[AVRO-3528] Optionally support strict LogicalType parsing

• Type: Improvement
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 1.11.0
• Fix Version/s: 1.13.0
• Component/s: java
• Labels: pull-request-available

Language

java

• Assignee: Andrei Leibovski
• Reporter: Andrei Leibovski
• Created: 02/Jun/22 22:52
• Updated: 05/Aug/24 17:25

Description

My organization uses Avro schemas extensively. We use Confluent schema registry for data governance, enforcing data contracts between various components. We are seeing proliferation of questionable LogicalType structures within our schemas, like the following:

{
  "namespace": "org.apache.avro.example",
  "type": "record",
  "name": "BadLogical",
  "fields": [
   {
     "name": "f0",
     "type":{
       "type": "string",
       "java-class": "java.math.BigDecimal",
       "logicalType": "decimal",
       "precision": 9,
       "scale": 2
     }
   }
  ]
} 

There are two issues in the above structure:

1. string is not allowed to back the decimal LogicalType
2. java-class property and some others, are incompatible for any LogicalType

Currently Avro allows such structures to pass validation and has no option to disallow them. Since Confluent schema registry delegates all avro schema validation to Avro, these structures are allowed to be registered.

Proposition:

Implement an option to switch avro Schema.Parser to LogicalType strict mode where such structures will be detected and disallowed.

The change is fairly trivial, most of the plumbing is already there. I have implemented the required functionality and covered it with with tests, but I would like to solicit some feedback on this proposal before submitting my PR.

Issue Links

• links to: GitHub Pull Request #1752