Should JSON files be counted as code?

[kadykov]

Hi all,

I’d like to discuss the current behavior of Tokei regarding JSON files and the code counting mode.

Background

Right now, Tokei includes .json files in the code count by default. However, JSON is a data format, not executable code. This means that projects with large or auto-generated JSON files (like package-lock.json) can have their code statistics heavily skewed, sometimes making the codebase look much larger than it really is.

For example, here’s my output:

vscode ➜ /workspaces/mcp-openapi-schema-explorer (main) $ tokei --no-ignore-dot
===============================================================================
 Language            Files        Lines         Code     Comments       Blanks
===============================================================================
 Dockerfile              1           32           14            9            9
 JavaScript              3          169          147           13            9
 JSON                   10        15136        15136            0            0
 TypeScript             42         6044         4538          817          689
-------------------------------------------------------------------------------
 Markdown               11         1249            0          909          340
 |- BASH                 3           18           12            5            1
 |- JSON                 3          183          181            0            2
 (Total)                           1450          193          914          343
===============================================================================
 Total                  67        22630        19835         1748         1047
===============================================================================

As you can see, the JSON files are much larger than the actual codebase.

Current Workaround

To get accurate code statistics, I have to add **/*.json to my .tokeignore file:

**/*.json

But this means I lose the ability to count all lines (including JSON) in lines mode, since the files are ignored entirely. (There is a --no-ignore-dot option to disable custom ignore rules, but that affects all ignored files, not just JSON.) Additionally, if I want to display two Tokei badges—one showing the total number of lines (including JSON) and another showing lines of code excluding JSON files—this is not possible with the current configuration.

Related Issues

• #313: Suggests excluding package-lock.json because it is big and auto-generated.

Badges

Tokei offers badges for lines, code, etc. Including JSON in the code count can make these badges misleading for some projects.

Summary

I believe excluding JSON from the code count, but keeping them in the lines count would be a more sensible default, as it better reflects the actual amount of source code.

What do you think? Please vote and share your thoughts below!