you mean.

input nodes cannot be corrected by a gflow

Do you mean that input nodes cannot can never be part of any correction set and thus have to be measured in the XY plane?

However, Pauli presimulation preserving the gflow is a desirable and expected property

Is it expected? I think it should preserve Pauli flow (cf https://arxiv.org/abs/2207.09368) but I'm not sure about gflow