W3C home > Mailing lists > Public > html-tidy@w3.org > July to September 1999

Recovering from an unexpected "<"

From: Randy Waki <rwaki@sun10.whizbanglabs.com>
Date: Mon, 13 Sep 1999 15:50:48 -0600
To: "HTML Tidy Mailing List" <html-tidy@w3.org>
Message-ID: <000101befe32$09e2f000$ce9946a6@whizbanglabs.com>
It would be nice if Tidy could recover from an unexpected "<" in the same
way as IE and Netscape.

Described below are what I think are the changes needed to do this.  I'm
using the 26-Jul-99 version of Java Tidy.  There should be a close
correspondence to the C version; my apologies for being unable to provide
the exact changes for C.  These changes seem to work for us although I'm a
bit concerned that I may have missed something because Tidy makes a special
effort to escalate an unexpected "<" from a warning to an error.

It looks like there are two cases to consider: "<" in attribute names and
"<" in attribute values.

For example,

-------- Example input document --------
<html <BAD1>
<head>
<title>t</title>
</head>
<body>
<a href=foo.fee<BAD2>This is a link</a>
</body>
</html>
----------------------------------------

26-Jul-99 Tidy issues errors for the "<" before "BAD1" and "BAD2", plus some
spurious warnings/errors, and does not produce an output document.  After
the changes described below, Tidy issues a warning about the unknown
attribute "<BAD1" and produces the following (which seems to correspond to
IE 5's and Netscape 4.5's interpretation):

-------- Example output document -------
<html>
<head>
<title>t</title>
</head>
<body>
<a href="foo.fee&lt;BAD2">This is a link</a>
</body>
</html>
----------------------------------------

The changes are as follows.

First, for "<" in attribute names, search Lexer.java (lexer.c) for the first
occurrence of "UNEXPECTED_GT" and change:

    this.in.ungetChar(c);
    Report.attrError(this, this.token, null, Report.UNEXPECTED_GT);
    return null;

to:

    // The '<' is unexpected; recover by skipping to the next '>',
    // '/', or whitespace character (mimics IE and Netscape).
    StringBuffer buf = new StringBuffer();
    buf.append('<');
    while (c != '>' && c != '/' && c != StreamIn.EndOfStream)
    {
        map = MAP((char)c);
        if ((map & WHITE) != 0)
            break;

        buf.append((char)c);
        c = this.in.readChar();
    }
    this.in.ungetChar(c);
    Report.attrError(this, this.token, buf.toString(),
Report.UNKNOWN_ATTRIBUTE);
    continue;

Second, for "<" in attribute values, search Lexer.java (lexer.c) for the
remaining occurrence of "UNEXPECTED_GT" and change:

    if (c == '<')
    {
        this.in.ungetChar(c);
        Report.attrError(this, this.token, null, Report.UNEXPECTED_GT);
        break;
    }

to:

    // Note: Accept unescaped '<' in attribute values (seems to mimic IE
    // and Netscape).

In addition, once these two changes have been made, it should be possible to
remove everything associated with the reporting of the "UNEXPECTED_GT" error
although I haven't tried to do so.

Randy
Received on Monday, 13 September 1999 17:51:26 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 April 2012 06:13:42 GMT