Refactored to ESM ("EcmaScript Modules")

[catamphetamine]

Changes:

• Added TypeScript definitions by copy-pasting them from DefinitelyTyped.
  • This allows using this package in TypeScript applications. But it also moves the responsibility to keep the file updated to the maintainers.
• Replaced dynamic import("@aws-sdk/client-s3") with optional predefined global variables global.GetObjectCommand and global.HeadObjectCommand.
  • This fixes a bug when bundlers can't bundle an app that uses unzipper package because of that dynamic import() statement when @aws-sdk/client-s3 package is not already installed by someone else in the application directory.
• Replaced package duplexer2 with a copy of duplexer3.
  • Reasons? Not really, just had a bit of fun. The readme of duplexer3 just says: "Modern version of duplexer2".
  • duplexer3 had to be copy-pasted (with the tests) because it didn't provide CommonJS exports. Called it DuplexStream.js.
• Replaced package bluebird with a combination of native Promises and a copy of p-map.
  • Reasons? Not really, just had a bit of fun. Bluebird source code seems to be no longer available, and it's unlikely that it supports "tree shaking". Hence, this change theoretically reduces the size of unzipper itself.
  • p-map had to be copy-pasted (with the tests) because it didn't provide CommonJS exports. Called it mapPromises.js.
• Removed package fs-extra.
  • Reasons? Not really, just had a bit of fun. This reduced the number of dependencies by 1. Called it ensureDir.js.

---

While this pull request is in review, the changes have been published as a separate package: https://www.npmjs.com/package/unzipper-esm

---

P.S.

There's a strange issue with the tests: the results seem visually same, yet the testing framework complains about some kind of a mismatch.

The same issue can be observed when cloning the original unmodified repo and running the tests. Therefore, I assume that tests are actually passing and it's just an operating system compatibility issue (I'm on Windows). Perhaps it's related to line endings being different between Windows and Linux.

[ZJONSSON]

This is awesome — really appreciate the work here.

Could you split this into a few smaller PRs? Right now it blends several different changes together, which makes it difficult to review each one properly.

Ideally, each PR should cover one logical change. For example, ESM changes only could be one PR, Bluebird removal another, and any unrelated cleanup or behavior changes should be separate as well.

[catamphetamine]

Hi,

Indeed, I agree. If someone submitted a PR in my repo, I'd probably have answered the same.

The thing is, I got a bit carried away while doing the refactoring and decided to also update all dependencies due to how massive the ESM change alone would be.

Frankly, all of the dependency replacements do make sense from my point of view. The general rule is: the smaller the count of dependencies, the better (in public opinion). And also: the fresher the dependency packages, the better. Both principles were applied when doing the refactoring.

Could I split it in several atomic PRs? I could but if all those are gonna eventually be merged anyway then there would be no point really and it would feel pointless doing that. So I think that I'll pass. I'd just prefer to be done with it and not open it again, haha.

If you're interested, I already shipped this PR in 'read-excel-file' package. For no reason, really. Just had some fun refactoring on a chill day.

[catamphetamine]

I can see you've added CI tests 👌

I'll see why it didn't pass when I get home later.

[ZJONSSON]

Well I really appreciate your efforts, so I will take a look and see if I can split them up.

Frankly, all of the dependency replacements do make sense from my point of view. The general rule is: the smaller the count of dependencies, the better (in public opinion). And also: the fresher the dependency packages, the better. Both principles were applied when doing the refactoring.

Could not agree more

I can see you've added CI tests

The CI tests have been there for years - but I had to permission them for each build. You should be able to npm run test locally and see if everything works.

[catamphetamine]

Ok, I’ll see if I could split them as well. I could create a new branch where I’d move the non-essential stuff.

…

On Sun, 21 Jun 2026 at 22:37, Ziggy Jonsson ***@***.***> wrote: *ZJONSSON* left a comment (ZJONSSON/node-unzipper#356) <#356 (comment)> Well I really appreciate your efforts, so I will take a look and see if I can split them up. Frankly, all of the dependency replacements do make sense from my point of view. The general rule is: the smaller the count of dependencies, the better (in public opinion). And also: the fresher the dependency packages, the better. Both principles were applied when doing the refactoring. Could not agree more I can see you've added CI tests The CI tests have been there for years - but I had to permission them for each build. You should be able to npm run test locally and see if everything works. — Reply to this email directly, view it on GitHub <#356?email_source=notifications&email_token=AADUP3ZSAI6DX5CRZVTSU3T5BA2PPA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZWGMYDONRRGI2KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-4763076124>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADUP3Y3ENALHF3EAZOKYBT5BA2PPAVCNFSNUABEKJSXA33TNF2G64TZHM3DENRUGA2TSOJ3JFZXG5LFHM2DMMZTGM2TCNJUG2QXMAQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[catamphetamine]

Here's a shortened version that's just about ESM: #360

[catamphetamine]

@ZJONSSON Hi. I can see you've merged the ESM refactor.

This PR has been updated to reflect the other changes.

These "other" changes are completely optional and see for yourself if you like some of them or not.

[gameroman]

You can just use Duplex.from

https://e18e.dev/docs/replacements/duplexer

[catamphetamine]

Replacing duplexer with Duplex.from results in errors in tests:

Error: test end() method called more than once

The reason is that there's code like this:

  const extractor = src.pipe(Extract({ path: dest }));
  extractor.on('close', function() {
    t.ok(fs.existsSync(path.join(dest, 'subdir', 'file.txt')), 'normal entry must be extracted');
    t.end();
  });

The close event can be emitted multiple times when using Duplex.from():

• When the readable stream closes
• When the writable stream closes
• When the duplex stream closes

Workarounds:

• Use import { finished } from 'stream' instead of stream.on('close', ...)
  • That would have to be replaced throughout the code (+ tests) and there's a hypothetical probability of stuff malfunctioning after the change, so I won't go this way.
• Use .once() instead of .on()
  • This one seems even hackier, so I'll pass too.

In summary, there seems to be no simple find-and-replace solution so DuplexStream.js is not removed.

[catamphetamine]

[catamphetamine]

"node.js duplex.from readable writable close event emitted multiple times"

[gameroman]

I wonder if there is a way to solve that