Hacker News Reader

Comment by Sawamara | Hacker News Reader

This is a very solid foundation to work with, for anyone who might have struggled with the contexts and how arrow functions, local variables, thises (hah) and prototypes fit into the bigger picture.

It also clearly shows that Javascript is not the mess that it looks like from a beginner's perspective. Yes, anyone can create a 15-20 min video about how == and === can mess up stuff, how you cant just pass around a function with a this in it without being careful, how NaN is wtf, and all that jazz. But at the end of the day, stick to this language long enough and it works for you.

View on HN

I wonder why nobody has created a version of JavaScript that just throws exceptions for all the WTF cases? That is, it would basically be a mode like 'use strict-at-runtime'; .

This would be incompatible, but it would be a useful development aid. Surely most frameworks like React and Angular avoid these corners of the language, and they could be trivially modified to run on such an interpreter? It would improve their code quality.

I know this would be a large amount of work, but it doesn't seem that huge compared to all the other JavaScript infrastructure out there (parsers, transpilers, etc.). There are more independent JavaScript interpreters written than interpreters of any other language, AFAICT. At least before ES6, writing a JS interpreter was a one-talented-person project, not a huge team project like a JIT.

I guess one reason is that the DOM and node.js bindings aren't easy to reproduce. But I would think people still run their unit tests in a limited environment and it would be useful for some code. It could even be based on narcissus -- JavaScript in JavaScript?

I think the most common way to accomplish that idea is to run a linter. Linters should be able to statically catch most of the unsafe operations you're probably describing.

Yeah, that's probably true -- linters do an OK job at attacking this problem.

But then there are still people complaining about JS semantics... I think there is a gulf between what tools professional JS programmers use, and what tools X programmers use when they write JavaScript, where X != JavaScript.

I think most JS fans these days recognize the inherent problems with JS, and work around them with best practices (on the lenient side) or linters and more advanced preprocessors like type checkers (on the strict side). At the end of the day, by far the main advantage of JS is the fact that there are mature and reliably-updated implementations with a relatively easy and open distribution platform on virtually every modern computing device (browsers and the World Wide Web). Some of the other fundamentals of the language are nice (particularly first-class functions), while the bad fundamentals (particularly the hopelessly messy type coercion) are now well-recognized and hopefully avoided through various techniques.

Probably because the WTF cases are exaggerated. After a couple of months coding JS you will know it well enough so it doesn't bite you. It really is a non-issue for people who work with it daily.

I agree it might not be an issue for professional programmers, given the right tools.

But there are plenty of people who would recommend Python over JS as a first programming language for exactly this reason. Python gives better error messages and has fewer WTFs.

I think JS could be a good teaching language if it were not for the confusion over types, which is an important thing for a beginning programmer to understand.

TypeScript, Flow? That's kind of what you're asking for.

Actually just basic linting.

> why nobody has created a version of JavaScript that just throws exceptions for all the WTF cases?

Google experimented with something they called "strong mode". Classes were read-only, accessing nonexistent properties was an error, and things like that.

I liked it. Unfortunately, the experiment didn't go anywhere and strong mode was removed from Chrome.

Dart is somewhat similar, though. It got more straightforward semantics than JS and most of those WTF things are an error.

Dart has strong mode.

Dart's strong mode is different (it's about using a sound type system) and with 2.0 (or if you're using Flutter), it's mandatory.

https://www.dartlang.org/guides/language/sound-dart

https://github.com/eslint/eslint

Sure this car get dust on the ventilation, and you can mess up if you use the 3rd gear and you can't really use the blinkers without being careful and the seat warmers are wtf.

But in the end of the day, stick to this car long enough and it works for you.

Plus the most popular road of the world only accept this car so take it or leave.

Not everybody can afford a Tesla or a BMW.

For some of us, A Toyota Corolla or a Honda Accord is more than capable of getting the job done.

It is so tempting to poke holes in this analogy but in general I like it!

Hey, Toyota and Honda are good quality cars, did you mean chevrolet, ford or renault?

Hey, I own a Fiesta SE and ST and they are great. Ford quality is way up.. Actually, maybe that's a perfect analogy for javascript :)

That's because they're designed in Germany now.

I've owned two generations of the Chevy Malibu. I'm pretty sure that's what he meant.

I fail to get the flak that javascript faces ! Coming from java it was a breath of fresh air. This was the car that let me drive without a seatbelt when i wanted, let the doors be open if i wanted, stepping on the gas accelerated without fail and the brakes functioned fine, i could change the gearbox orientation, choose which side the steering wheel i wanted to be in general let me do what i wanted to without getting in my way. Sure it was not foolproof and i dont want it to be. I like JS just the way it is and I am grateful to it.

>I fail to get the flak that javascript faces !

I personally don't like that Javascript has had to have a ton of work done on it to get it to the point of other languages that were better at their first release. It was just a ton of time spent on something that was weak to begin with.

If Javascript were better at the beginning, and this same effort was spent on it, it probably would be ruling the world, front and back-end, and be a really good language with great tooling around it.

There was a great amount of debate about nulls being a billion dollar mistake, but Javascript probably beats it in monetary damages by far.

Something wasn’t done right, right off the bat, thats grounds for not liking it, sure. But to dismiss it as immature despite it overcoming most (not all) of those flaws seems unfair.

I like it, some dont. I get it. All I ask is to not dismiss it.

Folks new to programming find these edge cases as a reason to not learn it. That irks me.

It has not overcome any of the flaws, we have just learned to work around them. They are still there and can bite you.

Some of us value our time and would like to improve our discipline. In other fields it's called being professional. It's a shame we care so little about it.

You've got me thinking now. Can you/anyone give me 3 solid cases where the flaw was in javascript and not the programmers understanding of the language ? I can't. (Not at the moment at least). Most of the times I got things wrong it was my understanding of how js works that was at fault.

Of course the flaw is always in the programmer's understanding. Langauges do not manipulate or lie to you, they are what they are.

However, better tools and languages reduce your cognitive burden. The less mental context you have to work with the more you can focus on the problem at hand. If you have to constantly worry about language edge cases and odd behavior, it really detracts from being able to focus on the problem you are trying to solve.

There is a reason languages like TypeScript and Flow exist, they are helping to reduce this large burden JavaScript carries with it.

JavaScript is really great for small scripts (which is the use case it was designed for).

Anything beyond a few hundred lines starts to carry a lot more cognitive load. Not to mention that we are now building full fledged applications with tens of thousands of lines in this language.

> Anything beyond a few hundred lines starts to carry a lot more cognitive load. Not to mention that we are now building full fledged applications with tens of thousands of lines in this language.

Agree completely.

> Anything beyond a few hundred lines starts to carry a lot more cognitive load. Not to mention that we are now building full fledged applications with tens of thousands of lines in this language.

Depends really on how much care was put into the application structure, mostly from the outset.

I've followed opinionated styleguides closely on some projects, and have also walked into other swamp projects that snowballed into something massive from a few snippets of jQuery. The difference between working on something that was intended to be maintainable from the start and something that evolved haphazardly is of course like night and day.

I still think it's more the people and approach than the tools.

I understand that Javascript will silently do something stupid and keep going if I accidentally call a function with an argument left off, but I'd still prefer to get an error message.

I understand that I need to use === instead of == to check if two things are actually equal, but I don't like that either.

It's not that you can't understand how the language works, it's that the language does some things that feel stupid when you're coming to it from any other programming language. Even if I can work around all of them it doesn't feel like I should have to when we have a lot of better planned programming systems, and that makes trying to get into js feel frustrating.

>Can you/anyone give me 3 solid cases where the flaw was in javascript and not the programmers understanding of the language ? I can't.

These were problems in various stages of JS's history, but not necessarily now:

Browser differences in javascript by not having a good specification or strong governing body.

Not having block-level scope despite convention in other C-like languages.

Optional semi-colons.

typeof inconsistencies.

Another problem with JS changing so much is developers picking up the language at various points in that development.

> Browser differences in javascript by not having a good specification or strong governing body.

Its the browser implementation that was at fault. The ambiguity in specs was and is minimal to negligible for the past 8 years (since I started following). But like you said, not relevant today.

> Not having block-level scope despite convention in other C-like languages. > Optional semi-colons.

Its a feature, not a bug!

> typeof inconsistencies.

Maybe. Never faced it.

> Another problem with JS changing so much is developers picking up the language at various points in that development.

What I see constantly changing is JS frameworks and preprocessors, not the language itself. Newer folks tend to learn some JS framework first before the language itself (just like I did) and the constant change can hurt and impede progress and confidence. The language itself has been well documented and forgiving in my experience.

So yeah, point taken. You do not like JS. I do. Let's make peace with that.

"it probably would be ruling the world, front and back-end"

So, what would be different about that? JavaScript does rule the world, front (definitely) and back-end (arguably, but there aren't a lot of contenders for the language that's more popular for new projects).

JS is nowhere near "ruling" the backend. That's just ridiculous.

What language(s) do you believe is more popular and with more lines of new code being written for it?

It gets flak for the same reason php gets flak.

They both more or less accidentally became popular. None of them were very well designed. They both lacked type safety.

Unlike php though javascript became not only popular but it actually ended up being the only alternative.

It's jealousy.

Most languages had major design flaws. Rubyists will gladly complain about how OO is bolted on in Python, Python users will gladly talk about how slow Ruby's interpreter was and how god awfully complex the syntax is. Both of them have limitations in their runtimes and are actually catching up to JS in some areas (see Python's adoption of async).

We can talk about how baroque CL is or how much Java sucked in 1995, how massively complex C++ is, etc, etc. Those arguments have been made for decades, though. JS is no worse than most languages in various ways, but people are pissed off because their chosen language isn't as popular.

Maybe that's a cynical point of view, but these criticisms never get past the most superficial concerns, so it's not obvious why people really care so much.

JavaScript and PHP are badly/barely designed languages, period; no amount of excuses and comparisons is going to change the fact. Anyone with a bit of experience in either Ruby or Python will tear their hair out when doing JS, because they know it doesn't have to be that difficult. These days it's looking more and more like C++ every day...

This is of course my personal opinion, but I think you're suffering from exactly what OP mentions. I'm not trying to be conflictive or anything, but this is in my experience the same argument that every language evangelist uses.

I started with C in my early teens (first book I found, lucky me) and I've gone through the VB, PHP, Java, Python, Ruby, and Javascript phases. Honestly I've seen great engineering in all those languages (maybe not VB) and those projects have all been easy and pleasurable to work with (maybe not Java). I've also seen hideous stuff that's a total sanity destroyer. At least for me the main contributing factor to language dislike is dealing with shitty code. Except Classic ASP... That stuff is poison.

I code in Ruby everyday and I think JS is a far superior language. Javascript has closures, modules and hoisting that complies and stores functions and variables before executing.

In Ruby code starts executing before everything is initialized making it unpredictable. Also JS is just so much smoother is composable. It's far superior to RUBY IMO

I don't think they accidentally became popular.

The features which make a language rise rapidly might just be different to the features that make a language "good".

PHP in particular is amazing for getting started at speed. I remember taking a HTML file and popping in a couple of lines of PHP in the middle. I still haven't found anything anywhere near that easy for making a server backed website.

You aren't the first person to think of this. It is otherwise known as Worse is Better[1]. Stuff that doesn't have all of that nice formalism but lets you get simple applications working without a lot of up-front training and boilerplate tends to be what people gravitate towards instead of "properly designed" languages.

[1] https://en.wikipedia.org/wiki/Worse_is_better

For very limited tasks that PHP is still oddly popular for: NGINX's Server Side Includes module. ;)

If that's already your web server, anyway.

I want my programming language to be predictable and easy to use. I'll "well design" my code, thats what I get paid to do. Type safety, yeah, never bothered me much anyway.

I agree, it sucks that js the only alternative in some cases. But hey, imagine if Java was the only alternative!

That says more about Java than Script though, and I've done enough of that to feel your pain. Coming from plenty of experience with sane dynamic languages (Common Lisp/Smalltalk/Clojure/Julia), it looks exactly like the madness it is. It's a difference in perspective, that's all. It's not about foolproof, it's about having any kind of intellectual integrity at all.

I just wish Javascript had a standarized working module system that doesn't require external dependencies (webpack/rollup etc.)...

It does now (ES2015 modules) but it was standardized too late, so people created their own solutions, and now it's super hard to implement the standard over them. That's why nodejs isn't transitioning to ES2015 modules anytime soon (although they're working on it).

I agree, this post is super-b, but it also shows why JS is the mess everybody knows it is. When you use a function as a constructor, the function creates an object and assigns as its __proto__ the value of the function's 'prototype', which is an object with the class' methods and that has an attribute 'constructor' that points back to the constructor function... and all that... for what? What purpose has this over-complex dance of references? Before trying to justify the current status quo, let's ask ourselves "could it be done easily?"

The answer is a resounding yes. Javascript is a prototype oriented language (nothing wrong with that) who is ashamed of being prototype oriented and wanted to look object (class) oriented, and that's a shame.

And of course, we have ECMA, a committee whose motto is "let's not change any past error no matter how flagrant it is, let's add just another layer of painting over it". That's why we have to live with hoisting and var/let, and crazy type conversions that are an affront to every living neuron in the universe, and typeof which return useless values, and yes, two comparison operators because the first one was worse than wrong.

>It also clearly shows that Javascript is not the mess that it looks like from a beginner's perspective.

Let's agree to disagree. I suspect your bar is too low. You could have made that argument for early versions of PHP (i.e. before 5)

Full disclosure: I've been writing JS for pretty much as long as I can remember and currently do it professionally, and I'm speaking mostly to my personal experience with having used both of them personally and professionally.

The question is how often you get bitten in the ass by it, specifically as a professional rather than a beginner, and in my experience it's been very little with Javascript but quite often with PHP. Once you know a few things like the difference between == and ===, that undefined and null are different, etc., it's pretty easy to avoid the warts because the warts become obvious. I'm not going to argue that Javascript is well designed, it's really not, just that I rarely if ever run into the typical laundry list of gripes in such a way that it has a profound effect on my job.

And having ESLint backing me up doesn't hurt, either. But that's not technically relevant.

PHP, especially the older versions, contained many subtle traps you could easily fall into. For all of its flaws, I've never really felt the same about Javascript. More to the point, beginners are always going to stumble. It comes with the territory. I tend to focus my criticisms more on problems that knowledgeable programmers can stumble into accidentally rather than ones that are commonly met by new programmers, if that makes sense. The former seems like a far worse problem, because the former starts to touch (bear with me) "real" code.

Just as an aside, the biggest single pain point for me when I still used PHP (a lifetime ago, now) was the way PHP used to report missing quotes extremely inaccurately. Egads that was an awful bug. I hope they fixed that.

What it boils down to in my mind is, when I'm writing JS, I'm comfortable not having a linter available. I don't always feel the same way with PHP, because I don't feel like the traps are as easily avoided. I certainly don't think I'm some sort of Javascript guru, it's just that I don't have many problems with it in a purely practical sense when compared to PHP.

I think this is a fair way to read it. I have made my bones on ripping PHP and JS up and down, left and right; I didn't trust them as far as I can throw them. SPAs sucked because JavaScript sucked, and PHP sucks because its existence has put more frustrating work on my plate than anything else.

I still feel that way about PHP. ES6 and ES2017 changed my mind about JS, though, along with the eventual maturity of the stack: babel, eslint, good testing in Jest. Everything I always hated about JS totally does still exist in its bowels, that's never going to change. But the guard rails have gotten very good and more modern flavors of JavaScript have added a lot of stuff that makes life easier (strongly opaque modules, async/await, and arrow functions are probably my top three). The build ecosystem is still bonkers, yeah--but there's a baseline level of sanity that I can work with. I know there are still land mines in there, but they're flagged and mostly out of the way.

On the other hand, any time I'm stuck back in PHP, it is fundamentally the same stuff that has been sticking needles in me since I was in high school. Yeah, the language has advanced--but I'm not sure the general practice has advanced all that much, and the parts of the language that are problematic are right up in your face.

In my case, my problem is that I like clean languages, languages whose mental model I can keep in my mind. I found "== is broken, use === instead" a terrible way of designing a language. It should have been "== is broken, let's fix it today so tomorrow will be better for everybody."

I can program in JS, and sometimes I even enjoy it. I had a friend who liked snakes and had several as pets, even some poisonous. I prefer kittens.

It's hard for me to take these comments seriously unless you feel this way consistently across most other dynamically-typed languages -- at least it'd be consistent.

I've used most dynamically-typed langs at this point and I've run out of reasons to use something aside from modern Javascript if I would have originally used Clojure or Ruby in the past.

Though it feels sheepish to reward your low-effort comment with a response.

Curious, for server side, why do you prefer JS over Clojure? (I assumed by your comment that you would still prefer modern JS already knowing Clojure. Maybe I misinterpret?)

Right, I would generally prefer JS over Clojure on the server side these days. Or, rather, I would default to JS and then possibly disqualify it based on my needs (like team composition).

I spent over three years using Clojure and eventually learned how unimportant technical superiority is for most of my projects. For example, the first time I met a co-founder and they simply weren't interested in using Clojure. Well, that was easy.

JS has major upsides like ubiquity and I'm-already-using-it-in-the-browser and its simple concurrency model that aren't so easy to paint over.

I'm not saying everybody should default to JS like I do. But I think a lot of these avoid-JS-at-all-costs posts need to update their intel. For example, JS has pretty great static analysis compared to anything I've used in the other dynamically-typed languages. It's also one of the few where gradual typing actually caught on.

For all its upsides, its warts aren't any worse than things like the overuse of metaprogramming in Ruby or the aversion to FP in Python or the tooling dependencies of Clojure.

I still find a lot of reasons to use Ruby, though like you I spend a lot of time doing JavaScript in places I wouldn't have a couple years ago. (For a long time I've been anything-but-JavaScript, but I was also using JavaScript before it was cool--somewhere on a hard drive somewhere is a JavaScript implementation I wrote as a scripting layer for a game I was working on in like 2006...)

My experience with API design in JavaScript is that it still kind of sucks. It's not the worst, that wonderful spot goes to PHP, but it's distinctly harder to get something usable and maintainable done than Grape/Ruby. Probably worse than Flask/Python, too. The Swagger tools out there for JavaScript aren't very good and none of the lesser-used alternatives seem any better; maybe it's the stateful/metaprogramming-friendly nature of Ruby, but `grape-swagger` is really nice and nothing comes close in JS, at least not that I've found.

Data representation worries me, too. I have not yet found a decent `grape-entity`/`representative` library in JS (think a `grape-entity` or a `representative`). On top of that, SQL library support out there is pretty bad, Sequelize being the best I've found but much worse than Ruby's Sequel or even ROM. Sequelize is still frustrating and clunky.

I'm also uncomfortable dealing with money and financials in JavaScript, in part because of the lack of a BigDecimal/fixed class that I feel like I can trust (as I'm not really qualified to judge the implementations of the billion gems out there). A standard library I feel like I can trust is still out there, somewhere.

If all I have to do is plop out an endpoint, then JS (shouts, Express!) is as easy as it gets, and that's fine. But these days I need more, I want my tools to help me with it, and I don't really have time or inclination to build them, either.

If I'm missing any that are worth checking out, though, by all means, I'm very interested.

Money comes in integers. Any function that works with monetary values needs to eat integers and output integers.

This is not pedantry, it's a real-world constraint.

I thought your "real-world constraint" existed too. Then I had to keep track of thousandths of a cent.

When you have experienced the "no, a cent is no longer granular enough" problem, the conversion of DECIMAL(7,2) to DECIMAL(9,4) and the attendant code compatibility is something you'll appreciate.

That's what fixed precision types are: integers + a scale factor -- whether it's cents, mils or something else.

Javascript does badly need a native fixed precision type beyond "everything is a double" or some web assembly hack.

And I like Javascript, but this is a real problem.

I don't like ORMs or abstractions like grape-swagger. I like working with the node-postgres client and knex when I need a query builder.

It sounds like if that's the sort of kit you want, then Ruby is for you compared to Clojure/Javascript.

What I do like about Javascript is that it's easy to write ad-hoc types via Typescript for my SQL functions to specify the shape of the data that comes from the database. I would never go back to something like Active Record that tried to make me canonicalize all of my data.

I have to wonder if these are more identity-level ecosystem differences.

I'm very confused as to why you don't care for "abstractions" like grape-swagger. Do you enjoy writing API client boilerplate?

As far as ORMs go, I'd certainly prefer not to use Active Record ones, but I also don't want to have to update my code when somebody else adds a field that I don't care about. My representations specify what I care about and how it should be presented to clients (for example, hashids rather than integer primary keys); I update them only when there's a new field that I care about. I can do this with functional mapping but the database boilerplate sucks more than Sequel does (it's a big part of why I've dropped ROM, though I much prefer it in theory). Computers exist to help me do less programming, not more, and what you describe is more programming for, to me, little benefit.

Reply on news.ycombinator.com