What causes "Cannot create view %s.%s that references temporary %s: %s" in apache/iceberg?

Persist the temp view's source as a table or permanent view first, then reference that in the new view definition.

apache/iceberg · error · IcebergAnalysisException

Cannot create view %s.%s that references temporary %s: %s

Error message

Cannot create view %s.%s that references temporary %s: %s

What it means

Iceberg's Spark 4.0 RewriteViewCommands rule throws this AnalysisException when a CREATE VIEW on a view catalog (e.g., Iceberg's REST view catalog) resolves to a plan referencing session-scoped temporary views. Permanent views must be reproducible in any session, so temporary views in their definition are rejected. verifyTemporaryObjectsDontExist performs the check during analysis.

Source

Thrown at spark/v4.0/spark-extensions/src/main/scala/org/apache/spark/sql/catalyst/analysis/RewriteViewCommands.scala:135

      case UnresolvedIdentifier(CatalogAndIdentifier(catalog, ident), _)
          if ViewUtil.isViewCatalog(catalog) =>
        Some(ResolvedIdentifier(catalog, ident))

      case _ =>
        None
    }
  }

  /**
   * Permanent views are not allowed to reference temp objects
   */
  private def verifyTemporaryObjectsDontExist(
      identifier: ResolvedIdentifier,
      child: LogicalPlan): Unit = {
    val tempViews = collectTemporaryViews(child)
    if (tempViews.nonEmpty) {
      throw invalidRefToTempObject(
        identifier,
        tempViews.map(v => v.quoted).mkString("[", ", ", "]"),
        "view")
    }

    val tempFunctions = collectTemporaryFunctions(child)
    if (tempFunctions.nonEmpty) {
      throw invalidRefToTempObject(identifier, tempFunctions.mkString("[", ", ", "]"), "function")
    }
  }

  private def invalidRefToTempObject(
      ident: ResolvedIdentifier,
      tempObjectNames: String,
      tempObjectType: String) = {
    new IcebergAnalysisException(
      String.format(
        "Cannot create view %s.%s that references temporary %s: %s",

View on GitHub (pinned to 86d9c8fc54)

Solutions

Persist the temp view's source as a table or permanent view first, then reference that in the new view definition.
Inline the temp view's SELECT statement into the new view's body.
Register temp views' data in the Iceberg catalog rather than the session catalog.
Use CREATE TEMPORARY VIEW for the outer view too if it is genuinely session-scoped.

Example fix

// before (fails)
CREATE TEMPORARY VIEW staging AS SELECT * FROM raw.events;
CREATE VIEW rest_cat.db.final_view AS SELECT * FROM staging;

// after (works)
CREATE TABLE rest_cat.db.staging AS SELECT * FROM raw.events;
CREATE VIEW rest_cat.db.final_view AS SELECT * FROM rest_cat.db.staging;

Defensive patterns

Strategy: validation

Validate before calling

// Spark 4.0: verify the view's source plan contains no temp views before CREATE VIEW
val analyzed = spark.sessionState.executePlan(df.queryExecution.logical).analyzed
val temps = analyzed.flatMap {
  case r: org.apache.spark.sql.catalyst.analysis.UnresolvedRelation
    if spark.sessionState.catalog.isTempView(r.multipartIdentifier) => Seq(r.multipartIdentifier)
  case _ => Seq.empty
}
if (temps.nonEmpty) throw new IllegalStateException(s"resolve temp views first: $temps")

Type guard

def bodyReferencesTempView(viewSql: String, spark: SparkSession): Boolean =
  spark.sessionState.catalog.listTempViews().exists(v => viewSql.contains(v.name))

Prevention

Treat session temp views as ephemeral staging; never build catalog views on them.
Persist the staging result with CREATE TABLE before defining a permanent view.
Inline temp view definitions into the permanent view's SQL where feasible.
Enforce in code review: CREATE VIEW bodies must reference only catalog-qualified objects.
Re-run view creation in a clean session as a CI check.

When it happens

Trigger: `CREATE VIEW iceberg_catalog.db.v AS SELECT ... FROM temp_view` where temp_view is a session temp view (CREATE TEMPORARY VIEW). The exception message includes the quoted temp view names and the target view identifier.

Common situations: Notebook pipelines that stage data as temp views then persist a 'final' Iceberg view referencing them; CI jobs that recreate temp views per run and then create permanent views; migrations from Spark session views to Iceberg views where the body was copied verbatim.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/9fbac6746a200e59. Report an issue: GitHub.