Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: ts-remove-unused – Remove unused code from your TypeScript project (github.com/line)
141 points by kazushisan 4 months ago | hide | past | favorite | 66 comments
ts-remove-unused is a command line tool for TypeScript projects that auto-fixes unused `export`s. It removes the export keyword from the declaration or the whole declaration based on its usage in the project.

There are some similar tools but they are focused on "detecting" rather than "removing" so I've built one myself. I wanted a solution that's as minimal as possible; config files to specify the files in your project shouldn't be necessary because that info should be already configured in tsconfig.json. All you need to do is to specify your entrypoint file.

Feedback is much appreciated!




I tried it on https://github.com/renovatebot/renovate

It deleted 100s of files, most of which were Jest test files, and potentially all of which were a mistake. I restored them all with `git restore $(git ls-files -d)`.

I then ran `tsc` on the remaining _modified_ files and `Found 3920 errors in 511 files.`

Obviously at that point I had no choice but to discard all changes and unfortunately I would not recommend this for others to even try.


You need a valid tsconfig that defines the scope of the project and it seems renovate’s tsconfig doesn’t meet this requirement. You can always --skip manually as an alternative option.


Renovate's TS tooling seems to work fine otherwise, and for ts-remove-unused as a new tool wanting to get adoption then you should be aiming to work the way your users already work, and not fail + tell them that they are the ones which need to change.

BTW the VSCode extension which someone else linked to discovered everything perfectly, which is another hint that your tool needs to improve, not your users.

Even if tsconfig.json was very important, and the users are all wrong and you're right, the only thing your readme says is: "The CLI will respect the tsconfig.json for loading source files." which is insufficient.


This is simple and configurable with Codegen - see the results on `renovate` here: https://www.codegen.sh/codemod/4553/public/diff

Most transformations like this are not possible with pure static analysis and require some domain knowledge (or repo-specific knowledge) in order to pull off correctly. This is because some code gets "used" in ways that are not apparent i the code.

Enjoy!


yeah the `--check` option should be the default mode imo


You should switch the default to not delete any files and modify/remove the files only with some flag (--dry-run=false, --rm, --delete, etc). I just deleted all files accidentally in a monorepo :D Luckily I didn't had any uncommitted changes and could recover using git


It's maybe reasonable to change the defaults, but I think you should also be mindful of running any random cli program in your codebase, the top of the Github readme does indicate that the --check mode runs it without deleting files.


I’d maybe recommend inverting the command so it doesn’t delete by default but only when you add the relevant flag. Far better to add a little rigmarole than deal with a sad person who wiped their code.


I’m not the author of the tool, I was just chiming in. But yes, inverting it probably makes sense.


Arguendo, having to type extra characters to enable the functionality that the tool says it will do in its very name seems backwards to me.


Besides, the common behavior is to do what it says on the tin by default, but offer flags to make cli's "simulate" their run with "--dry-run" or similar.

I'm not sure I know of any utility that would have the opposite behavior than that.


Well, explosives. They do in fact explode (what it says on the tin) but you have to put in some effort to make them do so. Similar to wiping out your codebase when testing an app. Personally I think developer will reduce support load from angry people by adding guardrails. Someone will not have committed their code.

An alternative is to have a check “Are you sure?” Just before delete, with a -fuckit_yolo flag to override the check.

I once had a desktop app and realised over a while that just making copious backups (50, IIRC) meant users were less likely to get angry when they had lots of extra safety, even from their own actions.


I guess the "explosives" in this analogue could be deleting files with "unlink", "rm", "find -delete" or "shred", neither of which require a flag to actually perform the job they're meant to perform.

> An alternative is to have a check “Are you sure?” Just before delete, with a -fuckit_yolo flag to override the check.

Alternatively, is there any point you'd run this "ts-remove-unused" tool and you don't want it to delete anything? The main point of the tool is to delete things, so not sure what other use cases you have if you don't want it to delete something.


The "install -y" pattern is kind of similar. Various tools have as a default that it gives a list of packages that would be installed, and then ask for confirmation.


`install -y` is a fun analogue to consider here.

A "Hey, this is really going to delete files. If you're just playing around here, maybe try it with the --dry-run flag" seems sane and (so long as it's asked for) means less syntax to have to know up front.


rm <file> does not ask for any confirmation. Neither do most of the infra tools we use on a daily basis.


You're totally right. I just quickly scanned the readme and had expected a tool that only check the files. But yes, it's a bad idea in general to run random cli programs without checking them in detail


I’m sorry that it may have been unclear that the tool will make modifications without the check flag. I’m glad you didn’t loose any unstated changes :)

I will add a cautionary note to README to make it clear!


> had expected a tool that only check the files

Isn't the name kind of a giveaway that it does more than just checking? "ts-remove-unused" is plastered all over the place, and you need to even type/paste the name into your terminal before you can use it.


Don't people do that all the time? I doubt that even you check everything that you run thoroughly. There's a lot of trust we have put into all those tools.


It would be kinda cool to use git status to avoid accidental data loss for tools like this.

I've never interacted with git programatically so I don't know how messy it would be to implement. But for tools that mostly operate on "whole files" rather than lines in files, I guess it shouldn't be that tricky?


It’d be very easy to programmatically check which files in a repo directory are unknown to git. Git status does indeed have a machine-parseable output mode.


I tried it and it's pretty cool, I might introduce it in our company project. I did notice one problem/caveat: It doesn't play nice with dynamic imports. Our project has a few files being lazily imported and the tool seems to think that those files are unused, which is not true, they're just imported with the `import('./file')` syntax.

But other than that it's pretty nice, I might look into the code to see if I can help with that small bug.


I've been using ts-prune[1] for years at this point. The project is in maintenance mode but works fine so I've kept using it. I've been looking into Knip[2] which is recommended by the authors of ts-prune though it's been slow mostly because there's little incentive with the current tool working fine.

[1]: https://github.com/nadeesha/ts-prune

[2]: https://github.com/webpro-nl/knip


I was in a similar spot, but knip offers more information that ts-prune, and ts-prune may trip on some newer TS syntax. I've been happy using knip so far.


Deleted 80% of my project. Glad I could revert it but yeah pretty useless at this point


really nice that you're tool focuses on removing. i've been using https://knip.dev/ for detection in monorepos, but it's cumbersome to remove manually.


Knip does have some auto fix capabilities but it’s not perfect yet. The detection is great, though. I use it with `dpdm` to keep things tidy. With these (plus eslint, prettier and typescript) in CI, I feel all warm and fuzzy.

OP, can you describe differences from knip?


My understanding is that Knip is for detecting unused things and that it's auto-fix feature is still experimental, but with that out of the way...

- In general, I personally do not like the idea of having to add another config file in my repo. I feel it's contradictory that I need to add more to my codebase to clean up the clutter in my codebase. I understand that relying on tsconfig to specify the target files has its pros and cons but I hope that my tool will encourage users to maintain their tsconfig. - Knip has its own ecosystem around removing unused stuff in your codebase (not limited to exports). I personally prefer tools that are more single-purposed and does one thing right so I don't have any ideas for expanding features; It will be focused on auto-fixing unused code caused by unncessary `export`s.


Thanks for for the detailed answer. I'll try it out some time :)


Will it remove exports that are only imported for the sake of testing? Eg: it is only imported by files ending in .test.ts or with __test__ as a parent directory?

I’ve tried tools like this in the past within projects that have high test coverage but I have never had any luck because of this edge case.


I may be opinionated but I believe that the best practice is to configure a separate tsconfig for test files with project references. As long as the test files are not included in the tsconfig passed to ts-remove-unused, it should remove exports that are only used in test files.

https://www.typescriptlang.org/docs/handbook/project-referen...


In the JavaScript API, it would be nice if there was the ability to have custom comparative functions for this sort of use case. A user defined function that gets called right before final output which allows the user to create custom filtering rules and return a boolean to indicate inclusion or exclusion. I could see this being useful not only for the case I just presented, but also other cases such as NextJS projects that export default functions in pages that are never imported elsewhere in the project.


I’ve been using it for my next.js project and passing something like `/pages\//` to `skip` was good enough for me… Is there something I’m missing out?


hey! I was wondering if you'd be willing to explain what project references are and why you would do test config this way?


I wont go into detail about what it is (since the docs are good enough) but in short, it's a feature to let you setup multiple TypeScript projects in the same repo.

Some things that project references will solve: - importing test files from your impl files will throw an error - you can specify what types are available. for example you shouldn't be able to use `@types/node` in your frontend code or the `DOM` type in your test code (or vice-versa) - no type checking the whole repository again just because you changed the test.

I think there's quite some control over how strict you can be when you say you're using TypeScript. Project references is another way of making your config more robust, and it may not be necessary depending on your needs (like when some people disable strict:true). I think it's a better choice and relying on a properly configured project shouldn't be a issue for users who are actually going to use this (who are likely maintainers of their project)


I think this can get very nuanced --

If you are providing a library, it's possible you are exporting a function that is meant to be used by downstream code, and that function is isolated from other parts of the code (so never used by other functions but only tests)

If you are writing "product" code, most likely this is just dead code. But there are also edge cases where a function is used as the entry point for other code outside the current repository etc.

Put it this way -- if you are given a codebase you have never seen before, and you see a function only imported by test on the first day. Would you remove it without hesitation? Probably not.

I feel this is likely something that must require human experience to be done "correctly".


A custom comparative function that returns true or false based on the special requirements of the project might fix that. I don’t think the project supports this based on a quick look at the documentation but I would love to see this added as a feature.


This is cool, as a place which uses Typescript for a lot of things, including back-end services we handle it differently. Basically we have a rather strict linting process setup which will warn developers of unused code in development environments and outright refuse it for staging deployment and forward. I’m not sure we would “dare” to automatically remove it, because for the most part there is a reason it is there. Maybe it’s something that needs to be deleted, but it’s almost always something which needs to be handled in some way.

Unused imports is perhaps the one area where I would be comfortable removing unused imports. I would never personally allow a third party package into our environment for something like this. I really don’t want to be rude about it, but it’s too “trivial” to rely on an external import for. Not because your code couldn’t be better than ours, it probably will be, but because it’s an unnecessary risk.

For smaller or personal projects I think many people will rely on something like prettier to do the same.


Thanks for the input! I think there may be a misunderstanding about what this does. Existing linters work great for detecting unused code within a file but once you add `export` to it, you can’t detect unused code with linters even if it’s not referenced from any file.

You’re right that this tool may not be useful for some codebases. If your modules are more like “scripts” that include side effects, deleting modules just because it’s not referenced may break things. That should not be the case for react projects that are based on components.

In our development process, we don’t allow the changes made by this tool to be deployed automatically. Instead we make sure the changes are okay by creating a pull request and reviewing. We treat it more like an assistant that does all the cumbersome clean up work in no time.


Could you provide examples of such risks? Because in my understanding, if some function/constant is exported but never imported anywhere, then it must be dead code and never run. And if it was reachable, then TypeScript would fail the compilation. As such, it sounds reasonably safe to me to remove it.

I only see these potential risks:

1. Using a mix of TS/JS and having some blind spots where we could accidentally delete non-dead code without the compiler noticing.

2. Having and relying on side effects. For example, `export const foobar = thisFnWillDoSthImportant()` and then, yes, removing that would break things.

3. Having separate projects/libs where some consumer might be accessing your exports directly.

Do you see other risks than those?


> Could you provide examples of such risks?

The risk is adding yet another dependency controlled by a third party that executes on your developer’s local machines and on your CI system.


Great tool! It uncovered a surprising amount of unnecessary exports in our codebase. Really streamlines things.

One interesting observation: when using it with our Next.js project, it flags all page TypeScript files as unused. This inadvertently highlights a potential drawback of file-system based routing - it can lead to less explicit code relationships.


Cool. A prettier plugin does this too. Cant remember which, but maybe https://www.npmjs.com/package/prettier-plugin-organize-impor...?


I'm not trying to be offensive here, but I've realized that the design choices of the tool and the explanation in README was a little too difficult and unfriendly for the average TypeScript user.

I've added a more detailed explanation to README so I hope it will change your mind if you've previously had a negative impression :)

https://github.com/line/ts-remove-unused#readme


Neat! I will add this to my toolbelt.

BTW a complimentary tool I've used in the past is depcheck, it is an npm package that removes unused dependencies from your npm package.json file. Smaller package.json files means faster "npm install" and also smaller docker files.

https://www.npmjs.com/package/depcheck


If you are using vs code, it has an extension: find unused exports. It worked so far pretty well in our projects.

https://marketplace.visualstudio.com/items?itemName=iulian-r...


I've always used https://github.com/pzavolinsky/ts-unused-exports

Has more features (like excluding enums) and works very well in large code bases.


There are also plugins that nicely integrate with ESLint to do this.


Ah the typical “here’s a cool thing OP build but I’m using something better-reply”


Been passively wanting something like this for a while now. We have a good few dynamic imports so I'll have to work around that (as per another comment), but this is a much appreciated tool in the belt!


Ohh my, this is something that I would love in my projects... OK currently I don't have any Typescript project, only Python.

For python there is a lib for this, but it is a bit crappy.


Any tool that helps delete code safely is a win in my book! Even more so if I happen to personally know the author and randomly stumble over their submission on hn


Isn’t this what tree-shaking is for?

That’s a genuine question — I’m only passingly familiar with tree-shaking so I may have a misunderstanding of what it does.


Similar but different. Tree-shaking generally means excluding unused stuff when bundling for production. This actually deletes unused stuff from your source files, i.e. more for code tidiness.


Ah got it — thank you!


I think the command in the first image in the readme is wrong. Shouldn't it be `npx @line/ts-remove-unused`?


Thanks! will work on a fix. my original assumption was that users would `npm i` beforehand.


Pretty cool!

Could this also work with .svelte files, which are essentially html-like files with <script lang="ts">?


possibly... not a priority at this moment but contributions are welcome!


Another good alternative. Archived but still in use at work. https://github.com/nadeesha/ts-prune


I was initially using it but decided to write my own tool because I was too lazy to go through the result of ts-prune and delete the code myself :P


Does not work quite well with monorepos, just tried but overall a good idea


Perhaps you need to specify skip option for the entry point file?


I searched far and wide for this a few years ago for a disastrous React codebase I inherited that had a lot of unused components, never found anything.

This looks great, particularly the `skip` and `mode` options (which I'm guessing several commenters here missed).

I suppose it should work just as well with monorepos or any other directory hierarchy right? It only "knows" files are unused because they're never referenced in any code within the defined `projectRoot, and it only knows exports are unused whenever they're never imported?

Cool project, will definitely try it soon.


This is basically malware




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: