How to win the coding interview

Am I the only one that is has been coding for years and don't expect people to know regex off the top of their head?

There are so many variants and keeping them straight between languages is a nightmare. Grep, egrep, POSIX, perl, python, php, javascript. They all have their idiosyncracies.

Last time I made interview rounds, about 50% of places gave whiteboard questions, 25% gave essentially the same questions but had me do them on a computer or screenshare rather than a whiteboard, and the rest had some kind of take home component that was usually larger than the whiteboard question but also more practical and relevant to my skill set. Obviously I prefer the last option, but even if you prefer to give whiteboard-style questions, why not at the very least allow them to complete it on their laptop in front of you if they so choose?

It seems incredibly arrogant to me that interviewers assume there is only one way to solve a problem, and that is by brainstorming on a whiteboard first. I don't feel comfortable standing up in front of the class to present a solution. I feel far, far, far less comfortable if it's a coding-related solution. I did significantly better, like pretty much 100% success rate on coding components when I could hit keys rather than display my chicken scratches on a whiteboard.| I move code around a lot as I prototype and it takes quite a bit of training for me to adjust to not having the option of restructuring as I code because the medium is a whiteboard or piece of paper. And every time I erase something my level of anxiety and nervousness just grows.

Why regexes combined with the pre-processing?

First, it's not so clear in what sense "efficient" is actually a real requirement if this is a model answer.

If the average input is very small then the cost of compiling/interpreting the regex is going to dominate any modest run-time improvement over even an extremely naive implementation.

And if the inputs are very large then the probability of contradiction in the first few characters even is extremely high, so all the upfront normalization is a huge waste.

I'm having trouble imagining the typical data set where this would be an efficient implementation and where there even exist non-ridiculous inefficient implementations.

Second, doesn't using regexes kind of defeat the whole point of the Palindrome exercise from an evaluation perspective? If someone asked if me if they could use regexes in a coding interview for this question my response would be "good instinct. If you have no other way then go for it. But I was really hoping to see you work through implementing/debugging/commenting/testing an implementation without regexes."

Overall quite subjective. You're not going to 'win' the interview every time. Sometimes the cards are practically stacked against you. You get asked a question no one had actually tested beforehand, you get an interviewer who's trying to prove how smart he is, you get someone who is too fixated on a particular skill you must have, you get someone having a bad day, etc... And when the process easily take 6 stages and unanimous thumbs up, one of these is likely to come up time to time no matter how prepared you were. Don't take it personally.

and return the original value