Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: My new JavaScript MVC framework (lhorie.github.io)
304 points by lhorie on March 18, 2014 | hide | past | favorite | 159 comments



I've become a huge fan of simple things. When I was just starting out developing software, I was a big fan of huge integrated solutions. But as the years went on I loved well scoped, lean options.

This is the reason why I like Backbone.JS instead of Angular. The reason why I like Go instead of Java.

Mithril looks really promising! Don't let feature creep turn it into a behemoth! Keep it lean and mean, and excel at the one thing it was made to do!


It's the power and pleasure of understanding. It makes your mental model of the project a closer approximation of omniscience.

The cycle makes sense. A new programmer finds everything overwhelming, and the simplifications offered by something like Mithril often aren't enough to mitigate that feeling in any meaningful way. It's going to feel like you're drinking from a fire hydrant no matter what.

But as you get a much firmer grasp on the fundamentals, you also gain a much finer perception of bloat and cruft and what's necessary and what's not. So you feel a push to eliminate large all-encompassing libraries, because for the first time you realize that some sort of omniscience is within reach.


And how much net current/future amortized cost (time + maintenance + pain) / benefit there is between DIY & using "off-the-shelf" + learning/configuring/donkey-punching


I think this is the same taste-maturation that many woodworkers experience. At first, you salivate over these crazy, overly specific, over engineered power tools. Over time, you start to really love things that are stupidly simple instead, like a card scraper, or a hand plane.


Cooking is the same way.

A few, simple, fresh ingredients prepared quickly and correctly can be much more amazing than a complicated 20 ingredient recipe that takes 2 hours to prepare.


I would buy your wood-things.


Go vs Java? Choose the right tool depending on your experience with either and the goal of your project. They're really different beasts. Go (2007) focuses on concurrency, is relatively young, and is compiled to machine code for native speed. Java (1995) is object-oriented, has thousands of libraries, and runs on JVM. For the same project you probably won't consider both of them at the same time.

Backbone vs Angular? Backbone (without 10 other plugins) is just a small handful of prototypes so you will need to write huge parts of the actual application framework yourself (NIH syndrome). Not really comparable to a fullblown framework like Angular or Ember where you learn their workflow to take advantage of all the features and to save 'technical' time to write more 'business' logic. So again apples and oranges in this case until you mention Thorax and that kind.


Hi, Mithril's author here. I'll just put a big comment here, and hopefully everyone can see it.

@stronglikedan: I don't have plans for extending the core (in fact, keeping it small and modular is a major focus point for me). I do have a list of things I want to tackle next (see the roadmap page), but I'll most likely release them separately from core.

@hcho: re: integrating w/ jQuery: see the integrating w/ other libraries page in the guide section. There's a simple example w/ select2 there.

@abjorn: those are excellent points. Re: turing completeness: my take is that things like good error messages in the view layer are more important than trying to prevent people from doing stupid things (that's what code reviews are for). Re: Bindings: I'm most familiar w/ Angular ones and yes, their bidi-bindings are really convenient, but they fail in some 5-10% of the use cases for me (either by being too aggressive or not expressive enough). So, that's a conciseness vs power trade-off in design from my personal experience. Re: templating, you can take the model-level utilities and integrate w/ other templating libraries. I do provide some comparisons w/ React and a few other frameworks in the misc section of the guide, as well as design rationales in the main guide page. TL;DR I use other frameworks full-time and I've done homework before I settled on the current implementation :)

@hanburglar I do provide a tool to convert HTML to Mithril (although not automated yet), see the "useful tools" page.

@timmiwil: you're right, jQuery is not MVC (I mention this in the comparisons page). The point is that with idiomatic jQuery, the app developer is responsible for knowing when to use .text() instead of .html(), whereas with, say, idiomatic Angular, that's not a concern, ever. jQuery "templating" tends to get pretty hard to audit as widgets become more complex (see select2 source code, for example)

@tzaman: I do contribute to Angular and other projects that I use as time permits (mostly bug reports)

@BaconJuice: it's a side project, but a scratch-an-itch one that I work on pretty much every night. Planning on continuing work on it for the foreseable future. I'm also considering introducing it at my day job as well.

@all: thank you for the feedback, I really appreciate :)


This is kind of an annoying format for answering questions. You can't see the context and it breaks the conversation thread.


  >  thank you for the feedback, I really appreciate :)

Well, thanks for the framework :)

Apparently, you missed my (now deleted) question:

  >  It would be nice if you could document the browser requirements.

We're even, since I had missed the answer, which lies at the very end of the misc section of the documentation:

  >  Mithril allows developers to support browsers all the way back to IE6 and Blackberry.

Wow! This should IMO be on the home page.

    ----    ----    ----
Edit: Is it possible, for SEO, to load Mithril with a pre-rendered page and have it take over like React?


>> Mithril allows developers to support browsers all the way back to IE6 and Blackberry. > Wow! This should IMO be on the home page.

Agreed! I'm in!


You've missed the point. You're comparing MVC frameworks to jQuery as if that made sense, which it doesn't.

As I said, Angular and Backbone use jQuery. So you're just comparing how you've used jQuery (as if that's how any jQuery user would do it) to how they've used jQuery. There is no such thing as "idiomatic" jQuery templating. There are only plugins and libraries that use jQuery to do templating. In fact, what you've done in the jQuery rendering test is incredibly bad practice and is greatly discouraged in the jQuery community. Again, your tests don't make sense.

If you need me to explicate my pointer further, the jQuery "security test" is you asking jQuery to do something stupid and it following orders. It's absurd. You might as well add a native test saying that document.documentElement.innerHTML = "a\"><img src=\"javascript:;\" onerror=\"alert('alert box should not appear')\">"; has a security hole. How dare the browser do what I said!

Stop comparing jQuery to MVC frameworks.


Many people are at a stage where they're using jQuery to build large apps and wondering what benefits a framework provides. Not having to worry about script injections is one of the benefits that frameworks bring to the table. This is a perfectly sensible thing to point out, imho, given both tools are used for the purpose of building web applications.

The technical implementation of the jQuery test is flawed on purpose, and that's the point. People make mistakes, some don't know better.

Are you arguing that every jQuery dev team in the world audits their code for stuff like:

  - bla.text(firstName + " " + lastName)
  + bla.html(firstName + "&nbsp;" + lastName) // bug fix 1234: make name not wrap
Or that somehow the jQuery community is gonna catch that in your application? Or that your innerHTML example (given a user-defined string) does not have a security hole? That you're gonna find that commit from the junior guy at 3am at crunch time? That "the developer should know better" is a good security strategy? Are you saying that one API that makes it hard to make mistakes not comparable to one that makes it easy to? That's the absurd claim, imho.

Like it or not, jQuery is the big elephant in the room, so yes, I'm going to compare to it, to the extent where it makes sense. Is it an apples to apples comparison? No. Neither is the comparison w/ React. But those comparisons are useful to some people, so I provide them.


Sigh. I would hope that you were not confusing my statements to mean that I don't think jQuery can be used badly (by the way, having seen jQuery used poorly is no justification for your tests) or that I don't understand the security hole presented. I want people to take responsibility for how they use tools like jQuery and I think your tests are an unfair and blatantly misleading representation of jQuery as a library.


Now, having gotten that off my chest, I'm not going to go off and disparage your framework. I want to thank you for your library and your contribution to open source, because I think that has eminent value often overlooked.


> your tests are an unfair and blatantly misleading representation of jQuery as a library

Sure, that's a fair point given that my explanation in the homepage is skimpy. I did try my best to word it in a way not to diss jQuery as a project: "if you see an alert box, ensuring security with that framework is more work for you". If you have a better idea on how to phrase this point, I'm open to suggestions.


> Many people are at a stage where they're using jQuery to build large apps

Nobody's using jQuery only to build large single page apps,that's not true. Comparing jQuery which a DOM manipulation library to any framework that manage the application lifecycle is dishonest.

It would be like comparing underscore to AngularJS,makes no sense.


Hmm, I think you're underestimating legacy code. I can name two very large projects in my company that do use jQuery as the main client-side library, as well as numerous other smaller ones (granted, these projects are not SPAs).

Back to the point, jQuery and Mithril are more similar than your analogy tries to make it sound. jQuery does AJAX/CORS and Deferreds and .data() and a whole lot of non-DOM stuff that people use when building their apps. There's some stuff that doesn't map like jQuery animations not having a Mithril counterpart or Mithril routing not having a jQuery-core counterpart, but that's cherrypicking the stuff that is different and saying "look, they're different!", while ignoring the similar, comparable aspects.

Also, I'm not sure how familiar you are w/ underscore or Angular, but both have a templating system and utilities to work with lists of data, so there are definitely points that can be compared. These comparisons are really not at all like comparing something like restangular to hoverIntent, so implying that they're equally imcomparable seems like the dishonest and non-productive argument to me, imho.


Can I ask how long you've been working on it before releasing? Your github only shows it was pushed two days ago.


Author of Vue.js here, I'm glad you included Vue in your framework comparisons. Here's some thoughts regarding your comments on Vue.js: the reason Vue chose to use ES5-only features is that it enables Vue to provide plain POJO syntax without having to resort to dirty checking or virtual DOM diffing. Granted the auto-magical POJO syntax is a somewhat leaky abstraction, but I'd argue that not all leaky abstractions are bad (since by definition any non-trivial abstractions are leaky) - it all depends on its leakiness vs. the benefits provided by the abstraction. TCP is the canonical example of a leaky abstraction in Spolsky's original post yet it has been the foundation of everything we build on the web all along. In fact, even the concept of virtual DOM is a leaky abstraction in itself. My point is, I don't think the leaky abstraction argument raises anything inherently problematic about Vue.js.

That said, I really like the project because I'm also a big fan of simplicity. Curious to see if there will be a TodoMVC implementation with Mithril for easier horizontal comparisons of actual code.


There is a partial TodoMVC implementation available here - https://github.com/jpmonette/todomvc-mithril


Have you successfully tried this out in a real app? Why would you want to structure your dom inside an array? Seem pretty crazy, my dude.

todo.view = function(ctrl) { return m("html", [ m("body", [ m("input"), m("button", "Add"), m("table", [ m("tr", [ m("td", [ m("input[type=checkbox]") ]), m("td", "task description"), ]) ]) ]) ]); };


I think the idea is that it's simple — it's just a basic data structure in the language and can be generated or manipulated with just simple JavaScript. This is similar to why Lisp programs being Lisp data structures is interesting. Intriguingly, it looks like the API is compatible with React's JSX. With that, you could write this as:

  @jsx m
  todo.view = function(ctrl) {
      return (
        <html>
          <body>
            <input/>
            <button>Add</button>
            <table>
              <tr>
                <td><input type="checkbox"/></td>
                <td>Desk description</td>
              </tr>
            </table>
          </body>
        </html>
      );
  }


That is ugly as heck - for some of the complaints that some people have with Angular's templating, it doesn't force you into tying your view so heavily to JS. That is far more of a travesty.


Adding some more css declarations, id, data-hooks, etc., and make it span 100+ lines, like in real life. Scares me just thinking about it.


Are those things you wouldn't have to do under any other system, or which would be significantly simpler there? Because I'm not seeing it. You seem to be saying "This involves building a web page" — which is true, but would be regardless.


Yes, it would be simplificantly simpler to edit html if it wasn't stuck in between two strings, for one, let alone inside an array.

todo.view = function(ctrl) { return m("html", [ m("body class='application", [ m("input class='add input input-sm input-md"), m("button class='btn btn-success button-green", "Add"), m("table class='table table-strped table-row", [ m("tr class='tr'", [ m("td", [ m("input[type=checkbox] class='checkbox input input-sm' id=firstName name=firstName") ]), m("td class='td", "task description"), ]) ]) ]) ]); };

I honestly can't tell you what the hell that is supposed to be anymore. Also, you guys are using invalid html for everything. Your example is not going to validate on anything.


I think you're too caught up on syntax and missing the big picture. The m() method is simply a helper utility that returns a POJO. You can write the code in coffeescript if js syntax bother you, or you can go nuclear and implement a frontend or a preprocessor that take some popular templating syntax and spit out the appropriate data structure. The conversion tool I provide is a small step in that direction, but, with Mithril being at a mere v0.1, it still needs more work done to be fully automated.

Also, just FYI, the selector syntax follows CSS rules, so you'd write m("body.application"), not m("body class='application'")


Programs are for humans to read and only incidentally for computers to execute. Syntax is the whole picture.


Well, I already told you which directions Mithril can go in terms of making syntax more to your liking. If you're too put off by js templates and not willing to spend time helping push the project in a direction you like, you can always try other frameworks or come back later when I have a more HTMLy frontend option available.


    > Yes, it would be simplificantly simpler to edit html 
    > if it wasn't stuck in between two strings
    > ...
    > m("body class='application", ...
You would put attributes in an object.

    m("body", {class: "application"}, [ ... ])
Sure, it's no Lisp, but to me it's just Javascript's commas that suck the thunder out of it.


This is exactly why JSX is such a big win for this method of creating content - it's just sugar for the same function calls, but expressed in a form we're already used to multiply-nesting, with closing tags to help keep context and no need for commas when you need to break a line.

Quite neat that you can use it for non-React stuff just by tweaking the pragma comment the JSX transformer requires.

Here's an actual example of a Bootstrap table in a React component using JSX: https://github.com/insin/lifequote/blob/master/src/component...

There's not much display logic in it, so it doesn't exhibit the usual progression I've seen when writing React components so far, in which most display logic happens up-front in a component's render() and often gets moved into another method and called directly from where it's going to go in the structure being returned from render(), so you generally don't end up with huge whacks of JSX in components which are doing more work.


To be clear, my example wouldn't actually work. I had a brain fart about JSX's output format. Sorry about that. I do think it would be pretty easy to adapt it to this format, though (it looks like you'd just need to change one line in the visitReactTag() function).


You could actually do this without a single change to the JSX transform, and instead just make a simple object that calls m() with the right params. Then you just change the @jsx pragma to point at that object. With a complete array of elements this should mostly work.

```js / @jsx m.dom */

function m(type, props, children) { // assuming m() looks something like this. return { type: type, props: props, children: children}; } m.render = function(lwDOM){ console.log(lwDOM)}; m.dom = {};

// React hard codes this array so we can do something like this to build the mapping ['a', 'span'].forEach((el) => { m.dom[el] = function(props, ...args) { return m.call(null, el, props, args); } });

m.render( <a href={"google.com"}> <span>hello</span> <span>world</span> </a> ); ```


While I admit I'm somewhat sympathetic to this reaction, it really is a cosmetic complaint. It's hardly uglier or more complex than ordinary html -- we're just used to html. Ofc there's some benefit to the technologies we're used to all things being equal, but I think architectural benefits trump concerns like this. I'm not making a case for or against Mithril, btw, just noting that this objection is a fairly surface one.


How many times have you messed up a comma inside an array? And just for looping, this is literally constructing a web page out of it. There is no architectural benefit to this, all that's going to happen is that your going to fat finger some shit, and the compiler will complain to no end.

Guys, rule of thumb, architecture === directory names. Performance is an implementation detail.

Frameworks live and die by their APIs. The easier it is to work with, to read and write and reason with, the better it is.

But you know what, this can be easily proved. Someone take a stab at todomvc with this and let me know if it hurt or not.


There's a sample TodoMVC posted here - https://github.com/jpmonette/todomvc-mithril. If you have any idea to improve, feel free to send a pull request!


Looks roughly how I'd expect. Personally I don't find it unreadable, except for the 8-spaces-in-a-tab thing (which is not really relevant anyhow). Anyways, few things I see:

- I'm not sure why you're doing `return m("#todo-app"), [...]` on line 15 of the view, it looks like a typo

- I would probably have written a keyboard utility function similar to m.withAttr to keep the e.keyCode stuff out of the controller and remove the anonymous function in the view

- I generally prefer inline ternary syntax instead of a if statement at the top for the "clear" button.

- the pluralize snippet can also be pulled out into a utility method

- I also tend to favor this syntax: m("input#new-todo[placeholder='What needs to be done?']") if the value of the attribute is static (line 18)


https://github.com/jpmonette/todomvc-mithril/blob/master/js/...

Looks pretty painful, but I do think adapting JSX could be a pretty painless solution so I don't look at it as a deal breaker. I am still yet to find a javascript templating solution that I don't have at least 1 or 2 issues with.


Here's a JSXified version which assumes the existence of an "mshim" object which adapts tag calls and arguments to what m() expects:

https://gist.github.com/insin/9639918


Honestly, that looks worse than I thought it would. Are you guys familiar with the whole, programs are for humans to read it, and incidentally for computers to execute?


Overall I like it, except for the templates.

It's perfectly possible to not have FOUC with traditional template languages like mustache, and I don't think having "turing completeness" in your templates is a good thing. Frameworks like Ember.js have also shown that you don't need to manually write bindings to handle when your models change.

That being said, I do see the advantage gained with the virtual DOM you generate from the templates. I'd be interested to see if it was possible to integrate other templating languages in via plugin.


Entirely agreed, this is what I was about to say as well. Being able to use html templates for me makes apps much less confusing and much more organized. Moving your html into your javascript just doesn't seem right. If there was a way that it would accept a mustache/underscore/other precompiled template I'd be all about it. But constructing html out of javascript functions and objects just doesn't sit right with me.


I actually had the same thought, but I also get that some people don't like HTML-based templates for various reasons. My next thought was whether or not you could just augment this framework by providing a utility that converts an HTML-based template into the view that mithril expects and make both camps happy. As an optional contrib module, of course -- I'm not suggesting he start heaping on additional framework features already. :)

edit: I just wanted to add that I read through the guide expecting to find yet another half-baked framework but the whole thing seems well thought out and I like the philosophy alluded to in the guide. I'm definitely going to try this for my next mini-project.


Well, there is such a tool, actually. http://lhorie.github.io/mithril/tools/template-converter.htm...

But while me and my coworker have been trying it out there seem to be quite a few bugs with it, and I'd much rather have it done at runtime (precompiled for production build) then have to translate it from my HTML-based template every time I make a change.


Can you file an issue on Github? I still need to make an automated version of this tool (and eventually be able to import from something like Mustache), and I want to make sure it's rock solid.


Well, that's a good start, but I was referring to a 'utility' in the sense of a utility method that does it at runtime, as you say.


Definitely, and there are other benefits in parallelizing the work stream if you have people that do layout and styling, but aren't necessarily up to constructing a page in a javascript framework.


>converts an HTML-based template into the view that mithril expects

So basically XAML?


The tests don't make sense. jQuery is not an MVC framework and should not be tested as if it were. It's worth noting that both Backbone and Angular use some version of jQuery (lite or otherwise), so comparing that to your own way of using jQuery doesn't correlate. I could use jQuery to append elements in a way that was much faster than and just as short as what you've done in your test. Also, did you know you can append image elements with events using document.appendChild? It must be a security issue! But seriously, don't even include it in the tests. It is a JavaScript library that serves a lower-level purpose than MVC frameworks.


It has reached the point where Javascript Framework coders don't know Javascript.

Looks like Javascript IS the Assembly of the Web.


But DOM is separate from JavaScript.


DOM is separate from Ecmascript, but is actually one of the 3 components of Javascript (along with Ecmascript and the BOM). This is at least according to Professional Javascript for Web Developers (you can see it on 'page 3' in the Amazon preview).


The sources I've read (Douglas Crockford or Wikipedia) clearly state that DOM is separate from JavaScript.


Only theoretically.

Javascript is a way to manipulate the DOM, no matter whan those nutters who run it on servers say.


Nope, it's the truth. You can have JavaScript without the DOM in a Node.js or a CouchDB environment, for example.


I would say that what node.js runs is ecmascript with it's own API that differs from the DOM. It's called javascript for convenience, but then what separates javascript from ecmascript?


Nothing. It is only different in name. Oracle owns the actual copyright on JavaScript via Sun. ECMAscript was used so the ECMA would not have to worry about this.


I guess this is a point of contention. Javascript is considered by many to be an implementation of ECMAscript that also provides additional features (the additional features, according to the book I cited, being the DOM and BOM).


I consider myself a JS-person, but if I were honest, I speak jQuery/backbone, not pure Javascript.


The approach we've taken in our anti-framework is that the model is your data (and the code you use to manipulate the data), the view is your HTML/CSS (not a special templating language -- actual HTML), and the controller is automatic for simple stuff and custom for complex stuff.

So the big example at the end of Mithril would, for us, be something like (we implement binding as a jQuery extension)

$('.display').bind(data);

Where .display selects the root node of the bound UI, and data is our object.

Which is simpler and less code, I think. (Oh and our binding library has jQuery as a dependency, but is sub 3kB minified and gzipped.)

That said, we haven't put our libraries on github yet :-(


Sounds a bit like the pure js library but with jquery.

http://beebole.com/pure/


Ours is a fair bit more elegant and works both ways. It's also being used for multiple pretty complex projects so it's reasonably "battle-tested" (although nothing is in production yet, so not as battle-tested as it needs to be).

We also do not bind by class because that's a really bad idea. (Actually the way we bind is a lot like Angular, but without being a templating language.)

A key design requirement is that the model does not get polluted with special methods and properties to support the binding -- so (for example) if you have nice RESTful services, you can GET, bind, edit, POST/PUT and everything just works. For moderately complex cases we support "decorators" that live "off to the side" (i.e. do not pollute the model). Again, you can stick getters and setters in your model if you want to, you just don't need to.


You're not using actual HTML, you're wrapping dom elements in your own special functions. You've made a templating langauge, like it or not.


Am I using code to do binding? Yes. Do we bind to HTML or something that -- after being parsed -- turns into HTML? HTML.

The advantages are significant. First -- you have to know HTML. Second -- all the code that works on HTML works on HTML.

We use data attributes to make the binding work. The data attributes are a "template language" if you will, but it's not breaking HTML (and if you read our attributes they make sense without much documentation).


No, if you're compiling at some point into HTML, then by definition, you're not writing HTML. HTML doesn't have functions in its spec (function m()) That's breaking HTML right there.


I want this


your "getting started" page is like a crash course in building websites with javascript - just forwarded it to a bunch of friends

it felt like you were giving a smart person a crash course in javascript

wish more apps sturctured their tutorials like this

great job!

One thing: it feels like it's aimed at people who are already competent at javascript

but your tutorial isn't too many steps away from a being solid "javascript for absolute beginners" guide

So would be cool if you made it slightly more beginner friendly, like I'm sure you're losing a lot of engagment and users because they read the first line of your guide and immediately think it's too complicated

"Mithril is a client-side Javascript MVC framework, i.e. it's a tool to make application code divided into a data layer (called "Model"), a UI layer (called View), and a glue layer (called Controller)"

Like I'd suggest adding a VERY simple sentence about why you'd want to use javascript AT ALL

both on the guide and the start page

like if someone landed on your page and they were good at HTML and css but never really programmed, they could probably use your framework to build their first programming project if you made it a little more beginner-friendly

great work! :)


why don't you

use some punctuation and

not separate your thoughts out on different lines?

Not trying to be mean, this is just hard(er) to read!


Sorry - hadn't planned on making my comment that long but will definitely keep your feedback in mind next time I reply to a comment.


Not

astroturfing

AT ALL


The framework is completely free and MIT-licensed, so I suspect that a literal interpretation of your comment is absolutely correct.


I am really interested in this, and am going to give it a go on my next project. Its areas of responsibility are contained and well-defined, the syntax is nicely concise, and it really feels like plain old javascript, which gives me the comfort I can get under the hood if I need to, without massive conceptual/abstraction overload.

If I am iffy on anything, it would be the templating language, but I suppose a React-type HTML syntax could be optionally layered over top without interfering with anything else. And there is something secure about javascript-rendered HTML in that you are much less likely to have unclosed tag issues, etc.

But for me, having experimented with some of the slower performers like angular, performance is a huge draw, and I am willing to write my templates in js if that's what it takes to get it.


Here are my thoughts, but you won't like them. Stop with the frameworks. Learn Javascript. Learn CSS. Learn HTML. You'll find pretty quickly that what you need are libraries, not frameworks. Things like jQuery, underscore, etc...


    >Stop with the frameworks. Learn Javascript. Learn CSS. Learn HTML. 
That's good advice to someone that might be asking about using a framework. But to someone that has written a framework they probably have a pretty good understanding of Javascript already, and if they didn't it will have improved greatly by the time they are done.


The Stop with the frameworks idea still applies doesn't it?

I don't code in JS. Maybe a few lines once in a while when it pops up. It just seems like an MVC framework is the javascript equivalent of writing your first recursive factorial function.


A developer I worked with several companies ago told me the same type of thing about learning OO "Its not necessary just learn to code better".

The fact of the matter is large code bases need to be supported over time by diverse groups of people and frameworks help enforce standards that get people up to speed quickly.

That developer I knew, he has been chronically unemployed for the last 5 years because he lacks the skills to work as part of a team. Great coder though!


I second that. I have met a number of devs who are not idiots, have good analytical thinking, but refuse using frameworks, libraries, etc. The result is unmaintainable code base, which no one wants to maintain, and re-writing is too expensive so the business owner is, well, fucked.

The majority of the Web apps is CRUD + a bit of logic. It makes much more sense to use a framework because 1) most of the things are already there 2) easier to maintain for someone from outside 3) and, most importantly, very few developers have skills to create a nice, maintainable design, accompanied with a useful documentation.

Yes, in ideal world, maybe you should be able to stay framework-less (just made that word up) but the world isn't ideal. We have to sacrifice philosophical ideals for the business because that's what pays the living.

In rare cases, when the app is much more than a simple CRUD, it might make sense to ditch frameworks... But it's very, very rare.


Sorry, he's not a great coder. I think there is a big difference between saying "just learn to code better" and "use the right tool for the job" (which was the intent of my post).

I'll add an addendum to my post that frameworks can be the right tool for the job. But 99% of the time they're used the wrong way and for the wrong purpose.


>Stop with the frameworks.

>Things like jQuery, underscore, etc...

Sounds great until you have a team of 5 developers maintaining at least a semi-huge code base over the course of some years. Some people leave, new join the team.

Then you really start appreciating frameworks if only for their conventions to structure the code. Top coders can write web apps with jQuery and underscore alone, but everyone else is better off following some conventions.


HTML, css, JS (jquery) have stable APIs, when new developers come they don't have to learn something new.

If your code degrade over time, sorry to tell you, a framework won't solve the problem.

If you have coders that don't care about maintainability of code correction, code organization in whatever technology, they won't care about what you wish.

As usual, they won't take the time to learn the frameworks, will diverge, and you are back to square one with no code that can be salvaged or reused. (at least with vanilla techs you may salvage stuffs)

I think you use frameworks to solve the wrong problems here: you are probably the problem by trying to ditch the lack of real engineering skills in IT (not the letter of engineering. The spirit: making things that are correctly built and are reliable over time... like a plane).


Well I think we'll just have to disagree here. I think you are patently wrong if you really believe you can keep up maintainability of a largish codebase by appealing to craftmanship in a real-world company.

This is all nice and well in academics and start-ups but some years down the road, your code is going to be a mess. Frameworks, especially opinionated frameworks are a tremendous help. Coding in jQuery or native JS without some prodding is at best an accident waiting to happen.


In my experience, the only reason the code is going to be a mess is because the company hired a bunch of junior developers and gave them "Architect" titles. I can show time and time again that with a team of seasoned engineers the code not only isn't a mess but in fact there is less code that does more, and does it more efficiently.


Actually, this is kind of the spirit of Mithril. It's really a tiny collection of standalone useful functions that also happen to work well together. Ironically, your examples of "libraries" are both quite big in the context of the "lean learner" mentality, and in comparison to Mithril.


That's similar to my viewpoint. Stop with all the frameworks - what new, powerful functionality will it provide your users that existing things can't do (ie JQuery, raw JS)?


Faster development, potentially less bugs, conventions, etc. All of that is highly dependent on what your projects needs but I can't imagine writing some of the Ember apps I've worked on in just plain jQuery while maintaining a clean and readable code base.

You eventually end up writing your own framework of sorts anyway when you take minimal path.


Frameworks add to codebase maintainability, not to user experience.


the point of a framework is speedy development. less lines of code for results. the tradeoff is often bloat, which one should (ideally) refactor out later. Angular can get overwhelmingly slow if you're not careful with your code, but gosh dang, you sure can whip an app up pretty fast


For most apps, I've found that slowness is not a problem with Angular. You can do some nasty things though if you're not careful though, definitely agree with that. If you have flaws elsewhere in your codebase, including the backend, it may be possible that in practice, the effects may trickle down to your frontend code with Angular - it has happened in an app I've worked on with a couple of other developers, in which flaws with the api performance caused us to implement some subpar workarounds that triggered many $digests & forced us to spend a day or two on just optimization.

Angular can be quite fast compared to jQuery in an app of substance. It is not constantly reading/writing to the DOM in most cases, which saves a lot on the performance front. It is also a smaller library than jQuery, and you're not naively applying global or complicated selectors. Plugins rewritten to be pure javascript + Angular also can have a much smaller footprint & be more powerful to boot, such as the Angular UI Bootstrap project (5 KB minified & gzipped as opposed to 15-35 KB? I forget Bootstrap's js size) & contains all of the power of Angular to modify your functionality with them & more.


> bloat, which one should (ideally) refactor out later

How do you refactor out bloat from a framework without throwing away the framework?

This is my whole point about not using frameworks. They're a siren that WILL eat you in the end.


I don't like rewriting boilerplate code for every new project-- as I see it, Angular is to jQuery what jQuery is to raw JS, just in terms of reimplementing basic functions for the nth time.

Of course, if a framework automates a ton of functionality, it'll naturally be inappropriate for certain projects. I wouldn't try to use Angular for an HTML 5 game, for example, though there's a good chance I wouldn't use jQuery either.


> I wouldn't try to use Angular for an HTML 5 game, for example, though there's a good chance I wouldn't use jQuery either.

I actually did use Angular for an HTML5 game last week as part of a demo for a conference. Admittedly it was a really simple game (tic tac toe), and it did use Polymer as well as angular, as sort of an experiment. But it worked pretty well, and only took about a day and a half to hack together, with realtime multiplayer and chat using socket.io.

Obviously you're going to run into things that Angular is not really ideal for, but we are trying to make it do what it does very, very well, so in the future it should be suitable for pretty much any mobile or desktop app.

---

This sounds like evangelism, which is not my intent, you're absolutely right about what you're saying, but it's true, sometimes a framework works (and saves a ton of time) that you'd otherwise spend hacking together something awful. Hopefully this mithril thing also saves people a ton of time without them having to worry about numerous other issues.


Ehh I know Javascript, and know it well, along with HTML+CSS. My experience is that a very light framework can not only help structure things in a more sane way, but can provide a set of tools that make a mindset easier.

After building/using Composer.js (http://lyonbros.github.io/composer.js/) extensively I inevitably end up reimplementing parts of it on projects that don't use it because it makes some things so easy.

That said, if you're learning javascript, stay away from frameworks at all costs.


When I saw the link title, my first thought was "why do we need yet another new MVC Javascript framework"?


That's sort of like the tone Microsoft took after releasing ie6. Landscapes change. Every single js framework in existence today will be unfit for purpose in 10 years. So we enhance the frameworks we have and try out new ones so in 10 years time we don't find ourselves stuck with tools that don't work anymore.


The problem is the shear volume of new frameworks. Writing your own framework has almost become a rite of passage for software developers. And I'm not knocking that. I did it myself. I learned a lot, the most important of which was that my framework was crap and that I didn't really need it at all.


Ehh. As a developer, I find the best way to understand a concept is to sometimes implement it from scratch. Sometimes your implementation may just be for your own edification on how the patten (doesn't) works, other times it can be useful for others.

Your view is limiting


First impression after reading across the docs: Compact, fast, seems to contain most important things. I like it.


Unfortunately, they all start out this way it seems. Hopefully, this one will remain this way.


The way to keep it that way is to never add even 1 new feature and only fix bugs. That's not a bad idea actually; if you don't, it'll end up like every other framework on earth. So it's a good niche to say 'this is it' and keep it exactly like that even though it 'lacks features'.


I like that idea. Keep the core frozen. With a good API, users can extend it to their liking, and complicate it all they want.


Sorta like Backbone... People definitely love adding plugins and definitely don't count it as a negative when comparing to other JS MVC frameworks......


But this was built to solve problems the author was having at the time he/she created it. In the future, as the web evolves, those problems will change, and you're saying the framework shouldn't adapt to solve new problems?


I'm saying that's the only way to keep it from entering the inevitable malignant growth phase frameworks enter in after a bit because the author / community wants it to do everything. And we already have a ton of those, some popular, most not.


Looks very clean. Especially having just tried a read through the angular docs. Bleurgh. Take the advice of the others in this thread and avoid bloating this and maybe it will take off!


Hi lhorie, I like this. I want to use this. How long will this supported is my concern. Or is this a fun side hobby you worked on and plan on moving on to the next with no future updates?

Cheers.


It's open source. You provide the support and updates.


I've been using the dform library [link] to create my DOM. I'm creating an input form used to query a database with PHP and generate a chart from the data using FusionCharts. I've been considering utilizing frameworks to help with form creation, manipulation, and query generation/execution/processing, but I've decided to stick with just these two libraries for the former and raw PHP for the latter. I also use PHP to enable/disable parts of my form via if's and file includes.

I really like dform's flexibility. I have been able to add every HTML elements that I've needed with the same syntax for every one, making it very simple to stick them into PHP loops to generate the elements based on the query results. I've learned a lot about writing jQuery functions while using it.

FusionCharts is very specific in what it does, and is compatible with many languages. I have actually learned to use it's classes in both PHP and JavaScript so that I can create any part of the chart I please in either language, and it is capable of transmitting data in both JSON and XML even though it uses XML internally. I've briefly taught myself the differences between the formats and have gone from manually creating XML, to learning that JSON is way better, to automatically converting my arrays into JSON data.

I feel that I've kept my project quite lean indeed, and was actually worried about not accomplishing a lot, not being complex enough. But I have intentionally been trying to avoid complexities so that I can make it easy to pick up and use, just include jquery, dform, and the FusionCharts.js/.php (and source code).

[link] https://github.com/daffl/jquery.dform


The templating language reminds me a little of what spacebars (htmljs) probably looks like if a nice mustache facade wasn't put in front of it. (which goes back to Gee's comment about react... but I guess anything these days that promotes a 'virtual dom' will probably be tarnished with that brush from first glance)

I love the simplicity and independent direction this micro-framework provides. It's very 'non-magical' which I think makes it far more appealing.

If you end up solving the HATEOAS/ember-data sideloading/'embedded foreign key data loading' problem I think this will be my goto library (though this probably falls out of the microframework requirements also ;)


tiny, small, idiomatic javascript;

uses POJOs; has just enough 'binding'; uses routing and basic promise/A+ implementation.

I'd say this is good and as high level as one should go if they want to churn really good performance out of an app.

Keep it coming mate, and link with me on Twitter. I used a number of micro frameworks and my own glue to make my own mini-framework using much of the same concepts, only I used a templating engine called doT instead of declaring my HTML structure in JS (although I know it is easily possible to generate view code with a tool/script with HTML as input).

@matthiasak mkeas dot org


Aside from templates, this is the same as https://github.com/moot/riotjs


I thought riot relied on jQuery and was basically decreed to be more marketing than substance. This seems to have a lot of substance.


I like the small, unmagical, well-defined API.

That strength is also a weakness though. Particularly because for coffeescript (or my favourite, the related LiveScript) it doesn't result in the super-cute syntax of e.g. TeaCup[1][2].

[1] https://github.com/goodeggs/teacup

[2] Markaby begat CoffeeKup begat CoffeeCup and DryKup which begat Teacup


It's pretty, I like the rendering. A more complex example in the docs could probably be useful though, something like a bootstrap form.


This actually fills a good niche, if i'm looking for a monolithic framework that trys to do a bunch of things, angular fits the bill. I like that this does one thing, and seems to do it exceedingly well (fast). I'm not in the market for a new framework right now, but the next time I am... i'd consider it, if it stays small and compact like it is today.


I know react is not an MVC framework, but I think you should add it in for comparisons since it uses similar rendering concepts


The author comments on react here: http://lhorie.github.io/mithril/comparison.html


Thanks for making this, this looks really really good!

I'm a backend guy beginning to venture into front-end stuff (recently picked up JS for a Node project at work), and the extensive documentation is incredibly helpful for newcomers like myself.

PS: I'd be very interested in some performance comparisons with Om and Vue.


I think the m.prop getter/setter creation is a good compromise between the "magical" creation of properties employed by Vue and Ractive and the cumbersome use of SomeFramework.set("propertyname", value) employed by other MV* frameworks.

I can't think of a good argument against it.


I like the license, I like the level and style of documentation, I like the lean scope, and I like the timing as I've just started looking into an MVC framework to learn. I was going to choose react over angular but now I'm going to choose this over either.


I just came back to say that the documentation really is beyond excellent.


Off-topic: how do you apply animation to your logo on activation of the browser tab? I loaded your page in a new tab, but the logo animation didn't fire until I viewed that specific tab.


The three rings are in a span. So it's content 'o' in span, the 'o' in :before and 'o' in :after.

Then he uses a -webkit-animation to translate the positions of each ring.

    @keyframes logo {
    	from {opacity:0;transform:scale(2) rotate(359deg);}
    	to {opacity:1;transform:scale(1) rotate(0deg);}
    }
    @-webkit-keyframes logo {
    	from {opacity:0;-webkit-transform:scale(2) rotate(359deg);}
    	to {opacity:1;-webkit-transform:scale(1) rotate(0deg);}
    }


But does that explain how it's not triggered until you click the tab?


This is browser specific to trigger animations on tab open


just a bit of a nitpick, but your getting started page uses the word "performant" http://english.stackexchange.com/questions/38945/what-is-wro...

the debate around the word just leaves a bad taste in my mouth. after reading arguments from both sides I would tend to agree: until the word is a word, it just sounds like marketing buzzword garbage.


> ... until the word is a word, it just sounds like marketing buzzword garbage.

Fair enough, but remember that a word becomes a word simply by usage, not by edict. The old Webster's rule was that, if a word appeared in ten recognized publications with a consistent meaning, it (the word and the definition) would be included in the next edition of the dictionary. I'm sure there's a similar rule in play today.

The fluid and pragmatic nature of words and their meanings can be gauged by noting that "literally" and "figuratively" now mean the same thing:

Link: http://www.merriam-webster.com/dictionary/literally

Quote:

1 : in a literal sense or manner : actually <took the remark literally> <was literally insane>

2 : in effect : virtually <will literally turn the world upside down to combat cruelty or injustice — Norman Cousins>


That SE page is 2 years old. Also, "performant" doesn't have any great synonyms, so unless there's a strong risk of being misunderstood (unlikely in this context), there's no reason not to use it.


it is a word in french ;)

It would translate to overperforming. Compared to what? No one knows. It is oddly very used as a buzzword in marketing.

If so (I could be wrong), maybe it is not marketing but "pardon my french" juste like "idiot savant", "touché" and other french words that appears in the mouth of some that wants to look smart: it make us laugh very hard as a mix of posture and failure.


This is very promising. I like the philosophies underpinning this framework, and once it is in a mature state, I could certainly see myself using it in a production environment.


This is some pretty amazing code, actually. An MVC in 400 lines? Awesome. I could learn from this for days, lack of semicolons and one-line loops notwithstanding.


So, how do I integrate existing JQuery plugins? In most real life scenarios there will be a few plugins which one would like to bring in.


With that style of view, I suppose it's probably pretty easy to adapt Mithril to support server-side rendering.


This is the kind of simplicity that brought me to Clojure for the past couple years.


I would love to see a comparasion with knockoutjs/durandal too...


Quite similar to React? Probably about the same performance?


I don't see this as solving the same problems as React.

React is not "a templating language" that "uses a Virtual DOM", React is an implementation of _autonomous nestable components_, and that requires rerendering of HTML. Virtual DOM is just a way to achieve that rerendering better.


I was actually wondering why this wouldn't just use React for the view. Then it could leverage all of the work going on with React (server-side rendering, possibly rendering from within a Web Worker, etc.)


Given the philosophy leo outlines, I'd guess he's going for something slimmer. This is 3k vs react's 29k. I am curious to know how it's achieving the same thing in so much less code, or rather, what react can do that mithril can't.


I don't mean this in a bad way, but do we really need another Javascript framework? Wouldn't author's energy be more useful when contributing to one of the existing ones?


The author provides an insightful comparison of Mithril to existing frameworks:

http://lhorie.github.io/mithril/comparison.html


Framework fatigue is understandable. Need? No. Want? Yes and maybe. Fortunately, no one's forced to use any one thing. That's because of the answer to your question: yes, no, and maybe. Do we really need another app? Another book? New languages? More music? Yet another program? People choosing their time and pursuits is what leads to countless dead-ends and countless progress. One might want to fear a day when the generality of a no displaces the spirit of a yes, about whether an activity is justified inherently.


That really depends on the author's motives. A lot can be learned by creating "yet another X". Not everyone thinks the same way and creating a framework from scratch gives the author an opportunity to get a deeper understanding for similar concepts and possibly a chance to invent a whole new approach. So many great things would not exist if everyone just settled to use or contribute to existing projects with similar goals to their own.


Normally I'd agree, but I think this has compelling enough reason to be on it's own. The lightweight nature means it's only real competition is Backbone, which is more of a library for making a framework than a framework itself. But Mithril seems to do more while still managing to have a much lighter footprint and feel more like "just JavaScript".


Not sure why I'm getting downvoted, it was a serious concern I expressed, with 0 intent of insulting anyone


You're getting downvoted because this guy thought it was an interesting problem to think about, sat down and put some effort in, came up with something he liked (and which appears to be very good) and wanted other people to see it, and your input was, "isn't this a waste of time? People have done this before!"

The fact that frameworks are appearing at a ridiculous rate is also the reason people still feel compelled to make their own: the existing ones sometimes feel thrown together and flawed in obvious ways, so people automatically think about ways they could be better and try those things out. The result is progress (edit: sometimes the result is a bag of crap, but that's beside the point).

No, we don't need more frameworks. And yet that still isn't a good reason to stop building them, if for no other reason than to explore for yourself what goes into building one (if that's your interest).


> I don't mean this in a bad way, but do we really need another Javascript framework?

Obviously the author thinks so.


Anyone who wants to make my coding life easier has my full support. If you think you got something that's gonna change things, I wanna see it.


That's the reason why I just shake my head every time I hear "HTML5 will overtake native for mobile apps anyday now". They still fight over what is the best MVC framework and have no idea what native offers. When did you last saw Android or iOS dev worried which MVC framework should they use, or maybe write their own?


Well that's just silly. The native platforms already have the frameworks in place, and have done since the beginning. If you were writing a game in a native system you'd have to choose from different frameworks - or write your own.

With JS the restrictions and limitations of the system are different. People are coming up with creative and innovative ways to work with them. There's no problem with that, it a process that's ongoing in all software, and forever will be.

Also, "You" should really avoid using "They", especially when it applies to a broad group of people that you're making assumptions and judgements about. Maybe "You" have no idea what HTML5 apps offer (like portability)? I happen to think that html5 is going to be a major player in apps development, which puts me in your "They" camp, then again, I know what native offers - so I'm one of the many contradictions to your statement.


I agree completely.


seems a lot like React. I haven't used it, but isn't it also built around the idea of Shadow Doms?


new function(window){}(this) is kinda weird to me. Sorry that's nothing valuable to add to your code.


Correct me if I'm wrong, but what this does is: 1) allows the code to be run in an environment where 'window' is not the global object (perhaps for testing), and 2) in a browser, gives the caller the option to isolate the module code so it's not allowed to touch 'window'.


do you have a todomvc example :)


The guide is based off the building of a TODO app. http://lhorie.github.io/mithril/getting-started.html


There is an example available here: https://github.com/jpmonette/todomvc-mithril


nice and clean


Why?




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

Search: