Length check added to simple_cycles

[vakker]

Adding a max_len parameter to simple_cycles according to #4797
Johnson's algorithm needs to be checked to see if this interferes with it.

[dschult]

Are you going to check whether this interferes with Johnson's algorithm?

[vakker]

So it's been discussed here but we haven't really reached a conclusion on how to combine the two.

[dschult]

It looks like that discussion ended here with a request that you construct a separate function for the length check version so that it wouldn't affect the runtime of the case max_len is None. Perhaps as a first check that this is true, you could time some graphs with your version and having max_len=None and the existing networkx version. Is there a difference in runtime?

If there is a difference, you could in-line the function call (python can be slow to call functions)
if nbrs and max_check_len(path): becomes:

if nbrs and (max_len is None or len(path) < max_len):

[rossbar]

Gentle ping @vakker have you had a chance to investigate the suggestion above? I suspect that the version without the function call overhead would indeed be more performant, though it'd be interesting to see what the practical effects are (if any).

[vakker]

@rossbar sorry for the delay, it didn't really have time to investigate this further. Probably during the holidays or early next year I can wrap this up.

[vakker]

@rossbar sorry, I don't have the bandwidth at the moment to further investigate this thoroughly, but if anyone comes along it should be straight forward to pick it up. Maybe @boothby has something to add here?

[dschult]

Fixed in #6461 :)