This is a story about an interview that I had with a Fortune 500 Tech Company some time ago.
Back in the days I came across a Twitter profile of an employee claiming they were looking for a frontend developer.
I usually do not apply for big companies, for a number of reasons. They usually do not respond to your application and frankly it’s hard to be noted when they have that many applications, and my CV does not look that appealing anyway.
If we sum up this with the fact that some of these are using Taleo (oh boy…), my default action is to run away.
However, if you see an employee, a developer, publicly shouting they’re hiring for their team, things are different.
It usually means they’re looking for people, for real. I’ve seen a lot of times companies putting dozens of job offers even though they’re not hiring. It’s a marketing strategy tipically used to scary the competition showing you’re growing that fast you need a lot of people - but at the end of the day, it’s not true. On the opposite, you might be on the way of closing the company (I’ve been there).
It’s very likely you’re going to speak and chat directly with the target team - therefore you know in advance if you like the people, the position, what they’ve been working on lately and - more importantly - you skip the hiring call where these folks sell things about the company you’re not really interested at that stage.
I decided to apply. I thought it was worth to give it a shot and if they refuse me, we’d be friends anyway.
I contacted the folk privately and I was surprised in a positive way: the response was almost immediate, congratulating with me for my profile and they were inviting me to send all my details via email so they could evaluate me properly.
I was happy - that fast response was validating the point #1 I mentioned in the paragraph above.
The things from there went very quickly: I had a phone call with an HR person and successively I spoke directly with the engineering manager of the team that was hiring (validating the point #2). Both of them went very smoothly and, given the great followups I received, the idea of joining in such company was kind of shaping in my mind.
Next step was the code challenge - whose solution has been put here
I took some time and solved the whole thing - then I submitted the solution and then I wrote a follow up email to the guys so I could explain how I decided to approach the challenge.
Good morning everybody,
yesterday I’ve completed the frontend task assigned to me; I hope I tamed it in a way you’ll like it.
I’d just like to share with you couple of notes I took during its resolution.
Clearly I focused on solving the bug first. My initial implementation (that’s not the final one I submitted) was tracking the number of disks on the screen with an internal counter. The increment was (and it’s still) happening when the element has just been appended to the DOM. On the other hand, the decrement was happening by tracking the progress variable in the update function. Once the value of progress was >=1 , it means the element isn’t visible anymore and I can safely decrement the value.
Then, I moved to the optimisation part.
Important foreword: the provided implementation didn’t look that clunky and sloppy (100-150 disks on screen). Not sure if it’s because I’m on high end computer - I think we’re missing a bit of context here (what’s the desidered performance? Where’s this thing supposed to be placed?); anyway I realised that there were some things not performed in an optimal way. In particular:
- The disks, once landed beyond the screen, were simply “abandoned” there. I considered this as a leak (as the reference is basically lost) and might blown the DOM if I would continuously click on the Go button. Therefore, I’ve decided to remove it once the animation was completed.
Second thing, I changed the fade in effect to use another animation instead of relying on jQuery for that; successively I modified the way the disks are moved on the screen animating the transform property rather than the left one. The reason is simple: opacity and transform are likely to be offloaded by the browser to the GPU. The other properties, as far as I know, are still handled by the CPU.
You can find my considerations about testing in the HackerRank test.
If I would have more time, I’d probably modify the code to try to cache and reuse the DOM elements, instead of creating and removing them every time we click on the Go button.
Then I would probably try to reorganise the code a bit (it’s a bit spaghetti, but I guess it’s on purpose to confuse the candidates); probably using React to write the view and manipulate the DOM so I can extrapolate the real logic in a simple function that’s way easier to test.
Thanks a lot for your time in reading this, and have a good day!
Anyway, couple of days later I had some spare time and I decided to keep working on the task and implement the DOM optimisation I was thinking about - so I wrote a second email.
Hello again - One last thing and then I’ll shut up, I promise.
I wanted to implement the DOM nodes reutilisation caching. You can find its code here:
So now the elements aren’t removed anymore when the animation is over, but eventually they will be “reutilised” when another disk is required. If there are not enough “stale” disks to satisfy the required number, new elements will be animated on the fly (and then kept as a pool).
Have a great weekend,
I was sure I made it.
After a couple of weeks I received an email announcing they didn’t want to move further with me.
I rechecked the things multiple times, I even asked some friends to have a look into the solution to see if I made any stupid mistakes - but nobody pointed out something really valuable. I was confused.
Therefore I replied and I asked for a feedback. If I had been missing something so basic - I wanted to know.
Here’s the answer I got (I hope the guy will apologise me for quoting, but for correctness of the information, I must write the entire thing)
The general feedback was that you had showed some positive signs in the coding question and good write up as well, however the exercise just missed the bar for this particular role in [team_name], and one example was the use of CSS animations to solve the problem, which is not ideal as it invalidates a lot of the performance-tuning the team expect.
[team_name] has a very large legacy codebase and felt that your approach wouldn’t suit what they are looking for right now, but definitely wanted to keep in touch about other roles in the future if you would like.
I was shocked - at that point, I really wanted to know what was wrong here, and I replied one more time.
Hello [human_name], thank you very much for the answer.
That’s kind of weird - I made multiple performances considerations in my write up I sent you where I motivated my choice of using CSS animations. There were precise reasons why I decided to not use requestAnimationFrame, for example.
One last thing - would it be possible to get in touch with the team that evaluated that? I’d be really willing to know what would have been a better solution, it would be a great learning experience.
Thanks a lot,
Unfortunately, I didn’t get any reply and they vanished.
Fundamentally I was surprised at first, and disappointed later. Surprised, because I genuinely thought the point of the interview was to make sure I had an idea of what I was doing in terms of solving a bug and as well as reason about using the right API for the job; disappointed because I hoped to have a detailed feedback to improve.
Instead I got silence.
I’ve been thinking about this for a while, and I wasn’t able to find any reasonable explanation around that. If you get a blank screen with a problem to solve that is completely isolated without any context around where this solution is going to be placed it’s hard to reason about a legacy codebases.
In my email, I also pointed out that there were not enough informations to make a complete analysis.
Important foreword: the provided implementation didn’t look that clunky and sloppy (100-150 disks on screen). Not sure if it’s because I’m on high end computer - I think we’re missing a bit of context here (what’s the desidered performance? Where’s this thing supposed to be placed?);
I’m firmly convinced we can do better than that. Providing more context, setting the right expectations are key parts of a interview process in order to verify some particular skills. I hope that [the_company] will take this as a genuine feedback and eventually adjust the process.
I still love you anyway.