Make default validator truncate long tokens included in judgemessage.txt

[gkreitz]

The default validator will, as a debugging hint, write the first incorrect space-separated token it finds. This can get extremely large when grading a wrong answer submissions where the submission doesn't emit any spaces.

As the hint is just to help humans, it would be much more helpful if the token is shortened to some relatively short maximum length.

[ghamerly]

I was thinking about this, and implemented something quickly. But I'm on the fence. What do you consider to be "relatively short"? I was initially thinking 100 characters, but it then occurred to me that there are likely problems that have judge tokens that are at least that long -- and truncating could hide differences we want to see (and therefore make it harder to understand what is going wrong). So maybe 100 is too short; is 1000 too short as well?

Generally: how do we do this without making it more difficult to debug in some cases?

[ghamerly]

Proof of concept: ghamerly@9a2382b

[gkreitz]

I've also started working on this issue, sorry for not assigning it to myself. I'm also adding some test coverage of default_validator, so it's a bit less scary to touch.

My plan is to truncate a token to approximately 30 characters if it's the only token being printed. However, for the case where the user token differed from the judge token, my plan is to shorten it like: firstXcharacters...firstDifferingcharacters..., to hopefully preserve usefulness in debugging (arguably even improving it - if you have tokens over 30 characters, it's not fun to manually find the first differing character if it comes a long way into the token.

One annoying aspect is that output is almost always utf-8, so one either needs to explode this into making default_validator utf-8-aware, or one needs to at least take care not to cut within a character.