prestodb/presto · error · IllegalStateException

Group By column (%s) definition not found in input selection

Error message

Group By column (%s) definition not found in input selections: %s

What it means

While building the column-index mapping from Pinot's output order to Presto's expected order, getIndicesMappingFromPinotSchemaToPrestoSchema looks up each group-by column in the accumulated selections. A missing definition means the group-by variable was never registered as a selection, indicating an internal inconsistency in the generated plan/context; it is thrown as IllegalStateException (no specific error code).

Source

Thrown at presto-pinot-toolkit/src/main/java/com/facebook/presto/pinot/query/PinotQueryGeneratorContext.java:411

        return new PinotQueryGenerator.GeneratedPinotQuery(tableName, query, indices, filter.isPresent(), forBroker);
    }

    private String updateSelection(String definition, ConnectorSession session)
    {
        final String overrideDistinctCountFunction = PinotSessionProperties.getOverrideDistinctCountFunction(session);
        if (!PINOT_DISTINCT_COUNT_FUNCTION_NAME.equalsIgnoreCase(overrideDistinctCountFunction)) {
            return definition.replaceFirst(PINOT_DISTINCT_COUNT_FUNCTION_NAME.toUpperCase() + "\\(", overrideDistinctCountFunction.toUpperCase() + "\\(");
        }
        return definition;
    }

    private List<Integer> getIndicesMappingFromPinotSchemaToPrestoSchema(String query, Map<VariableReferenceExpression, PinotColumnHandle> assignments)
    {
        LinkedHashMap<VariableReferenceExpression, Selection> expressionsInPinotOrder = new LinkedHashMap<>();
        for (VariableReferenceExpression groupByColumn : groupByColumns) {
            Selection groupByColumnDefinition = selections.get(groupByColumn);
            if (groupByColumnDefinition == null) {
                throw new IllegalStateException(format(
                        "Group By column (%s) definition not found in input selections: %s",
                        groupByColumn,
                        Joiner.on(",").withKeyValueSeparator(":").join(selections)));
            }
            expressionsInPinotOrder.put(groupByColumn, groupByColumnDefinition);
        }
        for (VariableReferenceExpression outputColumn : outputs) {
            Selection outputColumnDefinition = selections.get(outputColumn);
            if (outputColumnDefinition == null) {
                throw new IllegalStateException(format(
                        "Output column (%s) definition not found in input selections: %s",
                        outputColumn,
                        Joiner.on(",").withKeyValueSeparator(":").join(selections)));
            }
            expressionsInPinotOrder.put(outputColumn, outputColumnDefinition);
        }

        checkSupported(

View on GitHub (pinned to 55bb57d202)

Solutions

1. Rewrite the query to group by a plain column reference from the table instead of a derived/renamed expression alias.
2. If derived grouping is needed, materialize it via a supported project pattern the connector recognizes, or fall back to Presto-side grouping.
3. Capture the query and the selections listed in the message and file a bug against the Pinot connector; this indicates a generator bug.

Example fix

// before
SELECT user_region AS region, count(*) FROM events GROUP BY region; // alias not in selections
// after
SELECT user_region, count(*) FROM events GROUP BY user_region;

// Prefer grouping by plain column references, not aliases of projections
// Verify each GROUP BY term matches a base column name of the table
boolean plain = groupByTerms.stream().allMatch(tableColumns::contains);
if (!plain) { disablePushdown(); }

try {
    runQuery(sql);
} catch (IllegalStateException e) {
    if (e.getMessage() != null && e.getMessage().startsWith("Group By column")) {
        runQueryWithoutPushdown(sql); // report bug + fallback
    } else {
        throw e;
    }
}

When it happens

Trigger: A GROUP BY column variable is absent from the context's selections map, typically after a chain of project/aggregation nodes where the group-by column was renamed or derived but not added to selections before indices() is called.

Common situations: Queries grouping by an expression alias computed in a projection; connector/planner version mismatches where variable naming changed; internal bug reports rather than user-level query mistakes.

Related errors

• Output column (%s) definition not found in input selections:
• PINOT_UNSUPPORTED_EXPRESSION
• PINOT_UNSUPPORTED_EXPRESSION
• PINOT_UNSUPPORTED_EXPRESSION
• PINOT_UNSUPPORTED_EXPRESSION

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/de3006f714dabe25. Report an issue: GitHub.