prestodb/presto · error · PrestoException

PARQUET_UNSUPPORTED_ENCODING

PARQUET_UNSUPPORTED_ENCODING

Error message

Dictionary encoding is not supported: %s

What it means

Dictionaries.createDictionary() switches on the column's primitive type to select a dictionary implementation; BOOLEAN (and any unrecognized type) falls through the switch without returning, reaching the final throw of PrestoException with code PARQUET_UNSUPPORTED_ENCODING and message "Dictionary encoding is not supported: <columnDescriptor>". Booleans cannot usefully be dictionary encoded here (a boolean dictionary would be at most two values), so this library refuses to build a dictionary page for such columns and asks callers to decode them with a non-dictionary path.

Source

Thrown at presto-parquet/src/main/java/com/facebook/presto/parquet/batchreader/dictionary/Dictionaries.java:60

                case INT64:
                case DOUBLE:
                    return new LongDictionary(dictionaryPage);
                case INT96:
                    return new TimestampDictionary(dictionaryPage, timezone);
                case BINARY:
                    return new BinaryBatchDictionary(dictionaryPage);
                case FIXED_LEN_BYTE_ARRAY:
                    return new BinaryBatchDictionary(dictionaryPage, columnDescriptor.getPrimitiveType().getTypeLength());
                case BOOLEAN:
                default:
                    break;
            }
        }
        catch (Exception e) {
            throw new ParquetDecodingException("could not decode the dictionary for " + columnDescriptor, e);
        }

        throw new PrestoException(PARQUET_UNSUPPORTED_ENCODING, String.format("Dictionary encoding is not supported: %s", columnDescriptor));
    }
}

View on GitHub (pinned to 55bb57d202)

Solutions

1. Decode boolean columns with the RLE (non-dictionary) decoder — check column.getEncoding() and only call createDictionary for RLE_DICTIONARY/PLAIN_DICTIONARY supported types.
2. Rewrite the file with dictionary encoding disabled for boolean columns (parquet.enable.dictionary=false).
3. If you own the reader code, add a guard on the primitive type before attempting dictionary creation and fall back to plain/RLE decoding.
4. Upgrade Presto if a newer release added support for this column type's dictionary path.

Example fix

// before
Dictionary dict = Dictionaries.createDictionary(descriptor, dictPage, tz); // PrestoException PARQUET_UNSUPPORTED_ENCODING

// after
if (descriptor.getPrimitiveType().getPrimitiveTypeName() == PrimitiveTypeName.BOOLEAN) {
    values = rleBooleanDecoder.read(descriptor, page); // no dictionary needed
} else {
    Dictionary dict = Dictionaries.createDictionary(descriptor, dictPage, tz);
}

// Only request a dictionary for supported primitive types
PrimitiveType pt = descriptor.getPrimitiveType();
if (pt.getPrimitiveTypeName() == PrimitiveTypeName.BOOLEAN) {
    throw new IllegalStateException("Boolean column: decode with RLE, no dictionary");
}

try {
    Dictionary dict = Dictionaries.createDictionary(descriptor, dictPage, tz);
} catch (PrestoException e) {
    if (e.getErrorCode() == PARQUET_UNSUPPORTED_ENCODING.toErrorCode()) {
        return decodeWithoutDictionary(descriptor, page); // RLE/plain path
    }
    throw e;
}

When it happens

Trigger: createDictionary() is invoked for a column whose primitive type is BOOLEAN (or falls into the default case), i.e., the Parquet metadata advertises a dictionary page for a boolean (or otherwise unsupported) column, and the reader attempts to construct its dictionary.

Common situations: Writers that emit dictionary pages even for boolean columns; a BOOLEAN column mistakenly mapped to a dictionary-decodable type in a custom reader; custom/patched Parquet writers producing dictionary pages for unsupported primitive types; readers not checking the column's encoding before requesting a dictionary.

Related errors

• PARQUET_UNSUPPORTED_ENCODING
• not a valid mode
• not a valid mode
• not a valid mode
• not a valid mode

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/7ffc3161444b9194. Report an issue: GitHub.