Adjust allowed children some more
No reason to prohibit <h1><p>foo</p></h1>, since we allow
<pre><p>foo</p></pre>. Also expanded the rationale. This could use
revisiting later.
--- a/editcommands.html Wed May 25 13:35:50 2011 -0600
+++ b/editcommands.html Wed May 25 13:54:18 2011 -0600
@@ -309,14 +309,33 @@
<p>A <a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-node title=concept-node>node</a> or string <var title="">child</var> is an <dfn id=allowed-child>allowed child</dfn> of a
<a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-node title=concept-node>node</a> or string <var title="">parent</var> if the following algorithm returns true:
-<p class=XXX>This is very ad hoc and might need to be rethought. We can't use
-the HTML spec's definitions because those are too complicated, they don't take
-obsolete elements into account, and they're sometimes too restrictive. (We
-don't like having to contort the DOM to ensure that it's valid, because it can
-have unwanted side effects, so we want to minimize the number of cases we
-disallow.) Mostly this list covers only things that don't serialize as
-text/html. It's not intended to be complete, and in particular, it omits lots
-of cases that aren't likely to come up for us.
+<div class=XXX>
+<p>For the most part, right now we only disallow children when they wouldn't
+serialize to text/html, or in a couple of other cases where they'd behave very
+strangely (like a list item that's not the child of a list). It could well
+make sense to disallow children when they would be invalid per HTML5, but this
+has a few problems:
+
+<ol>
+ <li>We need to handle invalid elements like center, which have no conformance
+ requirements but can interfere with serialization (center cannot descend from
+ p).
+
+ <li>HTML5 validity requirements are not especially stable, so it would be
+ harder to stay up-to-date, while the parsing algorithm is quite stable.
+
+ <li>Sometimes users give instructions that have to produce invalid DOMs to
+ get the expected effect, like indenting the first item of a list.
+
+ <li>Making more children disallowed means we have to split parents more
+ often, and splitting parents can inevitably have side-effects, so we'd really
+ prefer to minimize it.
+</ol>
+
+<p>I didn't try to cover all serialization problems for now, particularly where
+they seemed implausible. Whatever happens, I'm pretty sure I'll revise this
+substantially sometime in the future, but I'm not sure exactly what to aim for.
+</div>
<ol>
<li>If <var title="">parent</var> is "colgroup", "table", "tbody", "tfoot", "thead",
@@ -349,11 +368,14 @@
<li>If <var title="">child</var> is one of the <a href=#prohibited-paragraph-children>prohibited paragraph
children</a> and <var title="">parent</var> or some <a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-tree-ancestor title=concept-tree-ancestor>ancestor</a> of
- <var title="">parent</var> is an <a href=#html-element>HTML element</a> with <a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-element-local-name title=concept-element-local-name>local name</a> "h1",
- "h2", "h3", "h4", "h5", "h6", or "p", return false.
- <!-- This cannot be serialized as text/html if the parent is a p, or if the
- parent and child are both h*. Something like <h1>foo<p>bar</p></h1> will
- actually work, but while we're here, we may as well disallow it. -->
+ <var title="">parent</var> is a <code class=external data-anolis-spec=html title="the p element"><a href=http://www.whatwg.org/html/#the-p-element>p</a></code>, return false.
+ <!-- This generally cannot be serialized either. -->
+
+ <li>If <var title="">child</var> is "h1", "h2", "h3", "h4", "h5", or "h6", and
+ <var title="">parent</var> or some <a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-tree-ancestor title=concept-tree-ancestor>ancestor</a> of <var title="">parent</var> is an
+ <a href=#html-element>HTML element</a> with <a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-element-local-name title=concept-element-local-name>local name</a> "h1", "h2", "h3", "h4", "h5",
+ or "h6", return false.
+ <!-- Nor this. -->
<li>Let <var title="">parent</var> be the <a class=external data-anolis-spec=domcore href=http://dvcs.w3.org/hg/domcore/raw-file/tip/Overview.html#concept-element-local-name title=concept-element-local-name>local name</a> of <var title="">parent</var>.
<!-- Further requirements only care about the parent itself, not ancestors,
@@ -397,11 +419,11 @@
<tr><th>Parent <th>Prohibited children
<tr><td>a <td>a
<tr><td>dd, dt <td>dd, dt
+ <tr><td>h1, h2, h3, h4, h5, h6 <td>h1, h2, h3, h4, h5, h6
<tr><td>li <td>li
<tr><td>nobr <td>nobr
+ <tr><td>p <td>All <a href=#prohibited-paragraph-children>prohibited paragraph children</a>
<tr><td>td, th <td>caption, col, colgroup, tbody, td, tfoot, th, thead, tr
- <tr><td>h1, h2, h3, h4, h5, h6, p <td>All <a href=#prohibited-paragraph-children>prohibited paragraph
- children</a>
</table>
<li>Return true.
--- a/implementation.js Wed May 25 13:35:50 2011 -0600
+++ b/implementation.js Wed May 25 13:54:18 2011 -0600
@@ -810,27 +810,28 @@
if (isHtmlElement(parent_)) {
// "If child is "a", and parent or some ancestor of parent is an a,
// return false."
- if (child == "a") {
- var ancestor = parent_;
- while (ancestor) {
- if (isHtmlElement(ancestor, "a")) {
- return false;
- }
- ancestor = ancestor.parentNode;
+ //
+ // "If child is one of the prohibited paragraph children and parent or
+ // some ancestor of parent is a p, return false."
+ //
+ // "If child is "h1", "h2", "h3", "h4", "h5", or "h6", and parent or
+ // some ancestor of parent is an HTML element with local name "h1",
+ // "h2", "h3", "h4", "h5", or "h6", return false."
+ var ancestor = parent_;
+ while (ancestor) {
+ if (child == "a" && isHtmlElement(ancestor, "a")) {
+ return false;
}
- }
-
- // "If child is one of the prohibited paragraph children and parent or
- // some ancestor of parent is an HTML element with local name "h1",
- // "h2", "h3", "h4", "h5", "h6", or "p", return false."
- if (prohibitedParagraphChildren.indexOf(child) != -1) {
- var ancestor = parent_;
- while (ancestor) {
- if (isHtmlElement(ancestor, ["h1", "h2", "h3", "h4", "h5", "h6", "p"])) {
- return false;
- }
- ancestor = ancestor.parentNode;
+ if (prohibitedParagraphChildren.indexOf(child) != -1
+ && isHtmlElement(ancestor, "p")) {
+ return false;
}
+ if (/^h[1-6]$/.test(child)
+ && isHtmlElement(ancestor)
+ && /^H[1-6]$/.test(ancestor.tagName)) {
+ return false;
+ }
+ ancestor = ancestor.parentNode;
}
// "Let parent be the local name of parent."
@@ -890,10 +891,11 @@
var table = [
[["a"], ["a"]],
[["dd", "dt"], ["dd", "dt"]],
+ [["h1", "h2", "h3", "h4", "h5", "h6"], ["h1", "h2", "h3", "h4", "h5", "h6"]],
[["li"], ["li"]],
[["nobr"], ["nobr"]],
+ [["p"], prohibitedParagraphChildren],
[["td", "th"], ["caption", "col", "colgroup", "tbody", "td", "tfoot", "th", "thead", "tr"]],
- [["h1", "h2", "h3", "h4", "h5", "h6", "p"], prohibitedParagraphChildren],
];
for (var i = 0; i < table.length; i++) {
if (table[i][0].indexOf(parent_) != -1
--- a/source.html Wed May 25 13:35:50 2011 -0600
+++ b/source.html Wed May 25 13:54:18 2011 -0600
@@ -263,14 +263,33 @@
<p>A [[node]] or string <var>child</var> is an <dfn>allowed child</dfn> of a
[[node]] or string <var>parent</var> if the following algorithm returns true:
-<p class=XXX>This is very ad hoc and might need to be rethought. We can't use
-the HTML spec's definitions because those are too complicated, they don't take
-obsolete elements into account, and they're sometimes too restrictive. (We
-don't like having to contort the DOM to ensure that it's valid, because it can
-have unwanted side effects, so we want to minimize the number of cases we
-disallow.) Mostly this list covers only things that don't serialize as
-text/html. It's not intended to be complete, and in particular, it omits lots
-of cases that aren't likely to come up for us.
+<div class=XXX>
+<p>For the most part, right now we only disallow children when they wouldn't
+serialize to text/html, or in a couple of other cases where they'd behave very
+strangely (like a list item that's not the child of a list). It could well
+make sense to disallow children when they would be invalid per HTML5, but this
+has a few problems:
+
+<ol>
+ <li>We need to handle invalid elements like center, which have no conformance
+ requirements but can interfere with serialization (center cannot descend from
+ p).
+
+ <li>HTML5 validity requirements are not especially stable, so it would be
+ harder to stay up-to-date, while the parsing algorithm is quite stable.
+
+ <li>Sometimes users give instructions that have to produce invalid DOMs to
+ get the expected effect, like indenting the first item of a list.
+
+ <li>Making more children disallowed means we have to split parents more
+ often, and splitting parents can inevitably have side-effects, so we'd really
+ prefer to minimize it.
+</ol>
+
+<p>I didn't try to cover all serialization problems for now, particularly where
+they seemed implausible. Whatever happens, I'm pretty sure I'll revise this
+substantially sometime in the future, but I'm not sure exactly what to aim for.
+</div>
<ol>
<li>If <var>parent</var> is "colgroup", "table", "tbody", "tfoot", "thead",
@@ -303,11 +322,14 @@
<li>If <var>child</var> is one of the <span>prohibited paragraph
children</span> and <var>parent</var> or some [[ancestor]] of
- <var>parent</var> is an <span>HTML element</span> with [[localname]] "h1",
- "h2", "h3", "h4", "h5", "h6", or "p", return false.
- <!-- This cannot be serialized as text/html if the parent is a p, or if the
- parent and child are both h*. Something like <h1>foo<p>bar</p></h1> will
- actually work, but while we're here, we may as well disallow it. -->
+ <var>parent</var> is a [[p]], return false.
+ <!-- This generally cannot be serialized either. -->
+
+ <li>If <var>child</var> is "h1", "h2", "h3", "h4", "h5", or "h6", and
+ <var>parent</var> or some [[ancestor]] of <var>parent</var> is an
+ <span>HTML element</span> with [[localname]] "h1", "h2", "h3", "h4", "h5",
+ or "h6", return false.
+ <!-- Nor this. -->
<li>Let <var>parent</var> be the [[localname]] of <var>parent</var>.
<!-- Further requirements only care about the parent itself, not ancestors,
@@ -351,11 +373,11 @@
<tr><th>Parent <th>Prohibited children
<tr><td>a <td>a
<tr><td>dd, dt <td>dd, dt
+ <tr><td>h1, h2, h3, h4, h5, h6 <td>h1, h2, h3, h4, h5, h6
<tr><td>li <td>li
<tr><td>nobr <td>nobr
+ <tr><td>p <td>All <span>prohibited paragraph children</span>
<tr><td>td, th <td>caption, col, colgroup, tbody, td, tfoot, th, thead, tr
- <tr><td>h1, h2, h3, h4, h5, h6, p <td>All <span>prohibited paragraph
- children</span>
</table>
<li>Return true.