Prevent spread `...obj` inside a `arr.reduce()`

[styfle]

I've seen this pattern used several times where the developer is using spread inside a reduce causing a O(N^2) algorithm when avoiding the reduce would lead to O(N).

Example usage we want to avoid:

O(N^2)

  const originalFiles = {}; // Imagine 100k files

  const remappedFiles = Object.keys(originalFiles).reduce(
    (mappedFiles, file) => ({
      ...mappedFiles,
      [`_next/${file}`]: originalFiles[file]
    }),
    {}
  );

Could we write a lint rule to prevent this pattern and guide the developer toward the O(N) solution?

Example usage we should encourage:

O(N)

  const originalFiles = {}; // Imagine 100k files
  const remappedFiles = {};
  for (const file of Object.keys(originalFiles)) {
    remappedFiles[`_next/${file}`] = originalFiles[file];
  }

[mrmckeb]

This is quite interesting, have you seen any lint rules around this that we could turn on?

[mrmckeb]

@styfle did you have any thoughts around how we might introduce a rule like this?

Obviously this would need some discussion, as a lot of JS engineers like functional programming.

[styfle]

Obviously this would need some discussion, as a lot of JS engineers like functional programming.

I think there is little to discuss when we consider performance.

In fact, @pvdz just found a case where changing the spread in the loop brought the build time down from 18 min to 4 minutes.

Also see The reduce ({...spread}) anti-pattern blog post.

[styfle]

did you have any thoughts around how we might introduce a rule like this?

I assume we need to write a eslint plugin. Something like this:

https://github.com/mkosir/eslint-plugin-no-array-reduce/blob/63b7792d1bbf3f8d3578fdb6fbde5982555f12c0/src/rules/disallowArrayMethod.ts#L11-L24

Or we could disable .reduce() completely with no-array-reduce, but that might upset some engineers as you mentioned.

[AlexErrant]

I googled "javascript spread reduce eslint" after reading an article on Tanstack Table and arrived here. I think a lot of devs (including myself) don't think about the implications of .... I'm not smart enough to help write a rule; just here to express interest :)

[Conaclos]

We added this rule to Biome several months ago. Any feedbacks are welcomed :)