Repeated choice plus other field

[Gaelan]

I've got some XML that looks like this:

<outer>
  <one />
  <two />
  <three />
  <one />
  <field>5</field> <!-- optional -->
</outer>

Which I want to parse with something like:

#[derive(Deserialize)]
#[serde(rename_all = "snake_case")]
enum Choice { One, Two, Three }
#[derive(Deserialize)]
struct Outer {
  #[serde(rename_all = "$value")]
  choices: Vec<Choice>,
  field: Option<u32>
}

But this fails, with:

unknown variant `field`, expected one of `One`, `Two`, `Three`

Is there a trick to making this work? I could just put field into the Choice enum, but that doesn't really make sense semantically, so I'd rather not.

[Mingun]

Could you post all code that reproduces the problem (ideally as Rust testcase)? Because we have exactly this testcase here:

quick-xml/tests/serde-de-seq.rs

Lines 3326 to 3339 in 64c4249

	/// In those tests non-sequential field is defined in the struct
	/// after sequential, so it will be deserialized after the list.
	/// That struct should be deserialized from the XML where these
	/// fields comes in an arbitrary order
	mod field_after_list {
	use super::*;
	use pretty_assertions::assert_eq;
	#[derive(Debug, PartialEq, Deserialize)]
	struct Root {
	#[serde(rename = "$value")]
	item: Vec<Choice>,
	node: (),
	}

[Gaelan]

Huh, strange.

The data in question is from the UK's train status API.

#[derive(Deserialize, Debug)]
enum ScheduleLocation {
    #[serde(rename = "OR")]
    Origin,
    #[serde(rename = "OPOR")]
    OperationalOrigin,
    #[serde(rename = "PP")]
    PassingPoint,
    #[serde(rename = "IP")]
    IntermediatePoint,
    #[serde(rename = "OPIP")]
    OperationalIntermediatePoint,
    #[serde(rename = "DT")]
    Destination,
    #[serde(rename = "OPDT")]
    OperationalDestination,
}

#[derive(Deserialize, Debug)]
struct Schedule {
    #[serde(rename = "@rid")]
    rid: String,
    cancelReason: Option<u32>,
    #[serde(rename = "$value")]
    locations: Vec<ScheduleLocation>,
}

#[derive(Deserialize, Debug)]
enum Event {
    #[serde(rename = "schedule")]
    Schedule(Schedule),
    #[serde(rename = "association")]
    Association,
    #[serde(rename = "TS")]
    TrainStatus,
    #[serde(rename = "trainOrder")]
    TrainOrder,
    #[serde(rename = "OW")]
    StationMessage,
    #[serde(rename = "trainAlert")]
    TrainAlert,
    #[serde(rename = "trackingID")]
    TrackingId,
    #[serde(rename = "alarm")]
    Alarm,
    #[serde(rename = "scheduleFormations")]
    ScheduleFormations,
    #[serde(rename = "formationLoading")]
    FormationLoading,
}

#[derive(Deserialize, Debug)]
struct UpdateInner {
    #[serde(rename = "$value")]
    events: Vec<Event>,
}

#[derive(Deserialize, Debug)]
enum UpdateType {
    // updates are contained in either an sR or uR element; we don't actually care which
    #[serde(rename = "sR")]
    Sr(UpdateInner),
    #[serde(rename = "uR")]
    Ur(UpdateInner),
}

#[derive(Deserialize, Debug)]
struct PportUpdate {
    #[serde(rename = "$value")]
    inner: UpdateType,
}

#[test]
fn quick_xml_weirdness() {
    let xml = r#"
        <?xml version="1.0" encoding="UTF-8" standalone="yes"?>
        <Pport xmlns="http://www.thalesgroup.com/rtti/PushPort/v16" xmlns:ns2="http://www.thalesgroup.com/rtti/PushPort/Schedules/v3" xmlns:ns3="http://www.thalesgroup.com/rtti/PushPort/Schedules/v2" xmlns:ns4="http://www.thalesgroup.com/rtti/PushPort/Formations/v2" xmlns:ns5="http://www.thalesgroup.com/rtti/PushPort/Forecasts/v3" xmlns:ns6="http://www.thalesgroup.com/rtti/PushPort/Formations/v1" xmlns:ns7="http://www.thalesgroup.com/rtti/PushPort/StationMessages/v1" xmlns:ns8="http://www.thalesgroup.com/rtti/PushPort/TrainAlerts/v1" xmlns:ns9="http://www.thalesgroup.com/rtti/PushPort/TrainOrder/v1" xmlns:ns10="http://www.thalesgroup.com/rtti/PushPort/TDData/v1" xmlns:ns11="http://www.thalesgroup.com/rtti/PushPort/Alarms/v1" xmlns:ns12="http://thalesgroup.com/RTTI/PushPortStatus/root_1" ts="2023-11-16T21:14:11.2579109Z" version="16.0">
            <sR>
                <schedule rid="202311168055118" uid="P55118" trainId="1C61" ssd="2023-11-16" toc="NT" trainCat="XX">
                    <ns2:OR wtd="19:29" tpl="MNCRIAP" act="TB" can="true" ptd="19:29"/>
                    <ns2:PP wtp="19:30:30" tpl="HLDGWJ" can="true"/>
                    <ns2:PP wtp="19:31:30" tpl="HLDG" can="true"/>
                    <ns2:IP wta="19:34" wtd="19:34:30" tpl="GATLEY" act="T " can="true" pta="19:34" ptd="19:34"/>
                    <ns2:PP wtp="19:36" tpl="EDIDBRY" can="true"/>
                    <ns2:PP wtp="19:37" tpl="BAGE" can="true"/>
                    <ns2:PP wtp="19:38" tpl="MLDTHRD" can="true"/>
                    <ns2:PP wtp="19:40:30" tpl="SLDLJN" can="true"/>
                    <ns2:PP wtp="19:44" tpl="ARDWCKJ" can="true"/>
                    <ns2:IP wta="19:47" wtd="19:50" tpl="MNCRPIC" act="T " can="true" pta="19:47" ptd="19:50"/>
                    <ns2:IP wta="19:51:30" wtd="19:53:30" tpl="MNCROXR" act="T " can="true" pta="19:52" ptd="19:53"/>
                    <ns2:IP wta="19:55" wtd="19:56" tpl="MNCRDGT" act="T " can="true" pta="19:55" ptd="19:56"/>
                    <ns2:PP wtp="19:57" tpl="WATSTJN" can="true"/>
                    <ns2:PP wtp="19:58" tpl="ORDSLLJ" can="true"/>
                    <ns2:PP wtp="19:59" tpl="SLFDCT" can="true"/>
                    <ns2:IP wta="20:07:30" wtd="20:08:30" tpl="BOLTON" act="T " can="true" pta="20:08" ptd="20:08"/>
                    <ns2:PP wtp="20:12" tpl="LOSTCKJ" can="true"/>
                    <ns2:PP wtp="20:13" tpl="HORWICH" can="true"/>
                    <ns2:PP wtp="20:15" tpl="BLRD" can="true"/>
                    <ns2:PP wtp="20:16" tpl="ADNL" can="true"/>
                    <ns2:IP wta="20:19:30" wtd="20:20:30" tpl="CHORLEY" act="T " can="true" pta="20:20" ptd="20:20"/>
                    <ns2:PP wtp="20:24" tpl="CHORBUK" can="true"/>
                    <ns2:PP wtp="20:26" tpl="EUXTONJ" can="true"/>
                    <ns2:PP wtp="20:30" tpl="PRSTRJN" can="true"/>
                    <ns2:IP wta="20:31:30" wtd="20:33:30" tpl="PRST" act="T " can="true" pta="20:32" ptd="20:33"/>
                    <ns2:PP wtp="20:34:30" tpl="PRSTNFJ" can="true"/>
                    <ns2:PP wtp="20:38" tpl="BBGHGL" can="true"/>
                    <ns2:PP wtp="20:41" tpl="GSTANG" can="true"/>
                    <ns2:IP wta="20:50" wtd="20:51" tpl="LANCSTR" act="T " can="true" pta="20:50" ptd="20:51"/>
                    <ns2:PP wtp="20:53" tpl="MORCMSJ" can="true"/>
                    <ns2:PP wtp="20:58" tpl="CRNFNJN" can="true"/>
                    <ns2:IP wta="20:59:30" wtd="21:00:30" tpl="CRNF" act="T " can="true" pta="21:00" ptd="21:00"/>
                    <ns2:IP wta="21:05" wtd="21:06" tpl="SDAL" act="T " can="true" pta="21:05" ptd="21:06"/>
                    <ns2:IP wta="21:09:30" wtd="21:10:30" tpl="ARNSIDE" act="T " can="true" pta="21:10" ptd="21:10"/>
                    <ns2:IP wta="21:15" wtd="21:16" tpl="GOVS" act="T " can="true" pta="21:15" ptd="21:16"/>
                    <ns2:IP wta="21:19" wtd="21:20" tpl="KTBK" act="T " can="true" pta="21:19" ptd="21:20"/>
                    <ns2:IP wta="21:23:30" wtd="21:24" tpl="CARK" act="T " can="true" pta="21:24" ptd="21:24"/>
                    <ns2:IP wta="21:30:30" wtd="21:31:30" tpl="ULVRSTN" act="T " can="true" pta="21:31" ptd="21:31"/>
                    <ns2:IP wta="21:39" wtd="21:40" tpl="DALTON" act="T " can="true" pta="21:39" ptd="21:40"/>
                    <ns2:PP wtp="21:41:30" tpl="DALTONJ" can="true"/>
                    <ns2:IP wta="21:45:30" wtd="21:46:30" tpl="ROOSE" act="T " can="true" pta="21:46" ptd="21:46"/>
                    <ns2:DT wta="21:53" pta="21:53" tpl="BAROW" act="TF" can="true"/>
                    <ns2:cancelReason>918</ns2:cancelReason>
                </schedule>
            </sR>
        </Pport>"#;
    let result = quick_xml::de::from_str::<PportUpdate>(xml);
    dbg!(&result);
    assert!(result.is_ok());
}

Results in:

running 1 test
test quick_xml_weirdness ... FAILED

failures:

---- quick_xml_weirdness stdout ----
[src/main.rs:139] &result = Err(
    Custom(
        "unknown variant `cancelReason`, expected one of `OR`, `OPOR`, `PP`, `IP`, `OPIP`, `DT`, `OPDT`",
    ),
)
thread 'quick_xml_weirdness' panicked at 'assertion failed: result.is_ok()', src/main.rs:140:5
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace


failures:
    quick_xml_weirdness

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

[Mingun]

The problem is in namespace prefix. Namespaces are not properly supported in serde yet. If you remove ns2: from cancelReason here, the test is passed:

#[test]
fn issue683() {
    #[derive(Deserialize, Debug, PartialEq)]
    enum ScheduleLocation {
        #[serde(rename = "DT")]
        Destination,
    }

    #[derive(Deserialize, Debug, PartialEq)]
    struct Schedule {
        cancelReason: Option<u32>,
        #[serde(rename = "$value")]
        locations: Vec<ScheduleLocation>,
    }
    let xml = r#"
        <schedule xmlns:ns2="http://www.thalesgroup.com/rtti/PushPort/Schedules/v3">
            <ns2:DT/>
            <ns2:cancelReason>918</ns2:cancelReason>
        </schedule>"#;
    let result = quick_xml::de::from_str::<Schedule>(xml);
    dbg!(&result);
    assert_eq!(result.unwrap(), Schedule {
        cancelReason: Some(918),
        locations: vec![ScheduleLocation::Destination],
    });
}

[Mingun]

Because your XML use namespaces, try to look at xmlserde which should support namespaces better. It is based on quick-xml, so should be as fast as quick-xml's deserializer.

[Gaelan]

Thanks!

(I don't actually care about the namespaces, so I'd be equally happy just stripping them; but there doesn't seem to be an easy way to do this.)

[leftmostcat]

I'm also experiencing this issue and, as with Gaelan, would be happy if namespaces could simply be ignored by the deserializer. xmlserde requires being significantly more explicit about structure, which in my case is impractical.

Changing https://github.com/tafia/quick-xml/blob/master/src/de/map.rs#L792 to use local_name() instead of name() solves the issue for me and all serde_de(_.*)? tests remain passing. Is that a reasonable solution, or is there a situation I'm missing?