How Do You Pick a Programming Language?

Beware of arguments related to programming speed. All things being equal, faster is better. But all things are never equal. Do you need the kind of speed that lets you get a website up and running quickly? Or the kind that allows you to rotate a few thousand polygons in 3D in real time? Do you need to convert 10,000 PDFs into text per hour? Or 10 million PDFs into text once? These are different problems. What do we need to do, how many times do we need to do it, and what existing code can we use to help us do it that many times? Ask those questions.

It’s possible to spend productive months preparing for a project without deciding on a language. It may be the sign of a fine manager, someone who assumes his people can learn new things, someone who’s built an agile team capable of experimenting with new technologies and getting ideas into production. It could also be that this person is totally useless. You’ll find out!
Let’s say your programmers are developing a huge website that serves 5 million people who each visit five times a month. Do you use Python, which is slower, or Go, which is fast, or Node.js, which is something in-between? Trick question! Twenty-five million Web page visits isn’t that big a deal, unless they involve some deep wizardry or complex database queries that are very different for each page (good example: Facebook).

Now, that number isn’t trivial; if it takes a minute to make a page, you’d need 48 years to make that many, which is way too slow. If it takes a second to make a page, that’s still too slow—there are only 2.6 million seconds in a month. So you need to figure out how to serve about 10 pages per second. You’ll probably want more than one computer, a little redundancy, some good server setup. It will take some doing and planning. But it can be done in any language. 

Programming Languages’ Greatest Hits

Assembly
Pac-Man, Centipede

C
Unix, Linux kernel, Python, Perl, PHP

C++
Windows, Google Chrome, software for F-35 fighter jets

Python
Instagram, Pinterest, Spotify, YouTube

PHP
Facebook, Wikipedia, WordPress, Drupal

Perl
BuzzFeed

Java
Google, EBay, LinkedIn, Amazon

Ruby
Twitter, GitHub, Groupon, Shopify 

 

 

 

 What if you are going to serve only a few hundred thousand pages a month? Then you’ve got tremendous breathing room. You don’t need too many engineers to create the system architecture. You still need to plan, but in general you can read some blog posts and follow along with what others have done. You can be pretty sloppy, to be honest. Again, any language will do.
What if you want to include a live, person-to-person chat on those pages, and you expect thousands of people to use that chat at once, all speaking to each other? Now you’re dipping your hand into that godforsaken river. But that is exactly the problem that Go was designed to solve. It’s a language for creating highly available servers that use as much of the computer’s processor as possible. It has other features as well, but this is where Go shines. Actually, Node.js works pretty well for that sort of server, too, and Clojure certainly has the capacity. Oh, right, Java works, too. If you really needed to, you could even do it in PHP.
This is why the choice is so hard. Everything can do everything, and people will tell you that you should use everything to do everything. So you need to figure out for yourself what kind of team you have, what kind of frameworks you like using, where people can be most productive, so they will stick around through the completion of the project. This is hard. Most places can’t do this. So they go with the lowest common denominator—Java, PHP—because they know that when people leave, they’ll be able to get more of them.
And that’s OK. The vast majority of technology projects don’t require original research, nor do they require amazing technological discoveries. All the languages under discussion work just fine. There are great coders in all of them.
But the choice of a main programming language is the most important signaling behavior that a technology company can engage in. Tell me that you program in Java, and I believe you to be either serious or boring. In Ruby, and you are interested in building things quickly. In Clojure, and I think you are smart but wonder if you ship. In Python, and I trust you implicitly. In PHP, and we sigh together. In C++ or C, and I nod humbly. In C#, and I smile and assume we have nothing in common. In Fortran, and I ask to see your security clearance. These languages contain entire civilizations.
You can tell how well code is organized from across the room. Or by squinting or zooming out. The shape of code from 20 feet away is incredibly informative. Clean code is idiomatic, as brief as possible, obvious even if it’s not heavily documented. Colloquial and friendly. As was written in Structure and Interpretation of Computer Programs (aka SICP), the seminal textbook of programming taught for years at MIT, “A computer language is not just a way of getting a computer to perform operations … it is a novel formal medium for expressing ideas about methodology. Thus, programs must be written for people to read, and only incidentally for machines to execute.” A great program is a letter from current you to future you or to the person who inherits your code. A generous humanistic document.
Of course all of this is nice and flowery; it needs to work, too.