apache/cassandra · error · MarshalException

Expected a float value, but got a %s: %s

Error message

Expected a float value, but got a %s: %s

What it means

FloatType.fromJSONObject() deserializes a parsed JSON value into a float; after the string and Number paths it catches ClassCastException and throws this MarshalException. It means the JSON value passed to fromJSONObject was neither a String nor a Number (e.g. a Boolean, Map, or List), so the float conversion cast failed. Cassandra throws it because a float column can only be materialized from a textual or numeric JSON representation.

Source

Thrown at src/java/org/apache/cassandra/db/marshal/FloatType.java:109

      catch (NumberFormatException e1)
      {
          throw new MarshalException(String.format("Unable to make float from '%s'", source), e1);
      }
    }

    @Override
    public Term fromJSONObject(Object parsed) throws MarshalException
    {
        try
        {
            if (parsed instanceof String)
                return new Constants.Value(fromString((String) parsed));
            else
                return new Constants.Value(getSerializer().serialize(((Number) parsed).floatValue()));
        }
        catch (ClassCastException exc)
        {
            throw new MarshalException(String.format(
                    "Expected a float value, but got a %s: %s", parsed.getClass().getSimpleName(), parsed));
        }
    }

    @Override
    public String toJSONString(ByteBuffer buffer, ProtocolVersion protocolVersion)
    {
        Float value = getSerializer().deserialize(buffer);
        if (value == null)
            return "\"\"";
        // JSON does not support NaN, Infinity and -Infinity values. Most of the parser convert them into null.
        if (value.isNaN() || value.isInfinite())
            return "null";
        return value.toString();
    }

    public CQL3Type asCQL3Type()
    {

View on GitHub (pinned to 88fd0f6a0e)

Solutions

  1. Ensure the JSON value for the float field is a number or numeric string before calling fromJSONObject
  2. Fix the client/producer so it serializes the field as a JSON number (e.g. 3.14) instead of a boolean/array/object
  3. If the field can legitimately be non-numeric, change the column type or drop/transform the value upstream
  4. Wrap fromJSONObject in try-catch for MarshalException to reject the record with a clear validation message

Example fix

// before
Object parsed = json.get("score"); // Boolean.TRUE
ByteBuffer v = FloatType.instance.fromJSONObject(parsed, ProtocolVersion.V5);
// after
Object parsed = json.get("score");
if (!(parsed instanceof Number || parsed instanceof String))
    throw new IllegalArgumentException("score must be a number or numeric string");
ByteBuffer v = FloatType.instance.fromJSONObject(parsed, ProtocolVersion.V5);
Defensive patterns

Strategy: validation

Validate before calling

if (!(parsed instanceof Number || parsed instanceof String))
    throw new IllegalArgumentException("Expected a float (number or string), got: " + parsed.getClass().getSimpleName());

Type guard

static boolean isFloatCompatible(Object v) { return v instanceof Number || v instanceof String; }

Try / catch

try { FloatType.instance.fromJSONObject(parsed, pv); } catch (MarshalException e) { /* reject record: non-numeric JSON value for float field */ }

Prevention

When it happens

Trigger: Calling FloatType.instance.fromJSONObject(parsed, protocolVersion) where parsed is a JSON object that parsed to a non-String/non-Number class such as Boolean, Map, or List (the cast ((Number) parsed).floatValue() throws ClassCastException, rethrown as MarshalException).

Common situations: Inserting JSON like {"f": true} or {"f": [1,2]} into a float column via the JSON path; buggy client serializers emitting booleans/nulls for float fields; schema drift where a column changed type but old JSON payloads still flow.

Understand the failure class

Background: Type mismatch errors: IllegalArgumentException, TypeError and type guards across 150 open-source libraries — this error's family across 150 libraries.

Related errors

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/1155dffb79e2e554. Report an issue: GitHub.