styleguide sanity-check for tc39 language-proposals to address javascript-fatigue

# kai zhu (10 months ago)

there are several factors for the current javascript-fatigue. one factor which tc39 could help mitigate is to provide a narrative on how to consistently apply proposed language-features (over existing-practices and interfacing with legacy-code).

i feel too many new and old javascript-programmers alike are unable to adopt a consistent programming-style for post-es5 features in production-code. style-issues which are problematic when a project has to deal with legacy libraries include:

  • when is it appropriate to use callback vs promise vs async-generator vs async/await, when interfacing with legacy-code (aka context-switching-hell or baton-passing-hell)?
  • when is it appropriate to use var vs let, when interfacing with legacy-code?
  • when is it appropriate to use function vs fat-arrow, when interfacing with legacy-code?
  • how can we apply destructuring in a consistent and readable manner?
  • when is it appropriate to use (proposed) pipeline-operator, and when is it not?

es6/es7/es8 introduces hundreds of these kinds of questions which distract us from actual coding and shipping features.

# Jordan Harband (10 months ago)

These questions have consumed programmers in most languages since forever. It's not TC39's place to tell people how to write code - but there's plenty of style guides that have answers to these questions.

# Naveen Chawla (10 months ago)

My opinions:

For existing code, just transition when encountering code, for convenience (no need to transition everything in a go).

For new code, always use await async where possible (very manageable vs callbacks) For new code, always use const (for references you don't expect to change) and let (for references you expect to change), and ditch var altogether (reason: var lets you accidentally override any global variables) For new code, always use fat arrow EXCEPT when using a library that relies on the use of "this" inside a function, where the "this" is different than the outer "this" (reason: more terse, and lets you access the outer "this"). Destructuring: as per your convenience. Pipeline: never unless they allow await in a pipeline (reason: could lead to inconsistency), otherwise always (instead of () for calling functions). However, pipeline operator is not yet accepted in the language.

# T.J. Crowder (10 months ago)

On Wed, Oct 18, 2017 at 6:44 AM, kai zhu <kaizhu256 at gmail.com> wrote:

there are several factors for the current javascript-fatigue.

I don't believe any such thing exists.

es6/es7/es8 introduces hundreds of these kinds of questions which distract us from actual coding and shipping features.

I've certainly never been distracted by these things, other than in the usual way ("Hmmm, in our house style are we always going to use () with arrow functions even if there's only one parameter, like we always use curly braces with if?"), which isn't significant and is certainly a very small price to pay for the added utility of these features.

Including libraries? Fine, use the API they provide (callbacks or promises) or wrap that API if you don't like it but need the lib anyway, and certainly don't worry about the style the library uses in its code. No distraction.

Frankly, this just seems like you're trying once again to push your "JavaScript should have frozen in time as of ES3 or maybe ES5" agenda, which I don't think anyone else on this list shares.

-- T.J. Crowder

# kai zhu (10 months ago)

Frankly, this just seems like you're trying once again to push your "JavaScript should have frozen in time as of ES3 or maybe ES5" agenda, which I don't think anyone else on this list shares.

its a legitimate agenda from someone who cares deeply about javascript and believes es6 was a mistake and a step in the wrong-direction for javascript and frontend-development.

-kai

On Oct 18, 2017 3:02 PM, "T.J. Crowder" <tj.crowder at farsightsoftware.com>

wrote:

On Wed, Oct 18, 2017 at 6:44 AM, kai zhu <kaizhu256 at gmail.com> wrote:

there are several factors for the current javascript-fatigue.

I don't believe any such thing exists.

es6/es7/es8 introduces hundreds of these kinds of questions which distract us from actual coding and shipping features.

I've certainly never been distracted by these things, other than in the

usual way ("Hmmm, in our house style are we always going to use () with arrow functions even if there's only one parameter, like we always use curly braces with if?"), which isn't significant and is certainly a very small price to pay for the added utility of these features.

Including libraries? Fine, use the API they provide (callbacks or

promises) or wrap that API if you don't like it but need the lib anyway, and certainly don't worry about the style the library uses in its code. No distraction.

Frankly, this just seems like you're trying once again to push your "JavaScript should have frozen in time as of ES3 or maybe ES5" agenda, which I don't think anyone else on this list shares.

# T.J. Crowder (10 months ago)

On Wed, Oct 18, 2017 at 8:18 AM, kai zhu <kaizhu256 at gmail.com> wrote:

its a legitimate agenda from someone who cares deeply about javascript and believes es6 was a mistake and a step in the wrong-direction for javascript and frontend-development.

I won't get into an argument with you about ES2015+ except to say I vehemently disagree with your conclusion, and challenge your characterization that there's some kind of "fatigue." I'm constantly meeting programmers who are thrilled with the new features -- three real standouts are arrow functions, class syntax, and async/await. I show async/await to people and they go crazy for it. "Whoo hoo, no more callback hell!"

But the point is: The ship has sailed. Re-litigating decisions that have been made is pointless and tiresome. When new proposals are made, it's perfectly valid to raise issues with them if you think there are issues (and perhaps cite concrete issues that have arisen from similar past work to support that argument -- with data, not innuendo), but complaining about arrow functions, let, Promises, etc., in October 2017 is not useful. Those decisions were finalized years ago.

-- T.J. Crowder

# Alexander Jones (10 months ago)

The beauty of (coding) standards is that there are so many to choose from. :)

IMO it’s a false dichotomy though. A respected and credible group of language contributors should pool some energy together and ratify some opinionated best practices, a la the C++ Core Guidelines and PEP-8. No, it’s not necessary—neither is the exponent operator—but it does have clear benefits.

I believe most in the community would rather not have to sell things like “const by default” to their team members, when it could be “official” guidance instead. It’s energy we’d rather be spending on other things!

Alex

# Naveen Chawla (10 months ago)

I disagree that the language contributors should be involved in best practice guidance. Patterns evolve over usage and experience with the constructs. I bet the implementors of && and || didn't necessarily expect them to be used so effectively for non-boolean logic e.g. car && car.drive() instead of if(car!==undefined) car.drive() or whatever... Or maybe they did. But the point is language usage is often a matter of opinion and preference, and not something that should be set as a tide against a possibly justifiable opposition. As a response to the original question, I gave my opinion and reason in brackets. If the reader prefers a different way for their own reasons, fine - I would just expect them to give their own reasons for superseding my reasons...

# Isiah Meadows (10 months ago)

In general, it's not the TC39 who should be dictating how code is written - in particular, even they have their stylistic disagreements (like with ASI and let vs const), and active TC39 representative maintain both JSHint (opinionated) and ESLint (unopinionated). Additionally, JSLint (the predecessor to JSHint) was created by a formerly active TC39 representative. If you want to see more of these broad stylistic disagreements, check out their meeting notes. A few things that come to mind are decorators, cancellation, recent class additions, and BigInt.

Instead, if you have your own strong opinions on everything, try introducing ESLint to your projects. They have numerous presets and rules built-in, and you can create your own custom presets, rules, and plugins. If you want to ban null, write a custom rule for it. If you want to ban anything not ES5, write a rule that catches every expression that isn't ES5. If you want to define local rules, use eslint-plugin-local. In my case, I decided I didn't want to use default exports, so I wrote a local rule banning all default exports. Not that I have a problem with those who use it - I don't. I just feel it's easier for me to wrap my head around named exports without introducing the cognitive overhead of default exports.


Isiah Meadows me at isiahmeadows.com

Looking for web consulting? Or a new website? Send me an email and we can get started. www.isiahmeadows.com

# Alexander Jones (10 months ago)

I used the term 'language contributors' rather than TC39 as an intentionally vague way of describing people like us.

The ISO C++ Committee also lacks a consensus on everything, but that doesn't mean those people and the people around them can't debate and establish a consensus on something. Hence, C++ Core Guidelines.

I think the reality is that people are averse to this because they don't want their pet practices at work being discouraged by anything resembling official guidance — having to justify a decision to use var instead of const by default is effort, right? But they're perhaps not always considering the benefits that an improvement in (not necessarily total) uniformity can bring.

I claim that the majority in the Python community would say that PEP-8 has been a net benefit. (Yes I break its rules from time to time. That's what rules are for. ;) )

# Michał Wadas (10 months ago)

Though, PEP-8 was created in 2001 and in significant part codified already used conventions. And JavaScript does not have such widely followed conventions - except camelCase for functions and PascalCase for classes.