discussion: support for TypeScript

[profnandaa]

This is a fresh take at #1261

Re-posting my initial thoughts:

So, a rewrite is out of question. Looks like viable proposals we have right now are:

1. Add types alongside the library (for easy maintenance), as opposed to doing it from DT
2. Find a way of keeping DT up-to-date (uphill task coz of lack of autonomy)
3. [my addition] Have a separate repo here in the validatorjs org specifically for types (which can be mirrored with DT)

Personally I'm leaning towards 3. I'm also trying to:

• protect a small footprint for the library.
• be less disruptive with what we already have.
• keep supporting those who already were using DT (@types/validator).

Let me know what you folks think.

/cc. @tux-tn @johannesschobel

[profnandaa]

Cross-ref #789 #247 #662 #1194

[deadcoder0904]

Would love support for this :)

How do I use Tree-Shaking with TS if anyone's got ideas?

I have already installed @types/validator from DT & would love to incorporate tree-shaking as well with it.

It currently gives errors if I directly use it from es/* libraries as they don't have types associated with them

[rcollette]

@deadcoder0904 - I imported like

import isEmail from 'validator/lib/isEmail';

That's as small as you'll get for tree shaking with TS alone, otherwise you need Webpack or something similar to get tree shaking.

[rcollette]

@profnandaa - I'm curious why the comment about not wanting to rewrite. JS is TS. There shouldn't be a need to rewrite anything.

You could just rename the files to TS and apply type information. You have the option to compile to multiple targets if you want ES5, etc.

If you don't want to write TS at all, you could use JSDoc, perform type checking with the compiler to ensure usage is consistent, and generate type declaration files from the .js

https://www.typescriptlang.org/docs/handbook/type-checking-javascript-files.html https://voxpelli.com/2019/10/use-type-script-3-7-to-generate/

[AubreyHewes]

@profnandaa You should not mirror own types to TD.. TD started as a repo of libraries that had no types. (read npm packages without types). It is not "the repo for types". The idea is that packages provide own types (authored by the package) and that TD / @types/etc slowly disappears

In the next version that has own types.. simply supply the types within the package.. anyone using TS will, after installing the package, have your types.. and will not need to install from @types/etc .. having to also install types from TD should be seen as a bad practice and can also introduce other problems (type packages with dependencies).

One thing you definitely do not want to do is keep a separate repo/branch with type definitions that need to be updated with each master source change.. I agree with @rcollette just move to TS (possibly with allowJS to get you going) and compile expected targets

[rubiin]

This is great. We can do the typescript version. I think the time consuming part is on the creating types @profnandaa

[fedeci]

On express-validator we have some incomplete type definitions that may work well as a starting point.

[rubiin]

On express-validator we have some incomplete type definitions that may work well as a starting point.

yeah that will help

[vlascik]

You already have a near complete rewrite at https://github.com/fireflysemantics/validatorts/ . Merge it here, replace that weird angular build system with tsc to create ambient declarations, use babel to generate es5/es6 etc. packages, and that's it. You can see an example of that e.g. here https://github.com/microsoft/TypeScript-Babel-Starter, but I'm sure more examples exist.

Every other solution is just pointlessly complicating things.

[AllusiveBox]

Curious if there was any decision actually made on how to go about using this? I am trying to argue for using this package instead of an internally managed one for work and they don't want to make the switch if there isn't any TypeScript support.

[WikiRik]

Curious if there was any decision actually made on how to go about using this? I am trying to argue for using this package instead of an internally managed one for work and they don't want to make the switch if there isn't any TypeScript support.

I don't believe so, but using @types/validator in combination with this package works fine in my projects

[AllusiveBox]

I don't believe so, but using @types/validator in combination with this package works fine in my projects

I missed that. Thanks, I'll let them know and see if that settles them for now.

[tux-tn]

I don't believe so, but using @types/validator in combination with this package works fine in my projects

I missed that. Thanks, I'll let them know and see if that settles them for now.

Agree with @WikiRik , @types/validator should cover all your needs.

[adamgen]

We can actually support types in a much better way.

DT offers type safety for the writing of the schema itself, however you cannot infer the end object's type from the schema like you can with other TS first libraries.

I've made a demo showcasing how we can easily infer types from our schema.

1. Github repo
2. Stackblitz link

@profnandaa I'd like to hear your thoughts.

[braaar]

As a TypeScript enthusiast I would be happy to contribute to this. Writing types for stuff is very satisfying :)

[profnandaa]

@adamgen -- definitely I'm ready revisiting this topic once again. Let us make this month's release and then resume this.

/cc @pano9000 @tux-tn @ezkemboi @rubiin

[rubiin]

I will create a draft then anyone who would like to contribute can push changes . CC https://github.com/validatorjs/validator-deno for reference

[WikiRik]

I think that it might be better to open a PR here that changes the infrastructure so that both TypeScript and JavaScript (including possible .d.ts files) are supported and we can migrate from JavaScript to TypeScript incrementally

[rubiin]

@WikiRik maybe you can start a PR and I can add in the changes. How does that sound. I am happy with either way

[fedeci]

For me the most viable solution is to convert the entire library to use TS. It is safer because of typings, prevents having to define types separately and it is quicker to understand the code for new contributors. I honestly cannot see drawbacks in adopting it.

[adamgen]

I have to take a step back from what I've proposed since it doesn't fit with the current state of validator.js. However, I'll take a few steps forward down the road.

Going type-heavy on validator.js in its current form will bring little to no value.

That's because the validator.js returns mostly primitive types. And a complex type system with inference, as zod has, is only relevant for schema validation, not for primitives. This gap between primitive vs. schema validation brings me to my next point.

Why wouldn't validator.js validate schemas? It already does on express-validator, where @gustavohenke has done a great job.

Now I'm entirely out of my domain with what I'm about to suggest, but this is how it goes.

I suggest having the schema validation already done by @gustavohenke in express-validator to be merged/combined with validator.js. Only then adding the inferred types I proposed earlier.

It'll be freaking awesome if you ask me. It'll give not less than scores of developers a lot of validation*type power without installing additional dependencies or migrating to new code, which I've found myself doing on a large project that maintained.

Also, I can be completely non-breaking, and since 80% of the work is done across the board, it requires a low time investment.

The 2 points above make a good open source ROI.

Is it too crazy? Is opening a new ticket for it will be a waste of time? Is there no PR that can move the needle forward on this topic?

[ffxsam]

I have one request if validator undergoes a major update:

Please simplify ESM tree-shaken imports:

import { isEmail, isURL } from 'validator';

validator/es/lib/isEmail is a bit inconvenient and clunky. The syntax I suggested will bring this package up to modern standards.

[ST-DDT]

@profnandaa / maintainers: What are your requirements for a community powered TS migration?

• No or very little extra work for the maintainers (only checking the types in future new PRs)
• No breaking changes for the users (excluding potential type incompatibilities with the unofficial @types/validatorjs)
• ?

I assume that there are quite a few community members (in addition to me) that would help with the migration itself. We just need the requirements and an "go ahead".

[WikiRik]

@ST-DDT indeed minimal breaking changes, but I want a major release either way when we ship our own types so it's clear to the users. I'm willing to update types if needed after we have merged a PR that copies the .d.ts files of @types/validator. And then preferably slowly migrate from .d.ts/.js files in the source code to .ts files

[ST-DDT]

but I want a major release either way when we ship our own types so it's clear to the users

I understand that and it sounds reasonable.

I'm willing to update types if needed after we have merged a PR that copies the .d.ts files of @types/validator.

Is there a good time to start the process (as an external contributor) or do you have your own roadmap for that change?

[WikiRik]

I'm willing to update types if needed after we have merged a PR that copies the .d.ts files of @types/validator.

Is there a good time to start the process (as an external contributor) or do you have your own roadmap for that change?

As far as I know there is no roadmap, but it would be good to have approval from @profnandaa @tux-tn on a possible roadmap before we start executing it