@Sanchises Conor did a check with a SEDE query, and I did another check with Suever's script. We got similar results: asterisk slightly more common, but nearly tied. I think hash is better because the body of the character is more centered in its "case", whereas the asterisk is placed a little above
@Sanchises Corrected now. Thanks! It helped me notice another bug in the same function. I have also applied the corrections here (which forms the basis if MATL's Za)
Commited:
* Corrected two bugs in Za (base2base): handling of input consisting of all "zeros" (thanks to @Sanchises for noticing), and properly left-padding the ouput with "zeros" if needed.
@Sanchises Also, duely credited in the ackonowledgments section of the doc
Pity that the bug makes it hard for MATL to compete in the base-97 challenge
@LuisMendo I really like that not to be a typo, and there's an actual duel going on in the docs now. Thanks, there was no need to credit me but I will print it out and hang it over our bed.
What do you mean by left-padding with zeroes?
You mean that ' 'T256Za produces more zeroes than ' 'T256Za?
You can never make that truly bijective anyway.
@LuisMendo I know how features made especially for a challenge are outlawed, but are bugfixes too? Anyway, it would make for a very fun answer to have someone work around that bug. At least then my easy generate-the-testcases program is out of the question.
@Sanchises I mean that something like {'abadca' 'aba'} 'abcd' 'ABC' Za would give ['BACBCA'; ' BB'] (left-padding with spaces) instead of the correct ['BACBCA'; 'AAAABB']. But actually that only happened in the old Matlab function; not in the MATL one. And now I see I have messed things up and the MATL function is giving the wrong result. I'll fix it in a minute.
Yes, in answers, bug fixes are disallowed too. Languages are defined by their implemenation, bugs and all
Working around the bug is not that costly... I think I got 19 bytes in total, although I didn't test thoroughly