Changed fixed tests as you suggest.

Oh, this is just to demonstrate that best-first search (which is faster than breadth-first search and A* when it's provided with an actual distance function, if I remember correctly) doesn't pass big random tests.

?

I guess that can be considered ok, yes.

I still think it would be easier for everyone to have the batch of tests on the small area. Currently, the last fixed tests are alrady asking for some kind of performances, so it would be more confortable for the user if the "area batch" was done before those "somewhat big" fixed tests.

Your call. Just close the issue if you really don't wanna do it.

• 1, 1 and 3 edge tests have been added to sample tests, example tests and more tests respectively.

• Changed the random test generator so that it now always generates 10% of edge tests.

Could you check and tell me if this is enough?

Yes, you can't just return the empty list.

I think it is better because you don't have to deal specifically with the case where the start and end points are the same (at least that's how it works in my solution).

you return [start]

• The start position may duplicate the end position.
• Returned path should be a list and nonempty(obviously).

So if start equals end, we need to run around some, we can't just return the empty list?

Oops, didn't know this eval way existed => Fixed.

About the side note: It's too embarrassing to see my original cocky name pop up again, but since it's already on the fork title, do whatever you like x/.

I tried it here. If I didn't mess something up, it looks like the current test suite is ok (failing in the medium tests already).

about the sidenote: can I report that publicly on github (issue) or do you prefer I keep it "private" (I can report directly to kazk about that strange behavior)

Since the O(1) distance function is easy to Google, I try to forbid those solutions that use best-first search with that distance function. However I found that those solutions don't even seem to pass the big random tests.

So, I made the following update:

• Python 3.8 -> Python 3.10

• Reduced the size of one last test, and now my solution will consistently pass all tests in 6-7s.

• Hid the previously exposed distance function in preloaded.

I'm going to close this, and if there are any problem, please feel free to raise a new one :).