ajklein / webkit Goto Github PK

View Code? Open in Web Editor NEW

This project forked from webkit/webkit-http

1.0 1.0 1.0 4.43 GB

Unofficial mirror of the WebKit SVN repository

Home Page: http://www.webkit.org/

webkit's People

Contributors

Stargazers

Watchers

Forkers

arv

webkit's Issues

design formal delegate API

right now it's not much of an API, just a convention

non-string model values cause crashes in v8::Value::toString

example test case that crashes (due to not having the proper v8::Context available):

<!DOCTYPE html>
<body>
<script>
if (window.testRunner) {
    testRunner.waitUntilDone();
    testRunner.dumpAsText();
    setTimeout(function() { testRunner.notifyDone() }, 1000);
}
document.body.model = [{key: 1}, {key: 2}, {key: 3}];
</script>
<div>Below, there should be a list of three items.</div>
<ol>
  <template iterate>
    <li>{{key}}
  </template>
</ol>
</body>

Fix is likely to pass a ScriptState to ScriptValue::toString and actually make use of it.

Avoid memory leaks by holding the .model and .modelDelegate properties in the JS wrapper

Right now, .model and .modelDelegate are in NodeRareData as strong references (in a ScriptValue). This will cause memory leaks if the model or modelDelegate have a reference back to the DOM. The easiest solution will be to move them to the JS wrapper, and store them in either internal fields or (more likely) hidden properties.

Provide an API on Element and Text to get a list of bindings

This is useful in cases where we want to replace on element with another (e.g., WebComponent upgrade) and need to copy bindings from one place to another. We already have addBinding(), so the obvious thing to add is getBindings().

removing a template element should cause its instances to be removed

This is consistent with the "strong connection" view of templates and instances

Wrong check for text binding

HTMLTemplateElement.cpp, addTextBinding, line 94

if (index != notFound && text->data().find("}}"))

this should be find("}}", index + 2) or we would match strings like this " abc }} def {{ ghi"

I should really just create my own clone and do pull requests.

design a safe mechanism for binding markup into a "hole" in a template.

need tests asserting that parser-created templates function properly

e.g.

<body>
<script>
  document.body.model = [1, 2];
</script>
<template iterate>
  Foo
</template>

should cause two instances to be produced

Attempt to implement better cloneNode(true) behavior for templates

The idea is that a template needs to take with it it's internal state (what model it's current iterator is pointing at, and how many instances it thinks it has), and it's instance nodes need to clone their terminators.

Add synchronous API to HTMLTemplateElement for template instantiation

Something like go() and stop(). go() would cause the template to synchronously begin processing (instantiating or iterating). stop() would cause the template to synchronously tear itself down (useful if the template is to be moved elsewhere in the tree, to avoid double-stamping).

re-hookup innerHTML for HTMLTemplateElement

Removing it was a mistake. If template parses special, it needs to serialize special a well (they should be symmetric).

It's fine that appendChild, etc... have different behavior. The goal here is that template contents serialize but aren't in the parent/child hierarchy.

design context-aware bindings

which property escape binding content based on the type of the attribute content

need the ability to reference "this" and ancestor values within bindings

It may be that this is left to delegates to implement, but if so, we need tests showing that it is possible.

implement <template instantiate/iterate ref>

Template's should be able to iterate or instantiate, and reference their prototype DOM from another template element.

design how model values propagate into shadow dom

Right now, model value is blinding forwarded across to the ShadowRoot. This was just something simple to get working initially, but in retrospect has some elegance to it. Specifically, it leaves the implementation of the component in full control of what happens. E.g. If the component actually wants to blindly accept the model it can do that. If it wants to "alias" path values to internal path values, it can initially set it's root model as null, require some kind of attribute API which declares explicit bindings and then synthesize an internal model based on the bindings.

In any case, it's worth either making a test that asserts this pattern is workable or coming up with a different design.

Error in Source/WebCore/mdv/ScriptValueBinding.cpp

Source/WebCore/mdv/ScriptValueBinding.cpp line 147

if (pathValues.hasNoValue() || !result.isObject())
    return false;

should be

if (pathValues.hasNoValue() || !pathValues.isObject())
    return false;

I guess if I return a delegate with a non object this is observable

added .model and .modelDelegate to ShadowRoot

ref needs to look in template owner document too

In HTMLTemplateElement.cpp near line 214:

Element* element = m_template->treeScope()->getElementById(m_template->fastGetAttribute(refAttr));

This does not work correctly since we might need to check both the ownerDocument of the template element as well as the template contents owner

Maybe an example will show what I mean.

Also see https://code.google.com/p/mdv/source/browse/template_element.js#510

Implement DOM stability for iteration

This requires synthesizing "splice" operations and transforming the existing array expansion from the old state to the new state with the fewest removals, additions. also, this should be willing to re-order existing instances. e..g

an iteration based on an array which is sorted should neither create nor destroy instances.

ajklein / webkit Goto Github PK

webkit's People

Contributors

Stargazers

Watchers

Forkers

webkit's Issues

Recommend Projects

Recommend Topics

Recommend Org

Jobs