java.lang.ClassCastException: org.apache.spark.sql.catalyst.expressions.GenericRow cannot be cast to org.apache.avro.generic.GenericRecord about abris

[cerveada]

Fixed in 3.0.2

[Inylschek]

I got the exact same error with 2_11, 3.0.1 when calling from_confluent_avro with a schema string.

Not sure if it's a bug or if I'm not using this correctly, but since this issue has been created, I'll hitch a ride on it

[cerveada]

Hi, thanks for repporting the error. Could any of you please provide more information? What schema were you using? How have you serialized the data?

[Inylschek]

My schema is very complex. Not sure about serialization. Just reading from a Kafka topic.

I'm just investigating what happens for a simpler schema first before coming back with more details...

[Inylschek]

Using from_confluent_avro, I have managed to read from a different Kafka topic with a different avro schema. This schema is much simpler than the schema that fails.

I amn't sure which serialization is being used, but I know that both the simple and complex schema are written in the same way.

Here is the complex schema that is failing

root
 |-- from_avro(value): struct (nullable = true)
 |    |-- a: string (nullable = false)
 |    |-- b: string (nullable = false)
 |    |-- c: string (nullable = false)
 |    |-- d: long (nullable = false)
 |    |-- e: integer (nullable = false)
 |    |-- f: struct (nullable = false)
 |    |    |-- g: struct (nullable = false)
 |    |    |    |-- value: double (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- h: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- i: struct (nullable = false)
 |    |    |    |-- value: string (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- j: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- k: struct (nullable = false)
 |    |    |    |-- value: integer (nullable = false)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- l: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- m: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- n: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- o: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- p: struct (nullable = false)
 |    |    |    |-- value: boolean (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- q: struct (nullable = false)
 |    |    |    |-- value: integer (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- r: struct (nullable = false)
 |    |    |    |-- value: integer (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- s: struct (nullable = false)
 |    |    |    |-- value: float (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- t: struct (nullable = false)
 |    |    |    |-- value: double (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)
 |    |    |-- u: struct (nullable = false)
 |    |    |    |-- value: integer (nullable = true)
 |    |    |    |-- tsMs: long (nullable = false)
 |    |    |    |-- status: integer (nullable = false)
 |    |    |    |-- changed: boolean (nullable = false)

I can confirm that the kafka topic containing the complex schema is being read by a microservice without any problems, so the data itself is not corrupt.

[Inylschek]

Also, we are using Confluent's schema registry for encoding, however I am still using from_confluent_avro with a jsonSchema string derived by directly reading and parsing the avsc file. In other words, the jsonSchema I am using does not include any additional information that Confluent adds.

At the moment I am still doing the proof of concept and as the schema registry is stuck behind some layers of security, I am passing in a schema string instead to make things easier

However, as I was able to parse data from the kafka topic with the simpler schema, I don't believe that this is a problem

[heng2018]

we use confluent schema registration
our code originally run on databricks notebook and we use databricks avro.functions.from_avro.
now we migrate the code off databricks and replace that part with ABRiS from_confluent_avro with schema registry signature, same producer , topic and messages and got this error

[Inylschek]

I've just seen on the wiki that abris does not support batch processing which is what I am trying to do (but I did get it to work for the simple schema!)

Interested to know if that could be the problem? Even if it is, heng2018 still has a problem here.

[cerveada]

I've just seen on the wiki that abris does not support batch processing which is what I am trying to do (but I did get it to work for the simple schema!)

The WIKI.wiki file is outdated. It is about the old API from Abris 2.x.x. The current API should work anywhere where you can do select.

Good news is that I was able to simulate this exception. The fix will be part of next release.

[heng2018]

great news!
when will the the next release?

[cerveada]

Most probably this week.