Atomic web development - Combining the power of npm, Browserify, Rework and more to build small, fully encapsulated client side modules
machinepacks
(but maybe there are alternative and better specs)
var componentAPI = componentName({
container : `domNodeOrSelector`, // maybe it should always be a dom node
options : {/* configuration options */},
data : `modelOrStreamOrEventEmitter`, // to initialize or update stuff
children : [ // this is optional, because maybe the component can use defaults if not provided
{ '__title' : titleComponent },
{ '__list' : listComponent },
{ '__sidebar : sidebarComponent }
]
});
This does say nothing about the Interface
that is returned from a call to any component, like componentName(..)
, titleComponent(..)
, listComponent(..)
or sidebarComponent(..)
.
But I imagine, that it should be possible to somehow formally define an Interface
, so that I can easily write a component that I can inject into an existing component to change it.
Maybe node machines are nice for defining those interfaces, but maybe there are alternatives :-)
What do you guys think of that?
menuComponent
or articleComponent
if i need too, or just build a new website component (by forking the existing one and changing it completely... The goal is to start with simple components and slowly build more complex ones over time
There is a standalone virtual dom library from matt esch.
Hyperscript is a pretty interesting library and a "virtual dom" version exists:
There are also some modules to use stream for updating the dom that i like a lot
You said earlier:
@joeybaker
possible, but I’m inclined to think that it’s to complex to have that many children?
why is the title and sidebar a part of this component?
and when i described stuff in more detail
@joeybaker
totally agreed there!
I guess that API you should would be for a “page” or some such?
I would actually like to brainstorm about a component interface that enables everything in a page to be a component.
The most "top level" components could be different kinds of "webpageComponents" (which set meta-tags
, html
, head
, body
, css-reset
and a lot of other things) so people could have something like:
<body><script="bundle.js"></script></body>
var webpage = require('webpage'); // encapsulates best practice boilerplate for certain kinds of projects
var webpageAPI = webpage({
container: document.body,
options: {/* some configuration depending on the project that should be built*/},
data: {/* some webpage specific data or e.g. RESTful endpoints to fetch the data */},
children: { /* inject some components to be used to render page in detail - if not provided might fallback to default components */}
});
/* Below the "webpageAPI" might be used to interact with the page or register some listeners to react on certain things */
I'm thinking of using leveldb
(a version that wraps indexdb
) to cache everything on the client side and potentially replicate with a remote leveldb
if provided in options
. A nice thing is, that leveldb
has a library sublevel
to provide a leveldb
interface to a subset of data in the database. This could be "passed down" into certain components in the DOM tree. Furthermore, I'm thinking of using BEM conventions in css and templates for all components and to mark "insertion points" for injected components from the "children: {..}" parameter.
componentAPI
?I'm not sure if the atomify chat is the right place to discuss these things, but I really enjoy the feedback from you and whether you imagine this kind of approach would be of any value or not (perhaps because there is already plenty of stuff out there that covers what i'm trying to achieve)
atomify init
option as an alternative to npm init
, that generates a "package.json" file with default values for the atomify configuration
. I'm currently tryin to found a coding school for kids and might teach them atomify. I'm basically copy&pasting always the same atomify configuration
from one project to the next - a init
option would be neat.