Hacker News Reader

Comment by swampthinker | Hacker News Reader

Max Howell's story really surprises me. You would figure that someone would've internally fast tracked a developer like him.

View on HN

He wrote a bit more about it on Quora [1]. He still feels he should have been hired, but I think the Quora answer makes it pretty clear that he's not what Google is looking for in software engineers.

[1] https://www.quora.com/Whats-the-logic-behind-Google-rejectin...

That one is famous. We had a quite lively discussion about it here shortly after the event.

Edit: to this day I have no fucking idea what “inverting a binary tree” means, other than the interviewer was too proud to admit they made a mistake.

Trees have two dimensions and “invert” is the one you can’t change.

It just means making

- 2

-- 4

-- 5

- 3

--7

into

- 3

-- 1

[union]

- 2

-- 1

[union]

- 2

-- 1

as one graph. I must say I'm surprised someone can't write this given enough time and proudly claim "they didn't hire me because I couldn't do this". Seems like CS 101 hw.

As he says, he wasn't a CS major. I can totally relate. I had a pretty popular DOS shareware back in the day (along with some other things). But I wasn't a CS major and would not have been able to answer most algorithmic CS questions unless I had specifically studied them for the purpose of an interview.

Fine, but don't you think companies should have some filter to choose candidates? In my company I interview often, and most candidates don't know how to code. Asking CS hw questions is a signal that predicts if they can code or not. If you're not a CS major it's ok but in general if you can't figure this question in 15 mins I don't want you to push code to my codebase. Some level of competence is ok. And I don't even work in a FAANG this is a random no name startup in Boston.

There's one particular category of question that I've started building up a question set in, in part because the last few coworkers who weren't nearly as useful as you would hope all struggled with it:

Data transformation.

Most of us are at least partly in the business of data transformation. Few of us should be implementing trees (you want coworkers who are informed consumers of data structures, not people who think reimplementing solved problems at work is a reasonable or even sane thing to do).

The simplest one I have is this: Hand them a data structure containing a list of lists, have them generate a list of the grandchildren ordered by some property of the grandchild.

Like take a list of cars or computers grouped by make and model and give me all of the models sorted by price, or by popularity. If that's too simple, filter by another criteria and do a little data cleanup.

I've seen people drop elements, get the order wrong, scope variables too broadly and confuse themselves into grabbing data from a previous iteration, or forget to grab it entirely. And this is exactly the sort of thing that puts your team on the spot for having dumb bugs in production. Not inverting binary trees, or fizzbuzz.

Reversing a tree (max element on the left) is absolutely CS homework (though probably 201). I'm sure I had to do that at least once or twice. But you can also avoid that most of the time by writing a second in-order traversal algorithm that runs from right to left.

But why would I need a data structure where I have the elements by the leaves instead of the roots? Make the nodes doubly linked (parent element) and you get the same effect without all the grandstanding.

It means to reverse it left-to-right, like a mirror.

You will never have to implement this yourself for real.

You won't because it's a toy problem, not because it something only greybeards did in libraries written decades ago.

Asking someone to describe the simplest possible transformations on one of the simplest data structures isn't a stretch, and this is coming from someone without a CS degree.

EDIT: And that's not to say Google doesn't have a screwed up hiring system. I know someone who implemented a dynamic programming problem correctly, with the appropriate big O complexity, but was rejected because the constant time factor was slightly higher than what the interviewer was looking for. That's a BS decision; Max Howell's was not.

There's another term for reversing a tree: reversing a tree.

'Reverse' is switching the orientation left to right or front to back. Top to bottom is 'inverting' (except in graph theory where inverting means something entirely... weird).

> other than the interviewer was too proud to admit they made a mistake.

Reply on news.ycombinator.com