Note that for this kata, the non-padding version (without = at the end) is used. (Reference: https://en.wikipedia.org/wiki/Base64#Implementations_and_history and the examples above that heading)

The encoding/decoding functions/test-cases are consistent in the sense that the encoding function expects non-padding strings as its outputs, and the decoding function receives non-padding strings as its inputs.

You don't get this kind of joy out of a stupid brute force solution if you knew beforehand that it was easy. -- i second that

No harm made! I think you feel frustated because you have to give up your beautiful O(nlogn) solution:-) Cheers!

I don't see clearly what is your problem, is it because the kata is too easy? You should understand that I wrote this kata for a lot of languages and not all languages have Big Integers (JS, CS, TS, C++). Moreover the difficulty level has to be almost the same in all languages. Last remark: this kata is not so easy since only 13 out of 294 completed it (4.5% whereas most kata have a success rate far above 10%) ...

See my answer to Mirzal below.

Sorry for your change! Thanks a lot for the feedback!

Did you see what I wrote below?

Do you think I should (could) add in the description that index i has to be as small as possible namely the first index which fits? (would seem the normal order: you choose a digit to remove and then you insert it, don't you think so?).

I think the results, though maybe ill explained, are consistent.
Please tell me what you think.

Edit: I modified the description.

The solution is to create a default value for the parameter according to the specifications of the kata :)

Was the no-arg Python test resolved? I can't get it to work.

You are right, thanks a lot!