David Coudert

Actually, there is another issue: `nauty_posetg` works only for $n\leq 16$. You could use a switch to use it only in this case and use the previous (slower) method otherwise.

I don't see how to avoid the topological sort. However, you can avoid a copy of the graph ```diff - yield FinitePoset(dig.relabel(label_dict, inplace=False)) + yield FinitePoset(dig.relabel(label_dict, inplace=True)) ``` note also...

These lines can be reached setting `data_structure=None`, `immutable=None` and `sparse=True` or `False`. In fact, lines 154-182 are copy-paste from method `copy` in `generic_graph.py`. May be I should make a method...

We'll see if the new tests are sufficient.

Thanks for the review.

@maxale, this is ready to be reviewed.

j'ai tout refait. Pas réussi le rebase ou autre.

@fchapoton, could you help reviewing this PR ? Thanks.

Je trouve ça très pratique. Ca évite de recopier systématiquement le même test / message d'erreur et permet d'unifier le type d'erreur.

Something like `TypeError: this graph is immutable and so cannot be modified` ?