Allow backslash-escaping within ElementSelector and CssIdentifier

[hannibal218bc]

There has been previous discussion about element names or CSS tokens that can currently not been selected through JSoup's selector, because they contain JSoup Selector operators (like #, . , ...).

In #1055 it was suggested to use escape characters within the selector syntax.

This (simple) implementation modifies TokenQueue so consumeElementSelector and consumeCSSIdentifier will continue matching if the preceeding character is a backslash (\). Finally, all backslashes are removed from the result.

It is a simple approach, and will not allow escaping backslashes themselves - but there should never be a backslash in an element name or CSS identifier?

Here's a simple test case:

	@Test public void testSelectorEscapes() {
		Document doc = Jsoup.parse("<root><e.dot>e with dot</e.dot><e class=\"dot\">e with class dot</e><f id=\"i.d\">f with i.d</f><x id=\"i\" class=\"d\">id i with class d</x></root>", "", Parser.xmlParser());

		assertEquals("e with class dot", 	doc.select("e.dot").text());
		assertEquals("e with dot", 			doc.select("e\\.dot").text());
		assertEquals("id i with class d", 	doc.select("#i.d").text());
		assertEquals("f with i.d", 			doc.select("#i\\.d").text());

	}

This PR would close issue #1441 .

[hannibal218bc]

Updated the PR to use TokenQueue.unescape to remove the escape characters, and added a flag to only do it when an escape char was detected in the loop.

[jhy]

Hi Hannes,

Thanks for this, and sorry for the late reply. This update looks great. One suggestion:

It is a simple approach, and will not allow escaping backslashes themselves - but there should never be a backslash in an element name or CSS identifier?

[Un]fortunately, HTML does indeed allow backslashes, both in tag names and in attribute names:

<p\p attri\bute>Foo</p\p>

is valid and parses as-is (both in jsoup and in Chrome)

I think that we should enable the backslashes to be escaped.

LMK if you want to do that, or if I should.

Finally, it would be great if you could update CHANGES and most importantly, add a test-case for each case, to make sure it works and we never regress.

[jhy]

(Somewhat amusingly, <table(╯°□°)╯>Hello!</table(╯°□°)╯> is also a valid tag name, although I wasn't able to create a selector for it, either with jsoup or Chrome. And I don't think that's a requirement for this PR, but a nice i18n stretch goal :) )