CDATA deserialization

[Mingun]

It seems that CDATA parsing performed incorrectly. CDATA should contain logically unescaped content (i.e. you don't need to unescape it). So the string <![CDATA[escaped&#x20;string]]> should be parsed into string escaped string in contrast to the ordinal text content where the result should be escaped string.

According to the https://www.w3.org/TR/xml/#sec-cdata-sect CDATA sections cannot contain any escape sequences. I've checked several online validators (for example https://jsonformatter.org/xml-parser) and they agree with that.

However, quick-xml has different opinion. This test seems invalid:

quick-xml/tests/unit_tests.rs

Lines 179 to 184 in 516f1ea

	#[test]
	fn test_cdata_open_close() {
	let mut r = Reader::from_str("<![CDATA[test <> test]]>");
	r.trim_text(true);
	next_eq!(r, CData, b"test &lt;&gt; test");
	}

As a consequence, all CDATA processing in the Deserializer is broken because it is simply handled the same as the Text event, and in several places handling of the CDATA just absent.

I preparing a PR with these fixes, but because I'm not working with XML very often, I'm not sure that I'm not wrong. So I open this issue to get feedback.

[iovxw]

From #244:

The BytesText buffer should always contain escaped string

That's good API design but bad for performance, maybe we need something like quick_xml::events::BytesTextUnescaped or quick_xml::events::BytesCdata?

[OliverBrotchie]

This is quite unintuitive behaviour. Is there a quick workaround for this?

From #244:

The BytesText buffer should always contain escaped string

That's good API design but bad for performance, maybe we need something like quick_xml::events::BytesTextUnescaped or quick_xml::events::BytesCdata?

BytesTextUnescaped defiantly seems like a good solution.

[OliverBrotchie]

Btw this works with the current API:

use quick_xml::{events::Event, Reader, Writer};
use std::{
    fs,
    io::{self, Cursor},
};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let file = fs::read_to_string("foo.xml")?;
    let mut r = Reader::from_str(&path);
    let mut w = Writer::new(Cursor::new(Vec::new()));
    let mut buf = Vec::<u8>::new();

    // Loop over the xml tags
    loop {
        match r.read_event(&mut buf) {
        
            // Special case for CDATA
            Ok(Event::CData(e)) => {
                w.write(format!(
                    "<![CDATA[{}]]>\n",
                    str::from_utf8(&e.unescaped()?.into_owned())?
                ).as_bytes())?;
            } 
            Ok(Event::Eof) => break,
            Ok(e) => w.write_event(e)?,
            Err(e) => panic!("Error at position {}: {:?}", r.buffer_position(), e),
        }
        buf.clear();
    }
    fs::write("foo.xml", w.into_inner().into_inner())?;
    Ok(())
}

[swi2012]

Hello.
I've got a problem with CData too, but not sure if it's quick-xml problem:
In xml i want to parse i have text inside tags as cdata, but it's look like this (escaped):
<NAME>&lt;![CDATA[GLOBAL RELIEF FOUNDATION (GRF)* ]]&gt;</NAME>
So Event::CData do not trigger, and Event::Start and read_text inside it gives me string
<![CDATA[GLOBAL RELIEF FOUNDATION (GRF)* ]]>.

I'm wonder how can i get actual text from cdata inside tag?

[Mingun]

You have a string representation of one XML inside other XML -- so you need to parse NAME content as an inner XML

[swi2012]

You have a string representation of one XML inside other XML -- so you need to parse NAME content as an inner XML

Thanks. It's worked, but * and , at the end of text remains. Code like clean this garbage for me:

 Ok(Event::CData(t)) => {
                        s = t.unescape_and_decode(&r).unwrap();
                        if !s.is_empty() {
                            let mut ast = Vec::<usize>::new();
                            let tmp: &str = s.trim_end().trim_end_matches(|c| c == '*' || c == ',');
                            if let Some(i) = tmp.find('*') {
                                ast.push(i)
                            }
                            s = tmp.to_string();
                            if ast.len() > 0 {
                                for idx in &ast {
                                    s.remove(*idx);
                                    s.remove(*idx + 1);
                                }
                            }
                            return s;
                        }
                    }