What causes "Expected a numeric value for field %s of type REAL: %s [%s]" in prestodb/presto?

Reindex documents so the field holds unquoted JSON numbers: "price": 19.99.

prestodb/presto · error · PrestoException

ELASTICSEARCH_TYPE_MISMATCH

ELASTICSEARCH_TYPE_MISMATCH

Error message

Expected a numeric value for field %s of type REAL: %s [%s]

What it means

RealDecoder requires the mapped REAL (float) column's document value to be an instance of java.lang.Number; it writes floatBits via REAL.writeLong. Any non-numeric JSON value (string, boolean, object, array) raises ELASTICSEARCH_TYPE_MISMATCH. No string-to-float coercion is performed.

Source

Thrown at presto-elasticsearch/src/main/java/com/facebook/presto/elasticsearch/decoders/RealDecoder.java:48

    private final String path;

    public RealDecoder(String path)
    {
        this.path = requireNonNull(path, "path is null");
    }

    @Override
    public void decode(Hit hit, Supplier<Object> getter, BlockBuilder output)
    {
        Object value = getter.get();
        if (value == null) {
            output.appendNull();
        }
        else if (value instanceof Number) {
            REAL.writeLong(output, Float.floatToRawIntBits(((Number) value).floatValue()));
        }
        else {
            throw new PrestoException(ELASTICSEARCH_TYPE_MISMATCH, format("Expected a numeric value for field %s of type REAL: %s [%s]", path, value, value.getClass().getSimpleName()));
        }
    }
}

View on GitHub (pinned to 55bb57d202)

Solutions

Reindex documents so the field holds unquoted JSON numbers: "price": 19.99.
Map the Presto column as VARCHAR and CAST in queries, or use try_cast to tolerate bad rows.
Correct the Elasticsearch dynamic/strict mapping so numeric inputs are not coerced into text fields.

Example fix

// before
{"price": "19.99"}
// after
{"price": 19.99}

Defensive patterns

Strategy: validation

Validate before calling

SELECT t.price FROM idx t WHERE t.price IS NOT NULL AND TYPEOF(t.price) NOT IN ('real','double','integer') LIMIT 1;

Type guard

function isNumericValue(v) { return typeof v === 'number' && Number.isFinite(v); }

Try / catch

try { SELECT price FROM "idx" } catch (PrestoException e) { if (e.getErrorCode().getName().equals("ELASTICSEARCH_TYPE_MISMATCH")) { /* switch column to VARCHAR or fix data */ } throw e; }

Prevention

Never quote numbers when generating JSON in pipelines.
Set an explicit float mapping in Elasticsearch before writing.
Validate document schema with a JSON Schema check on ingest.

When it happens

Trigger: decode() sees e.g. {"price": "19.99"} or {"price": true} for a field the Presto table declares as REAL.

Common situations: JSON numbers quoted during export/import; Elasticsearch mapping is float but some documents store the value as text (multi-type index); a CSV-to-JSON pipeline left everything as strings.

Related errors

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/6db179e296f235e1. Report an issue: GitHub.