Kata

Length of lists has been dropped.

In the first example it would be [1,...] from list b since it has an exact match. For the second example it would be [3,...] since it minimizes the difference and is within tolerance.

I re-worded the bullet point in the 'Notes' section where it's mentioned.

What about identical distances?

a = [2, A]    b = [0, X]    t = 2
                  [4, Y]

always the first match, so [0,X]

This info should be added to the description, and such cases must be tested in the fixed tests (and preferably sample tests).

Also what about this?

a = [2, A]    b = [2, X]
                  [2, Y]

Or are such cases guaranteed to not be generated?

it would be [2,X]

I can modify the description and and fixed tests. I'm a little reluctant to change the random tests, but I could add more? The pattern here is that we're minimizing the absolute difference between list a and list b in column 1. If that is minimized and the value is within tolerance, it will always pick the first value it can find. There are likely cases of this in the random tests.

Edit: By the way, this is mimicking a pandas function and I've checked it's intended behaviour with this function. Both line up ;)

The random tests are alright. I've implemented everything except the "pick first closest" scenario and failed 200+ tests. But all this stuff must still be tested in the fixed tests where the input size is small, expected result doesn't change all the time, and different values are used in the last column, e.g.:

a = [[2, 5]]    b = [[2, 1], [2, 9]]    result = [[2, 5, 1]]

a = [[2, 5]]    b = [[2, 9], [2, 1]]    result = [[2, 5, 9]]

Also, it'd be good if the example in the description covered as many rules/details as possible, similar to how I rewrote the example in your other kata, and now it shows that: multiple columns are aggregated separately, rows with identical keys don't have to come one after another to be grouped, different combinations of keys are not grouped together.

That's half the fun of guessing what the hell the author wrote ;p

I'll let you know I've finished changes.

Okay, I've changed the fixed tests and implemented the two examples you gave above. Not sure about description. I'm on the fence since it lays out pretty much everything I can think of.

There're still no tests like these:

Equal difference - pick first:
a = [[2, A]]    b = [[0, X], [4, Y]]    t = 2    result = [[2, A, X]]
a = [[2, A]]    b = [[4, Y], [0, X]]    t = 2    result = [[2, A, Y]]

Non-equal difference - pick one with small difference:
a = [[2, A]]    b = [[0, X], [1, Y]]    t = 2    result = [[2, A, Y]]
a = [[2, A]]    b = [[3, X], [4, Y]]    t = 2    result = [[2, A, X]]

Changed. I think the user should be good to visit the test cases to answer any questions they might have. One thing about long kata descriptions is people generally don't read them and immediately jump into the coding.

(still waiting for a btach of small random tests... => trying another kind of implementation and it's passing the fixed tests but not the random ones)

EDIT: mmmh... XD Sounds like they are still a bit "too big", then.