I was wondering if it’s worth our time. Can we all just adopt the most popular style if there is one, and stop arguing? So I’ve analyzed 100 most depended-upon npm packages and ran a survey with the same questions that ended up with 1169 respondents.
Tabs or spaces?
- Two-space indentation is a clear winner, especially for open source projects.
- Some developers don’t really know what they use. Here’s my favorite answer: “I just press tab and the editor formats it to something else I don’t even know”.
Semicolons or no semicolons?
- Most open source projects do not use any linter presets.
- Many projects have no code style linting at all.
- Huge projects with many packages use their own presets: Babel, Gulp, PostCSS, Facebook.
- Sindre Sorhus uses xo.
- I’ve included eslint:recommended (ESLint) because it’s relatively popular but it has no code style rules.
- Results are almost the opposite.
- AirBnb is the most popular preset.
- Standard is popular too (looks like all developers who don’t use semicolons use Standard).
- Which makes some developers quite angry: “Definitely not that ‘standard’ BS”.
- Many open source projects have very basic linting: not more than 10 rules.
- Some projects have
.jshintrc but do not run linter as a part of tests.
- Some projects have both
.jshintrc, some even with conflicting rules.
- Prettier was released just a few weeks ago but already adopted by quite a few people.
- And has made some people super excited: “Waiting for Prettier to gain more traction!” or “Will use Prettier once major bugs have been fixed”.
- Some developers don’t really care: “Code style doesn’t matter”.
The results show less diversity in code style of popular open source projects: more than half use 2 spaces with semicolons. Many still don’t use any tooling to validate code style.
It could also be because many popular open source projects are old but in the survey people described what they used on their latest project.
Note: two projects of 100 are written in CoffeeScript. CoffeeScript itself is one of them. They and one with only native code were excluded from the results.
All data and chart sources are on GitHub.