Tapa answer checker with ? is wrong

[Riverness]

Tapa answer checker considers ? clues with no shaded cells around them to be correct.

I would except the first one to be erroneous.

Not sure if related or not, but it's possible to enter a single 0 as a clue. Maybe this is the root cause?

[robx]

Just to explain some of the mess around question mark clues:

Across puzzle types, question marks primarily exist in the pzprjs applet as a placeholder during construction. For most types, the basic puzzle type is understood not to have any question mark clues, and the rules are implemented as such.

The applet has always allowed exporting puzzles with such placeholders as vanilla puzzles, so now we have this unfortunate mix of standard puzzles and question-mark-variant puzzles. The latter typically have more subtle rule sets, and it would be appealing to keep the base type free of such subtleties. (E.g. the rules page shouldn't have to talk about question marks.)

Not sure there's a good way to fix this. Perhaps we could have a general "with placeholder" variant that activates when there are any placeholders. Then this would be a bug not of "Tapa", but of "Tapa (placeholders)".

A bug either way, of course. :)

[x-sheep]

I've now documented the behavior on the rules page: https://puzz.link/rules.html?tapa

For Tapa specifically, the zero clues have always been a bit of a sticking point. Even without the question mark, they don't look right since clues are actually sets. (You don't see zeroes in multi-digit clues...)

Changing question marks to disallow zero is guaranteed to break the answer check for existing puzzles which rely on the question mark being zero.