Text node split into multiple BytesText events when entities are present

[ImJeremyHe]

I noticed that when parsing XML with quick-xml, a text node can be split into multiple BytesText events if it contains XML entities.

For example:

    #[test]
    fn test_text_node() {
        let xml = r#"<f>_xlfn.DISPIMG(&quot;ID_72CA26DEC13E452487646D77B0F1058F&quot;,1)</f>"#;
        let mut reader = quick_xml::Reader::from_reader(xml.as_bytes());
        let mut buf = Vec::<u8>::new();
        loop {
            match reader.read_event_into(&mut buf) {
                Ok(Event::Eof) => break,
                Ok(e) => {
                    println!("{:?}", e);
                }
                _ => {}
            }
        }
    }

The output is like:

Start(BytesStart { buf: Borrowed("f"), name_len: 1 })
Text(BytesText { content: Borrowed("_xlfn.DISPIMG(") })
GeneralRef(BytesRef { content: Borrowed("quot") })
Text(BytesText { content: Borrowed("ID_72CA26DEC13E452487646D77B0F1058F") })
GeneralRef(BytesRef { content: Borrowed("quot") })
Text(BytesText { content: Borrowed(",1)") })
End(BytesEnd { name: Borrowed("f") })

produces multiple BytesText events instead of a single one containing the entire text.

Is this intentional? If so, is there a recommended way to reconstruct the complete text node from the emitted events? I was using v0.39 which didn't have this issue

[Mingun]

Yes, this is intentional. You need to expand GeneralRef references manually, which in general case may expand to another XML fragment. That is one of breaking API change in 0.40. The simplified way to process this demonstrated in custom_entities.rs, but it cannot handle namespaces correctly. The right approach you can find in #948, but I'm not sure that it is possible to implement it outside of quick-xml right now (maybe some private API is used).