What regex expression will operate with Java's "replaceAll" function to remove the html tag and its contents from an html string?

Question

What regex expression will operate together with the Java replaceAll() method to remove the <p> html tag and its contents in between the tag from an HTML string?

For example, after applying the method,

"<div><p>table <b>test</b> title</p><table><tbody><tr><td>this is table cell value</td></tr></tbody></table><p>miscellaneous contents</p><span>blah</span></div>"

becomes:

"<div><table><tbody><tr><td>this is table cell value</td></tr></tbody></table><span>blah</span></div>"

Note: This is an "academic" exercise. I am not seeking a solution that uses an XML/HTML parser.

---

UPDATE:

Getting closer to a solution on this (thanks, jlordo!)... You pattern seems to work somewhat...

However, the suggested regex string ("<[pP]>.*?</[pP]>") does not appear to have an effect on a <p> tag that contains an attribute (i.e., in this case a "style" attribute) -- see below,

    public static void main(String[] args)
    {
        String htmlstring = "<div><p style='text-align: center; font-style: italic'>[click the <b>submit</b> button to create the new company.]</p><table><tbody><tr><td>this is table cell value</td></tr></tbody></table><p>miscellaneous contents</p><span>blah</span></div>";
        htmlstring = htmlstring.replaceAll("<[pP]>.*?</[pP]>", "");
    }

htmlstring (before scrubbing):

<div><p style='text-align: center; font-style: italic'>[click the <b>submit</b> button to create the new company.]</p><table><tbody><tr><td>this is table cell value</td></tr></tbody></table><p>miscellaneous contents</p><span>blah</span></div>

htmlstring (after scrubbing):

<div><p style='text-align: center; font-style: italic'>[click the <b>submit</b> button to create the new company.]</p><table><tbody><tr><td>this is table cell value</td></tr></tbody></table><span>blah</span></div>

Is there anything we can do to "tweak" it so that it handles this issue?

Answer 1

try

    htmlstring = htmlstring.replaceAll("(?i)<p.*?>.*?</p>", "");

note that (?i) means turn on case-insensitive flag

Answer 2

Pattern.compile(
  // A start p tag.
  "<p(?![a-z0-9:\\-])([^>\"']|\"[^\"]*\"|'[^']*)*>"
  + ".*?"   // Phrasing content that does not handle comment, RCDATA or raw text boundaries
  // An end p tag
  + "</p(?![a-z0-9:\\-])[^>]*>",
  Pattern.DOTALL | Pattern.CASE_INSENSITIVE);

The Pattern.DOTALL flag will cause .*? to match newlines which is necessary because your original regex would not match any paragraph that contained a newline in its body.

The Pattern.CASE_INSENSITIVE flag is specified without Pattern.UNICODE_CASE because it's unnecessary and I'm not confident that Turkish case-folding wouldn't create a subtle maintenance hazard were this regex modified to deal with <i>.

The ([^>"']|"[^"]*"|'[^']*) part matches any tag body character or quoted attribute. It will misbehave on certain non-validating attribute names like <p ain't-this=confusing>. The attribute grammar is regular, but doing a full treatment of quote characters in attribute values vs names would hugely expand the size of this regex, and would not likely help since anything requiring a full treatment will have to deal with the fact that backticks can quote attributes on a few browsers which means that no single regular expression can find value boundaries for arbitrarily messy HTML.

The (?![a-z0-9:\\-]) makes sure the name of the tag is "p" and not "plaintext" or "p-" or "p:foo" or some other HTML identifier of which "p" is a prefix.

This may behave on some constructs like:

• <p><!-- </p> -->Not an orphaned end tag</p>
• <p><textarea>Not a paragraph</p></textarea></p>
• <noscript><p>Not a paragraph contextually</p></noscript>
• <p ain't-this=confusing>Foo</p> <p>Isn't recognized as separate</p>.
• <p><script>alert("Not a real </p> tag");</script></p>