mirror of https://github.com/n2geoff/um
dom diffing for more effecent rerendering
This commit is contained in:
parent
247475ef8b
commit
ed2e84ee0a
|
@ -7,11 +7,11 @@ Um, is an experimental composable UI builder that takes ideas from early [hypera
|
|||
Um, because you should think about, um, NOT using it.
|
||||
|
||||
## Features
|
||||
- No Virtual Dom
|
||||
- Real DOM
|
||||
- [Non-destructive (re)Rendering](https://github.com/bryhoyt/emerj)
|
||||
- No Build System
|
||||
- No Over Engineering
|
||||
- ~1kb minified
|
||||
- Totally INEFFICIENT rendering (at scale)
|
||||
- ~2kb minified
|
||||
|
||||
## Install
|
||||
|
||||
|
@ -81,4 +81,5 @@ The `h()` is an **optional** hypertext build utility that weighs in around **~25
|
|||
|
||||
### TODO
|
||||
|
||||
- Improve Update
|
||||
- Some tag attributes do not work, like rowspan on td
|
||||
- Rethink State Management, might be ok
|
||||
|
|
2
index.js
2
index.js
|
@ -1,5 +1,5 @@
|
|||
import h from "./src/tag.js";
|
||||
import app from "./src/app.js";
|
||||
import h from "./src/tag.js";
|
||||
|
||||
export {app};
|
||||
export {h};
|
|
@ -1,3 +1,7 @@
|
|||
import diff from "./emerj.js";
|
||||
|
||||
/*! Um v0.5.0 | MIT LICENSE | https://github.com/n2geoff/um */
|
||||
|
||||
/**
|
||||
* App Builder
|
||||
*
|
||||
|
@ -54,7 +58,7 @@ export default function app(opts) {
|
|||
|
||||
/** update dom */
|
||||
const update = () => {
|
||||
document.querySelector(mount).replaceChildren(view(state, actions));
|
||||
diff.merge(document.querySelector(mount), view(state, actions));
|
||||
}
|
||||
|
||||
// mount view
|
||||
|
|
|
@ -0,0 +1,122 @@
|
|||
/*! Emerj v1.0.0 | MIT LICENSE | https://github.com/bryhoyt/emerj */
|
||||
export default {
|
||||
attrs(elem) {
|
||||
const attrs = {};
|
||||
for (let i=0; i < elem.attributes.length; i++) {
|
||||
const attr = elem.attributes[i];
|
||||
attrs[attr.name] = attr.value;
|
||||
}
|
||||
return attrs;
|
||||
},
|
||||
nodesByKey(parent, makeKey) {
|
||||
const map = {};
|
||||
for (let j=0; j < parent.childNodes.length; j++) {
|
||||
const key = makeKey(parent.childNodes[j]);
|
||||
if (key) map[key] = parent.childNodes[j];
|
||||
}
|
||||
return map;
|
||||
},
|
||||
merge(base, modified, opts) {
|
||||
/* Merge any differences between base and modified back into base.
|
||||
*
|
||||
* Operates only the children nodes, and does not change the root node or its
|
||||
* attributes.
|
||||
*
|
||||
* Conceptually similar to React's reconciliation algorithm:
|
||||
* https://facebook.github.io/react/docs/reconciliation.html
|
||||
*
|
||||
* I haven't thoroughly tested performance to compare to naive DOM updates (i.e.
|
||||
* just updating the entire DOM from a string using .innerHTML), but some quick
|
||||
* tests on a basic DOMs were twice as fast -- so at least it's not slower in
|
||||
* a simple scenario -- and it's definitely "fast enough" for responsive UI and
|
||||
* even smooth animation.
|
||||
*
|
||||
* The real advantage for me is not so much performance, but that state & identity
|
||||
* of existing elements is preserved -- text typed into an <input>, an open
|
||||
* <select> dropdown, scroll position, ad-hoc attached events, canvas paint, etc,
|
||||
* are preserved as long as an element's identity remains.
|
||||
*
|
||||
* See https://korynunn.wordpress.com/2013/03/19/the-dom-isnt-slow-you-are/
|
||||
*/
|
||||
opts = opts || {};
|
||||
opts.key = opts.key || (node => node.id);
|
||||
|
||||
if (typeof modified === 'string') {
|
||||
const html = modified;
|
||||
// Make sure the parent element of the provided HTML is of the same type as
|
||||
// `base`'s parent. This matters when the HTML contains fragments that are
|
||||
// only valid inside certain elements, eg <td>s, which must have a <tr>
|
||||
// parent.
|
||||
modified = document.createElement(base.tagName);
|
||||
modified.innerHTML = html;
|
||||
}
|
||||
|
||||
// Naively recurse into the children, if any, replacing or updating new
|
||||
// elements that are in the same position as old, deleting trailing elements
|
||||
// when the new list contains fewer children, or appending new elements if
|
||||
// it contains more children.
|
||||
//
|
||||
// For re-ordered children, the `id` attribute can be used to preserve identity.
|
||||
|
||||
// Loop through .childNodes, not just .children, so we compare text nodes (and
|
||||
// comment nodes, fwiw) too.
|
||||
|
||||
const nodesByKey = {old: this.nodesByKey(base, opts.key),
|
||||
new: this.nodesByKey(modified, opts.key)};
|
||||
|
||||
let idx;
|
||||
for (idx=0; modified.firstChild; idx++) {
|
||||
const newNode = modified.removeChild(modified.firstChild);
|
||||
if (idx >= base.childNodes.length) {
|
||||
// It's a new node. Append it.
|
||||
base.appendChild(newNode);
|
||||
continue;
|
||||
}
|
||||
|
||||
let baseNode = base.childNodes[idx];
|
||||
|
||||
// If the children are indexed, then make sure to retain their identity in
|
||||
// the new order.
|
||||
const newKey = opts.key(newNode);
|
||||
if (opts.key(baseNode) || newKey) {
|
||||
// If the new node has a key, then either use its existing match, or insert it.
|
||||
// If not, but the old node has a key, then make sure to leave it untouched and insert the new one instead.
|
||||
// Else neither node has a key. Just overwrite old with new.
|
||||
const match = (newKey && newKey in nodesByKey.old)? nodesByKey.old[newKey]: newNode;
|
||||
if (match !== baseNode) {
|
||||
baseNode = base.insertBefore(match, baseNode);
|
||||
}
|
||||
}
|
||||
|
||||
if (baseNode.nodeType !== newNode.nodeType || baseNode.tagName !== newNode.tagName) {
|
||||
// Completely different node types. Just update the whole subtree, like React does.
|
||||
base.replaceChild(newNode, baseNode);
|
||||
} else if ([Node.TEXT_NODE, Node.COMMENT_NODE].indexOf(baseNode.nodeType) >= 0) {
|
||||
// This is the terminating case of the merge() recursion.
|
||||
if (baseNode.textContent === newNode.textContent) continue; // Don't write if we don't need to.
|
||||
baseNode.textContent = newNode.textContent;
|
||||
} else if (baseNode !== newNode) { // Only need to update if we haven't just inserted the newNode in.
|
||||
// It's an existing node with the same tag name. Update only what's necessary.
|
||||
// First, make dicts of attributes, for fast lookup:
|
||||
const attrs = {base: this.attrs(baseNode), new: this.attrs(newNode)};
|
||||
for (const attr in attrs.base) {
|
||||
// Remove any missing attributes.
|
||||
if (attr in attrs.new) continue;
|
||||
baseNode.removeAttribute(attr);
|
||||
}
|
||||
for (const attr in attrs.new) {
|
||||
// Add and update any new or modified attributes.
|
||||
if (attr in attrs.base && attrs.base[attr] === attrs.new[attr]) continue;
|
||||
baseNode.setAttribute(attr, attrs.new[attr]);
|
||||
}
|
||||
// Now, recurse into the children. If the only children are text, this will
|
||||
// be the final recursion on this node.
|
||||
this.merge(baseNode, newNode);
|
||||
}
|
||||
}
|
||||
while (base.childNodes.length > idx) {
|
||||
// If base has more children than modified, delete the extras.
|
||||
base.removeChild(base.lastChild);
|
||||
}
|
||||
},
|
||||
};
|
Loading…
Reference in New Issue