SchemaCompatibility.Resolve incorrectly resolves enums #556
Description
When using schema resolution with enums that have reordered or removed symbols, resolved schema is incorrect if writer does not have new incompatible symbols.
Expected behavior
When I have these two schemas, I expect to be able to correctly parse data after performing schema resolution.
Reader schema:
{"type": "enum", "name": "Enum", "symbols": ["A", "B", "C", "D"]}Writer schema:
{"type": "enum", "name": "Enum", "symbols": ["B", "C", "D"]}Actual behavior
The parsed data uses the reader schema's symbol positions to parse the data, instead of the writer schema's positions.
Example code to reproduce the issue
Playground link: https://go.dev/play/p/6nuMheyX0j9
package main
import (
"fmt"
"github.com/hamba/avro/v2"
)
func main() {
reader := avro.MustParse(`{"type": "enum", "name": "Enum", "symbols": ["A", "B", "C", "D"]}`)
writer := avro.MustParse(`{"type": "enum", "name": "Enum", "symbols": ["B", "C", "D"]}`)
data, err := avro.Marshal(writer, "D")
if err != nil {
panic(err)
}
sc := avro.NewSchemaCompatibility()
rs, err := sc.Resolve(reader, writer)
if err != nil {
panic(err)
}
var value string
err = avro.Unmarshal(rs, data, &value)
if err != nil {
panic(err)
}
// Expected output: "Resolved schema: D"
// Observed output: "Resolved schema: C"
fmt.Println("Resolved schema:", value)
}Probable root cause
I believe this issue is caused by this if statement in the resolution logic:
Line 341 in 07bcb49
checkEnumSymbols method only checks for names of the symbols, but not positions, so if all names are compatible, but positions are not, the reader schema is incorrectly returned as-is. A possible solution might be setting the encodedSymbols field when order is mismatched, instead of simply returning the reader schema.