prestodb/presto · error · ParquetDecodingException

could not decode the dictionary for

Error message

could not decode the dictionary for 

What it means

Thrown in AbstractColumnReader.init when the dictionary page for a column chunk exists but its encoding fails to initialize a Dictionary object. The original IOException is wrapped in a ParquetDecodingException naming the column descriptor. This means the file declares a dictionary for the column but the dictionary page bytes/encoding cannot be decoded by this reader.

Source

Thrown at presto-parquet/src/main/java/com/facebook/presto/parquet/reader/AbstractColumnReader.java:115

    @Override
    public boolean isInitialized()
    {
        return pageReader != null && field != null;
    }

    @Override
    public void init(PageReader pageReader, Field field, RowRanges rowRanges, Optional<DateTimeZone> timezone)
    {
        this.pageReader = requireNonNull(pageReader, "pageReader is null");
        this.field = requireNonNull(field, "field is null");
        DictionaryPage dictionaryPage = pageReader.readDictionaryPage();

        if (dictionaryPage != null) {
            try {
                dictionary = dictionaryPage.getEncoding().initDictionary(columnDescriptor, dictionaryPage);
            }
            catch (IOException e) {
                throw new ParquetDecodingException("could not decode the dictionary for " + columnDescriptor, e);
            }
        }
        else {
            dictionary = null;
        }
        checkArgument(pageReader.getValueCountInColumnChunk() > 0, "page is empty");
        valueCountInColumnChunk = pageReader.getValueCountInColumnChunk();
        indexIterator = (rowRanges == null) ? null : rowRanges.iterator();
    }

    @Override
    public void prepareNextRead(int batchSize)
    {
        readOffset = readOffset + nextBatchSize;
        nextBatchSize = batchSize;
    }

    @Override

View on GitHub (pinned to 55bb57d202)

Solutions

1. Verify the file integrity (re-copy / checksum) and re-read; a truncated or corrupted dictionary page is the most common cause.
2. Rewrite the file with dictionary encoding disabled (e.g. parquet.writer.dictionary... or parquet-tools rewrite) so no dictionary page is present.
3. Check the writer version that produced the file; if a known writer bug, re-export the data with a fixed/newer writer.
4. Upgrade Presto in case initDictionary lacks support for the encoding used by the writer.
5. As a fallback, disable predicate pushdown/scan that column via a full rewrite to PLAIN encoding.

Example fix

// before (reading corrupt file directly)
SELECT col FROM corrupted_table;  -- ParquetDecodingException: could not decode the dictionary for ...

// after (rewrite without dictionary encoding, then read)
// $ parquet-tools rewrite --disable-dictionary bad.parquet fixed.parquet
SELECT col FROM fixed_table;

// Validate the file before scanning: dump dictionary pages
// $ parquet-tools dump file.parquet | head   # fails/errs on corrupt dictionary
boolean readable = new ParquetMetadataReader().readFooter(file) != null;

try {
    // scan / initColumnReader
}
catch (ParquetDecodingException e) {
    if (e.getMessage().startsWith("could not decode the dictionary for")) {
        logger.warn("Corrupt dictionary, falling back to rewrite/plain read", e);
        // fall back to reading a rewritten (non-dictionary) copy
    } else {
        throw e;
    }
}

When it happens

Trigger: dictionaryPage != null and dictionaryPage.getEncoding().initDictionary(columnDescriptor, dictionaryPage) throws IOException — corrupt or truncated dictionary page, dictionary encoded with PLAIN_DICTIONARY/RLE_DICTIONARY bytes this reader cannot parse, or a dictionary page whose contents don't match the column schema.

Common situations: Files corrupted in transit or truncated by a failed copy; files written by a writer producing dictionary pages with encodings or bit widths Presto's reader mishandles; HDFS/S3 reads returning partial data; writer bugs (e.g. old versions of certain writers writing malformed dictionaries).

Related errors

• not a valid mode
• could not decode the dictionary for
• Dictionary is missing for Page
• Unsupported Parquet encoding:
• Failed to decode.

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/32416d441e293ad1. Report an issue: GitHub.