mirror of
https://github.com/ezyang/htmlpurifier.git
synced 2025-07-31 19:30:21 +02:00
Rewrite FixNesting implementation to be tree-based.
This mega-patch rips out the FixNesting implementation and the related ChildDef components. The primary algorithmic change is to convert from use of tokens to tree nodes, which are far more amenable to the style of processing that FixNesting uses. Additionally, FixNesting has been changed to go bottom-up rather than top-down, in order to avoid needing to implement backtracking. This patch simplifies a good deal of the relevant logic, since we no longer need to continually recalculate the nesting structure when processing things. However, the conversion to the alternate format incurs some overhead, so for small inputs these changes are not a win. One possibility to greatly reduce the constant factors here is to switch to entirely using libxml's representation, and never serializing tokens; this would require one to rewrite injectors, however. The iterative post-order traversal in FixNesting is a bit subtle, but we have essentially reified the stack and continuations. We've removed support for %Core.EscapeInvalidChildren. Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
This commit is contained in:
Edward Z. Yang
@@ -59,230 +59,164 @@ class HTMLPurifier_ChildDef_Table extends HTMLPurifier_ChildDef
|
||||
}
|
||||
|
||||
/**
|
||||
* @param array $tokens_of_children
|
||||
* @param array $children
|
||||
* @param HTMLPurifier_Config $config
|
||||
* @param HTMLPurifier_Context $context
|
||||
* @return array
|
||||
*/
|
||||
public function validateChildren($tokens_of_children, $config, $context)
|
||||
public function validateChildren($children, $config, $context)
|
||||
{
|
||||
if (empty($tokens_of_children)) {
|
||||
if (empty($children)) {
|
||||
return false;
|
||||
}
|
||||
|
||||
// this ensures that the loop gets run one last time before closing
|
||||
// up. It's a little bit of a hack, but it works! Just make sure you
|
||||
// get rid of the token later.
|
||||
$tokens_of_children[] = false;
|
||||
|
||||
// only one of these elements is allowed in a table
|
||||
$caption = false;
|
||||
$thead = false;
|
||||
$tfoot = false;
|
||||
|
||||
// whitespace
|
||||
$initial_ws = array();
|
||||
$after_caption_ws = array();
|
||||
$after_thead_ws = array();
|
||||
$after_tfoot_ws = array();
|
||||
|
||||
// as many of these as you want
|
||||
$cols = array();
|
||||
$content = array();
|
||||
|
||||
$nesting = 0; // current depth so we can determine nodes
|
||||
$is_collecting = false; // are we globbing together tokens to package
|
||||
// into one of the collectors?
|
||||
$collection = array(); // collected nodes
|
||||
// INVARIANT: if $is_collecting, then !empty($collection)
|
||||
// The converse does NOT hold, see [WHITESPACE]
|
||||
$tag_index = 0; // the first node might be whitespace,
|
||||
// so this tells us where the start tag is
|
||||
$tbody_mode = false; // if true, then we need to wrap any stray
|
||||
// <tr>s with a <tbody>.
|
||||
// <tr>s with a <tbody>.
|
||||
|
||||
foreach ($tokens_of_children as $token) {
|
||||
$is_child = ($nesting == 0);
|
||||
$ws_accum =& $initial_ws;
|
||||
|
||||
if ($token === false) {
|
||||
// terminating sequence started
|
||||
} elseif ($token instanceof HTMLPurifier_Token_Start) {
|
||||
$nesting++;
|
||||
} elseif ($token instanceof HTMLPurifier_Token_End) {
|
||||
$nesting--;
|
||||
foreach ($children as $node) {
|
||||
if ($node instanceof HTMLPurifier_Node_Comment) {
|
||||
$ws_accum[] = $node;
|
||||
continue;
|
||||
}
|
||||
|
||||
// handle node collection
|
||||
if ($is_collecting) {
|
||||
if ($is_child) {
|
||||
// okay, let's stash the tokens away
|
||||
// first token tells us the type of the collection
|
||||
switch ($collection[$tag_index]->name) {
|
||||
case 'tbody':
|
||||
$tbody_mode = true;
|
||||
// fall through
|
||||
case 'tr':
|
||||
$content[] = $collection;
|
||||
break;
|
||||
case 'caption':
|
||||
if ($caption !== false) {
|
||||
break;
|
||||
}
|
||||
$caption = $collection;
|
||||
break;
|
||||
case 'thead':
|
||||
case 'tfoot':
|
||||
$tbody_mode = true;
|
||||
// XXX This breaks rendering properties with
|
||||
// Firefox, which never floats a <thead> to
|
||||
// the top. Ever. (Our scheme will float the
|
||||
// first <thead> to the top.) So maybe
|
||||
// <thead>s that are not first should be
|
||||
// turned into <tbody>? Very tricky, indeed.
|
||||
|
||||
// access the appropriate variable, $thead or $tfoot
|
||||
$var = $collection[$tag_index]->name;
|
||||
if ($$var === false) {
|
||||
$$var = $collection;
|
||||
} else {
|
||||
// Oops, there's a second one! What
|
||||
// should we do? Current behavior is to
|
||||
// transmutate the first and last entries into
|
||||
// tbody tags, and then put into content.
|
||||
// Maybe a better idea is to *attach
|
||||
// it* to the existing thead or tfoot?
|
||||
// We don't do this, because Firefox
|
||||
// doesn't float an extra tfoot to the
|
||||
// bottom like it does for the first one.
|
||||
$collection[$tag_index]->name = 'tbody';
|
||||
$collection[count($collection) - 1]->name = 'tbody';
|
||||
$content[] = $collection;
|
||||
}
|
||||
break;
|
||||
case 'colgroup':
|
||||
$cols[] = $collection;
|
||||
break;
|
||||
}
|
||||
$collection = array();
|
||||
$is_collecting = false;
|
||||
$tag_index = 0;
|
||||
} else {
|
||||
// add the node to the collection
|
||||
$collection[] = $token;
|
||||
}
|
||||
}
|
||||
|
||||
// terminate
|
||||
if ($token === false) {
|
||||
switch ($node->name) {
|
||||
case 'tbody':
|
||||
$tbody_mode = true;
|
||||
// fall through
|
||||
case 'tr':
|
||||
$content[] = $node;
|
||||
$ws_accum =& $content;
|
||||
break;
|
||||
}
|
||||
|
||||
if ($is_child) {
|
||||
// determine what we're dealing with
|
||||
if ($token->name == 'col') {
|
||||
// the only empty tag in the possie, we can handle it
|
||||
// immediately
|
||||
$cols[] = array_merge($collection, array($token));
|
||||
$collection = array();
|
||||
$is_collecting = false;
|
||||
$tag_index = 0;
|
||||
continue;
|
||||
case 'caption':
|
||||
// there can only be one caption!
|
||||
if ($caption !== false) break;
|
||||
$caption = $node;
|
||||
$ws_accum =& $after_caption_ws;
|
||||
break;
|
||||
case 'thead':
|
||||
$tbody_mode = true;
|
||||
// XXX This breaks rendering properties with
|
||||
// Firefox, which never floats a <thead> to
|
||||
// the top. Ever. (Our scheme will float the
|
||||
// first <thead> to the top.) So maybe
|
||||
// <thead>s that are not first should be
|
||||
// turned into <tbody>? Very tricky, indeed.
|
||||
if ($thead === false) {
|
||||
$thead = $node;
|
||||
$ws_accum =& $after_thead_ws;
|
||||
} else {
|
||||
// Oops, there's a second one! What
|
||||
// should we do? Current behavior is to
|
||||
// transmutate the first and last entries into
|
||||
// tbody tags, and then put into content.
|
||||
// Maybe a better idea is to *attach
|
||||
// it* to the existing thead or tfoot?
|
||||
// We don't do this, because Firefox
|
||||
// doesn't float an extra tfoot to the
|
||||
// bottom like it does for the first one.
|
||||
$node->name = 'tbody';
|
||||
$content[] = $node;
|
||||
$ws_accum =& $content;
|
||||
}
|
||||
switch ($token->name) {
|
||||
case 'caption':
|
||||
case 'colgroup':
|
||||
case 'thead':
|
||||
case 'tfoot':
|
||||
case 'tbody':
|
||||
case 'tr':
|
||||
$is_collecting = true;
|
||||
$collection[] = $token;
|
||||
continue;
|
||||
default:
|
||||
// [WHITESPACE] Whitespace is added to the
|
||||
// collection without triggering collection
|
||||
// mode. This is a hack to make whitespace
|
||||
// 'sticky' (that is to say, we ought /not/ to
|
||||
// drop whitespace.)
|
||||
if (!empty($token->is_whitespace)) {
|
||||
$collection[] = $token;
|
||||
$tag_index++;
|
||||
}
|
||||
continue;
|
||||
break;
|
||||
case 'tfoot':
|
||||
// see above for some aveats
|
||||
$tbody_mode = true;
|
||||
if ($tfoot === false) {
|
||||
$tfoot = $node;
|
||||
$ws_accum =& $after_tfoot_ws;
|
||||
} else {
|
||||
$node->name = 'tbody';
|
||||
$content[] = $node;
|
||||
$ws_accum =& $content;
|
||||
}
|
||||
break;
|
||||
case 'colgroup':
|
||||
case 'col':
|
||||
$cols[] = $node;
|
||||
$ws_accum =& $cols;
|
||||
break;
|
||||
case '#PCDATA':
|
||||
// How is whitespace handled? We treat is as sticky to
|
||||
// the *end* of the previous element. So all of the
|
||||
// nonsense we have worked on is to keep things
|
||||
// together.
|
||||
if (!empty($node->is_whitespace)) {
|
||||
$ws_accum[] = $node;
|
||||
}
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
if (empty($content)) {
|
||||
return false;
|
||||
}
|
||||
// INVARIANT: all members of content are non-empty. This can
|
||||
// be shown by observing when things are pushed onto content:
|
||||
// they are only ever pushed when is_collecting is true, and
|
||||
// collection is the only thing ever pushed; but it is known
|
||||
// that collections are non-empty when is_collecting is true.
|
||||
|
||||
$ret = array();
|
||||
$ret = $initial_ws;
|
||||
if ($caption !== false) {
|
||||
$ret = array_merge($ret, $caption);
|
||||
$ret[] = $caption;
|
||||
$ret = array_merge($ret, $after_caption_ws);
|
||||
}
|
||||
if ($cols !== false) {
|
||||
foreach ($cols as $token_array) {
|
||||
$ret = array_merge($ret, $token_array);
|
||||
}
|
||||
$ret = array_merge($ret, $cols);
|
||||
}
|
||||
if ($thead !== false) {
|
||||
$ret = array_merge($ret, $thead);
|
||||
$ret[] = $thead;
|
||||
$ret = array_merge($ret, $after_thead_ws);
|
||||
}
|
||||
if ($tfoot !== false) {
|
||||
$ret = array_merge($ret, $tfoot);
|
||||
$ret[] = $tfoot;
|
||||
$ret = array_merge($ret, $after_tfoot_ws);
|
||||
}
|
||||
|
||||
if ($tbody_mode) {
|
||||
// a little tricky, since the start of the collection may be
|
||||
// whitespace
|
||||
$inside_tbody = false;
|
||||
foreach ($content as $token_array) {
|
||||
// find the starting token
|
||||
// INVARIANT: token_array is not empty
|
||||
$t = NULL;
|
||||
foreach ($token_array as $t) {
|
||||
if ($t->name === 'tr' || $t->name === 'tbody') {
|
||||
break;
|
||||
// we have to shuffle tr into tbody
|
||||
$current_tr_tbody = null;
|
||||
|
||||
foreach($content as $node) {
|
||||
switch ($node->name) {
|
||||
case 'tbody':
|
||||
$current_tr_tbody = null;
|
||||
$ret[] = $node;
|
||||
break;
|
||||
case 'tr':
|
||||
if ($current_tr_tbody === null) {
|
||||
$current_tr_tbody = new HTMLPurifier_Node_Element('tbody');
|
||||
$ret[] = $current_tr_tbody;
|
||||
}
|
||||
} // iterator variable carries over
|
||||
if ($t->name === 'tr') {
|
||||
if ($inside_tbody) {
|
||||
$ret = array_merge($ret, $token_array);
|
||||
$current_tr_tbody->children[] = $node;
|
||||
break;
|
||||
case '#PCDATA':
|
||||
assert($node->is_whitespace);
|
||||
if ($current_tr_tbody === null) {
|
||||
$ret[] = $node;
|
||||
} else {
|
||||
$ret[] = new HTMLPurifier_Token_Start('tbody');
|
||||
$ret = array_merge($ret, $token_array);
|
||||
$inside_tbody = true;
|
||||
$current_tr_tbody->children[] = $node;
|
||||
}
|
||||
} elseif ($t->name === 'tbody') {
|
||||
if ($inside_tbody) {
|
||||
$ret[] = new HTMLPurifier_Token_End('tbody');
|
||||
$inside_tbody = false;
|
||||
$ret = array_merge($ret, $token_array);
|
||||
} else {
|
||||
$ret = array_merge($ret, $token_array);
|
||||
}
|
||||
} else {
|
||||
trigger_error("tr/tbody in content invariant failed in Table ChildDef", E_USER_ERROR);
|
||||
break;
|
||||
}
|
||||
}
|
||||
if ($inside_tbody) {
|
||||
$ret[] = new HTMLPurifier_Token_End('tbody');
|
||||
}
|
||||
} else {
|
||||
foreach ($content as $token_array) {
|
||||
// invariant: everything in here is <tr>s
|
||||
$ret = array_merge($ret, $token_array);
|
||||
}
|
||||
$ret = array_merge($ret, $content);
|
||||
}
|
||||
|
||||
if (!empty($collection) && $is_collecting == false) {
|
||||
// grab the trailing space
|
||||
$ret = array_merge($ret, $collection);
|
||||
}
|
||||
|
||||
array_pop($tokens_of_children); // remove phantom token
|
||||
|
||||
return ($ret === $tokens_of_children) ? true : $ret;
|
||||
return $ret;
|
||||
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user