Heap out of memory error

[reneerevis]

I'm using vitest with a typescript project. Fairly early on and I'm new to vitest, but it's been working well so far. However, I have hit a wall and I'm baffled.

I am running only one simple test and if I run it within the vitest harness, I get a JavaScript heap out of memory, but if I run the exact same code outside of vitest, it runs fine with no errors.

If I run the same test with a sentence that is a bit simpler, it works fine. Debug shows the memory used is minimal:

Before the parsing begins:
{"level":"debug","message":"Starting parse of sentence: 'The children are ready.'","timestamp":"2025-12-02T02:23:44.851Z"} {"arrayBuffers":811837,"external":4408972,"heapTotal":69435392,"heapUsed":46342568,"level":"debug","message":"Memory Usage (bytes):","rss":164937728,"timest\ amp":"2025-12-02T02:23:44.852Z"}

After the parsing has finished:
{"arrayBuffers":943622,"external":4541422,"heapTotal":69697536,"heapUsed":50438520,"level":"debug","message":"Memory Usage after parse (bytes):","rss":16767\ 3856,"timestamp":"2025-12-02T02:23:45.130Z"}

But if I add one additional word to the sentence (e.g., 'The first children are ready'), it fails during the initialization phase before the main parsing even begins. And again, this same sentence runs with no errors outside of vitest.

I upgraded from vitest 3.0.8 to 4.0.14 hoping that might help, but am still seeing the same issue. After upgrading to 4.0.14, I set vmMemoryLimit: '1024mb' in vitest.config.ts.

I really don't know how to figure out what is causing the issue and would appreciate any help!

Edit: I did one other thing and added debug to the initialization routine that fails so I could monitor how much the memory changes as I parse each word/part of speech. The changes are as expected and are very minimal. When I parse the sentence outside of vitest, it does bump the heapTotal slightly on the last word, but nothing unexpected. When I run it with vitest, that is the point where it fails with heap out of memory.

[tt-a1i]

This looks like Vitest is hitting V8 memory limits in its worker pool, while your code runs fine in a normal Node process.

Likely causes:

1. VM isolation overhead: Vitest runs tests in isolated VMs with their own memory limits
2. Worker memory constraints: Each worker has its own heap limit

Try these solutions:

1. Disable VM isolation:

export default defineConfig({
  test: {
    isolate: false,
    pool: "forks",
    poolOptions: {
      forks: { singleFork: true },
    },
  },
});

2. Increase Node memory limit:

NODE_OPTIONS="--max-old-space-size=4096" npx vitest run

3. Use forks pool:
Forks get their own process with full memory access.

4. Debug the initialization:
Since it fails during initialization, not parsing, check your test setup.

The "one additional word" clue:
This suggests your parser might have exponential complexity - each additional word dramatically increases memory usage. This is common with certain parsing algorithms (CYK, Earley with ambiguous grammars). The parser works with simple sentences but explodes with slight complexity increases.

What kind of parser is this (NLP, programming language, etc.)?

[sheremet-va]

This is not correct. Vitest does not use vm to isolate test files. Vitest runs a single file in a single worker.

There is also no such thing as poolOptions anymore, see migration guide: https://vitest.dev/guide/migration.html#vitest-4

And forks pool will also be limited by memory

[tt-a1i]

This is not correct. Vitest does not use vm to isolate test files. Vitest runs a single file in a single worker.

There is also no such thing as poolOptions anymore, see migration guide: https://vitest.dev/guide/migration.html#vitest-4

And forks pool will also be limited by memory

thanks for the correction, my bad - i was working off outdated info. appreciate the clarification on how vitest actually handles isolation

[reneerevis]

Thanks to both of you for the feedback.

I ended up getting through it. I think there were multiple issues, some on my end and some with vitest config.

I'm now using only one config option with the latest version: vmMemoryLimit and I'm not seeing the issue anymore.

I also found an infinite loop issue that happened in a particular circumstance which was what was affecting that particular sentence (i.e., it wasn't actually about the number of words).

The parser is my own very specific algorithm for English sentences, and I've thankfully been able to confirm doesn't have memory explosion issues.

I appreciate the help from both of you!