diff options
| author | Andrew Balholm <andybalholm@gmail.com> | 2011-11-01 11:42:54 +1100 |
|---|---|---|
| committer | Nigel Tao <nigeltao@golang.org> | 2011-11-01 11:42:54 +1100 |
| commit | 9db3f78c392643769fd46fc7900a6deb1fd2692f (patch) | |
| tree | c094631796fafea58dc427c208678bb2e017e82b | |
| parent | cae23f036ac639141153a73551bcabbb7169e9e0 (diff) | |
| download | go-9db3f78c392643769fd46fc7900a6deb1fd2692f.tar.gz go-9db3f78c392643769fd46fc7900a6deb1fd2692f.zip | |
html: process </td> tags; foster parent at most one node per token
Correctly close table cell when </td> is read.
Because of reconstructing the active formatting elements, more than one
node may be created when reading a single token.
If both nodes are foster parented, they will be siblings, but the first
node should be the parent of the second.
Pass tests1.dat, test 77:
<a href="blah">aba<table><a href="foo">br<tr><td></td></tr>x</table>aoe
| <html>
| <head>
| <body>
| <a>
| href="blah"
| "aba"
| <a>
| href="foo"
| "br"
| <a>
| href="foo"
| "x"
| <table>
| <tbody>
| <tr>
| <td>
| <a>
| href="foo"
| "aoe"
R=nigeltao
CC=golang-dev
https://golang.org/cl/5305074
| -rw-r--r-- | src/pkg/html/parse.go | 8 | ||||
| -rw-r--r-- | src/pkg/html/parse_test.go | 6 |
2 files changed, 10 insertions, 4 deletions
diff --git a/src/pkg/html/parse.go b/src/pkg/html/parse.go index 49c7a65bab..2538ea9811 100644 --- a/src/pkg/html/parse.go +++ b/src/pkg/html/parse.go @@ -126,6 +126,7 @@ func (p *parser) addChild(n *Node) { // fosterParent adds a child node according to the foster parenting rules. // Section 11.2.5.3, "foster parenting". func (p *parser) fosterParent(n *Node) { + p.fosterParenting = false var table, parent *Node var i int for i = len(p.oe) - 1; i >= 0; i-- { @@ -986,7 +987,12 @@ func inCellIM(p *parser) (insertionMode, bool) { case EndTagToken: switch p.tok.Data { case "td", "th": - // TODO. + if !p.popUntil(tableScopeStopTags, p.tok.Data) { + // Ignore the token. + return inCellIM, true + } + p.clearActiveFormattingElements() + return inRowIM, true case "body", "caption", "col", "colgroup", "html": // TODO. case "table", "tbody", "tfoot", "thead", "tr": diff --git a/src/pkg/html/parse_test.go b/src/pkg/html/parse_test.go index 861070d103..caf3c92bc7 100644 --- a/src/pkg/html/parse_test.go +++ b/src/pkg/html/parse_test.go @@ -132,7 +132,7 @@ func TestParser(t *testing.T) { rc := make(chan io.Reader) go readDat(filename, rc) // TODO(nigeltao): Process all test cases, not just a subset. - for i := 0; i < 77; i++ { + for i := 0; i < 78; i++ { // Parse the #data section. b, err := ioutil.ReadAll(<-rc) if err != nil { @@ -161,8 +161,8 @@ func TestParser(t *testing.T) { continue } // Check that rendering and re-parsing results in an identical tree. - if filename == "tests1.dat" && i == 30 { - // Test 30 in tests1.dat is such messed-up markup that a correct parse + if filename == "tests1.dat" && (i == 30 || i == 77) { + // Some tests in tests1.dat have such messed-up markup that a correct parse // results in a non-conforming tree (one <a> element nested inside another). // Therefore when it is rendered and re-parsed, it isn't the same. // So we skip rendering on that test. |