Dear Spark Community, I’m currently managing a data platform that uses Trino with Hive Metastore integration. Our Hive Metastore contains a mix of legacy Hive tables and views, alongside newer views created via Trino.
As expected, Trino stores views in the metastore with viewOriginalText containing a Base64-encoded JSON blob (e.g., /* Presto View */<base64>). This format is not directly readable by Spark when using a session with .enableHiveSupport() and the config spark.sql.hive.convertMetastoreView=true (or equivalently, spark.enableHive=true), as Spark expects standard SQL strings in the metastore view definition. When attempting to read a Trino-created view in Spark, Spark fails because it cannot parse the encoded format stored by Trino. My question is: - Is there any built-in or recommended way to allow Spark to interpret or access Trino views within a Spark session using enableHiveSupport()? - Alternatively, is there any planned support or extension mechanism that could allow Spark to decode Trino views at runtime? Currently, the only workaround is to extract and decode the Base64 manually in Python or Scala and recreate the view within the session — which is not scalable in a multi-user, interactive Spark environment. Would appreciate any guidance, best practices, or roadmap insights related to this use case. Best regards.
