[
https://issues.apache.org/jira/browse/AVRO-3786?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17745628#comment-17745628
]
Evan Blackwell commented on AVRO-3786:
--------------------------------------
I'll keep that in mind if I need to make more examples going forward.
The large test in the last comment is there more to serve to see if there's
even more issues hiding behind the one that I broke down in smaller tests in
the second to last comment. The issue seems to be with defined_in_record, or
more generally with enums defined in a record in a union (compared to enums
used as a reference in a record in a union which is what the ticket originally
highlighted).
I edited my comment with all the smaller tests to provide better explanation
and highlight what could be the problem. It might actually be two problems:
* Failing to deserialize if the reader is given an unknown symbol, even if it
has a default. Similar to this other ticket but the enum is defined in the
record instead of as a reference https://issues.apache.org/jira/browse/AVRO-3787
* Failing to deserialize if the reader and writer both know the symbol but
have it at different indices (and possibly only errors if the defaults also
have different indices)
> [Rust] Deserialization results in FindUnionVariant error if the writer and
> reader have the same symbol but at different positions
> ---------------------------------------------------------------------------------------------------------------------------------
>
> Key: AVRO-3786
> URL: https://issues.apache.org/jira/browse/AVRO-3786
> Project: Apache Avro
> Issue Type: Bug
> Components: rust
> Reporter: Evan Blackwell
> Assignee: Martin Tzvetanov Grigorov
> Priority: Major
> Labels: pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> When calling from_avro_datum with a reader schema that has the same enum
> symbol but in a different position as the writer schema, a FindUnionVariant
> error occurs. The value in the message is available in both schemas. The enum
> is defined elsewhere in the schema, so it's worth checking if the problem is
> with the reference type or the enum itself.
>
> Sample test
> {code:java}
> #[test]
> fn deserialize_union_with_different_enum_order() -> TestResult {
> #[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize,
> serde::Serialize)]
> pub struct BarUseParent {
> #[serde(rename = "barUse")]
> pub bar_use: Bar,
> }
> #[derive(Debug, PartialEq, Eq, Hash, PartialOrd, Ord, Clone,
> serde::Deserialize, serde::Serialize)]
> pub enum Bar {
> #[serde(rename = "bar0")]
> Bar0,
> #[serde(rename = "bar1")]
> Bar1,
> #[serde(rename = "bar2")]
> Bar2,
> }
> #[derive(Debug, PartialEq, Eq, Clone, serde::Deserialize,
> serde::Serialize)]
> pub struct Foo {
> #[serde(rename = "barInit")]
> pub bar_init: Bar,
> #[serde(rename = "barUseParent")]
> pub bar_use_parent: Option<BarUseParent>,
> }
> let writer_schema = r#"{
> "type": "record",
> "name": "Foo",
> "fields":
> [
> {
> "name": "barInit",
> "type":
> {
> "type": "enum",
> "name": "Bar",
> "symbols":
> [
> "bar0",
> "bar1",
> "bar2"
> ],
> "default": "bar0"
> }
> },
> {
> "name": "barUseParent",
> "type": [
> "null",
> {
> "type": "record",
> "name": "BarUseParent",
> "fields": [
> {
> "name": "barUse",
> "type": "Bar"
> }
> ]
> }
> ]
> }
> ]
> }"#;
> let reader_schema = r#"{
> "type": "record",
> "name": "Foo",
> "fields":
> [
> {
> "name": "barInit",
> "type":
> {
> "type": "enum",
> "name": "Bar",
> "symbols":
> [
> "bar1"
> ],
> "default": "bar1"
> }
> },
> {
> "name": "barUseParent",
> "type": [
> "null",
> {
> "type": "record",
> "name": "BarUseParent",
> "fields": [
> {
> "name": "barUse",
> "type": "Bar"
> }
> ]
> }
> ]
> }
> ]
> }"#;
> let writer_schema = Schema::parse_str(writer_schema)?;
> let foo = Foo {
> bar_init: Bar::Bar1,
> bar_use_parent: Some(BarUseParent { bar_use: Bar::Bar1} ),
> };
> let avro_value = crate::to_value(foo)?;
> assert!(
> avro_value.validate(&writer_schema),
> "value is valid for schema",
> );
> let datum = crate::to_avro_datum(&writer_schema, avro_value)?;
> let mut x = &datum[..];
> let reader_schema = Schema::parse_str(reader_schema)?;
> let deser_value = crate::from_avro_datum(&writer_schema, &mut x,
> Some(&reader_schema))?;
> match deser_value {
> types::Value::Record(fields) => {
> assert_eq!(fields.len(), 2);
> assert_eq!(fields[0].0, "barInit");
> assert_eq!(fields[0].1, types::Value::Enum(1,
> "bar1".to_string()));
> assert_eq!(fields[1].0, "barUseParent");
> // TODO: assert barUseParent value
> }
> _ => panic!("Expected Value::Record"),
> } Ok(())
> } Ok(()) } {code}
> Resulting error:
> {code:java}
> apache_avro::error::Error: Could not find matching type in union {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
