< Rocky> how do i fix this error ```parquet column reader no StarRocks #questions-and-troubleshooting

<@U0A71G31CDV> how do i fix this error? ```parquet...

Jaya Kasa

03/12/2026, 1:51 AM

@Rocky how do i fix this error?

Copy code

parquet column reader: not supported convert from parquet `FIXED_LEN_BYTE_ARRAY` to `VARBINARY`

Rocky

03/12/2026, 1:52 AM

The error message

parquet column reader: not supported convert from parquet FIXED_LEN_BYTE_ARRAY to VARBINARY

occurs because the StarRocks Parquet reader currently supports direct mapping to

VARBINARY

only from Parquet's

BYTE_ARRAY

(variable-length) type, and not from its

FIXED_LEN_BYTE_ARRAY

(fixed-length) type. In Parquet,

FIXED_LEN_BYTE_ARRAY

is commonly used for two things: High-precision Decimals or Fixed-size Binary data (like UUIDs or Hashes). Depending on what your data actually is, here are the ways to fix this: 1. If the column is actually a Decimal If the Parquet file was generated from a system that stores Decimals (especially Decimal128) as fixed-length byte arrays, you should change the column type in your StarRocks table to

DECIMAL

. * Fix: Alter your table or recreation script to use

DECIMAL(precision, scale)

instead of

VARBINARY

. 2. If the column is fixed-length Binary data (e.g., UUIDs) If the data is truly binary but stored in a fixed-length format in Parquet, StarRocks' reader currently doesn't support the automatic conversion to the

VARBINARY

type. * Workaround (Use VARCHAR): Change the StarRocks column type to

VARCHAR

. StarRocks can often read fixed-length binary arrays as strings. Once loaded, you can use functions like

hex()

base64()

to process the data if needed. * Workaround (Change Source): If you have control over how the Parquet file is created, try saving that column as a standard

BYTE_ARRAY

(variable length binary) instead of

FIXED_LEN_BYTE_ARRAY

. 3. If using
FILES()
or External Tables If you are using the

FILES()

table function or an External Catalog and cannot change the schema easily, you can try to cast the column explicitly in your SQL query, though the reader error may still persist if it hits the column during the initial scan. * Example:

SELECT CAST(col AS VARCHAR) FROM FILES(...)

Summary of Support: * Supported:

parquet::BYTE_ARRAY

StarRocks::VARBINARY

* Not Supported:

parquet::FIXED_LEN_BYTE_ARRAY

StarRocks::VARBINARY

(This is a known limitation in the current C++ reader). References • sql-reference/data-types/semi_structured/VARIANT.md • sql-reference/data-types/string-type/BINARY.md • sql-reference/sql-functions/table-functions/files.md • administration/management/FE_configuration.md • unloading/Flink_connector.md

Jaya Kasa

03/12/2026, 1:53 AM

@Rocky how do i read a uuid column from iceberg?

Rocky

03/12/2026, 1:54 AM

Hello! I'm Rocky. Dealing with

UUID

columns from Iceberg in StarRocks depends slightly on your version, as the mapping behavior was updated recently to improve consistency and performance. Here is how you can read a UUID column: 1. Data Type Mapping * StarRocks v3.3.11 and later: The Iceberg

UUID

type is automatically mapped to the VARBINARY
type in StarRocks. * Earlier Versions: It was typically mapped to STRING
(VARCHAR) or might have required manual mapping if using the older "External Table" syntax. 2. How to Read/Query If you are on a recent version where it is mapped to

VARBINARY

, a simple

SELECT uuid_col

might display the raw binary or a hex representation depending on your SQL client settings. To ensure it is human-readable, you should use the

hex()

function:

Copy code

sql
SELECT hex(uuid_column) FROM iceberg_catalog.db.table_name;

3. Formatting as a Standard UUID String StarRocks stores the UUID as a 16-byte binary. If you need it in the standard hyphenated format (

aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee

), you can use string manipulation:

Copy code

sql
SELECT 
    lower(concat_ws('-', 
        left(hex(uuid_col), 8), 
        substr(hex(uuid_col), 9, 4), 
        substr(hex(uuid_col), 13, 4), 
        substr(hex(uuid_col), 17, 4), 
        right(hex(uuid_col), 12)
    )) AS formatted_uuid
FROM iceberg_catalog.db.table_name;

Summary Tips: * Check the Schema: Use

DESC <table_name>

to see if the column is currently interpreted as

VARBINARY

VARCHAR

. * Filtering: You can filter using hex strings:

WHERE uuid_col = unhex('550e8400e29b41d4a716446655440000')

. * Catalog: Ensure you are using the Iceberg Catalog (recommended) rather than the deprecated External Table syntax for the best type-mapping support. Let me know if you need help with a specific version or query! References • data_source/catalog/maxcompute_catalog.md • data_source/catalog/paimon_catalog.md • release_notes/release-3.3.md • data_source/External_table.md • data_source/catalog/iceberg/iceberg_catalog.md

Jaya Kasa

03/12/2026, 1:55 AM

@Rocky how do i read a uuid column that is stored as

FIXED_LEN_BYTE_ARRAY

from iceberg?

Rocky

03/12/2026, 1:55 AM

To read a UUID column stored as

FIXED_LEN_BYTE_ARRAY

in Iceberg, the behavior depends on the version of StarRocks you are using and whether you are using an Iceberg Catalog or a manual external table. 1. Mapping via Iceberg Catalog When using an Iceberg Catalog (the recommended way to query Iceberg), StarRocks handles the type mapping automatically: * StarRocks v3.3.11 and later: The Iceberg

UUID

type (which uses

FIXED_LEN_BYTE_ARRAY(16)

physically in Parquet) is mapped to the BINARY
data type. * Versions before v3.3.11: The Iceberg

UUID

type was typically mapped to VARCHAR
. 2. How to Read/Format the Column If your column is mapped to

BINARY

(v3.3.11+), querying it directly might return raw bytes or a hex-encoded string depending on your client settings. To view it in a human-readable format, you can use the

HEX()

function. To get a 32-character Hex String:

Copy code

sql
SELECT HEX(uuid_column) FROM your_iceberg_table;
-- Result: 550E8400E29B41D4A716446655440000

To get the standard hyphenated UUID format (
8-4-4-4-12
): Since StarRocks does not currently have a built-in

BIN_TO_UUID

function, you can format the binary column manually using

SUBSTR

and `CONCAT`:

Copy code

sql
SELECT 
    LOWER(CONCAT_WS('-', 
        SUBSTR(HEX(uuid_column), 1, 8), 
        SUBSTR(HEX(uuid_column), 9, 4), 
        SUBSTR(HEX(uuid_column), 13, 4), 
        SUBSTR(HEX(uuid_column), 17, 4), 
        SUBSTR(HEX(uuid_column), 21, 12)
    )) AS uuid_string
FROM your_iceberg_table;
-- Result: 550e8400-e29b-41d4-a716-446655440000

3. Manual External Table (Deprecated method) If you are defining a manual Iceberg External Table (not using a Catalog), you should specify the column type as

VARCHAR

BINARY

in your

CREATE EXTERNAL TABLE

statement. According to StarRocks documentation, the Iceberg

UUID

type should be mapped to

VARCHAR

in these definitions for legacy support, but

BINARY

is preferred in modern versions for better performance and alignment with Parquet physical types. Summary * Check your version: Upgrading to 3.3.11+ ensures standard mapping to

BINARY

. * Use `HEX()`: This is the easiest way to inspect the values. * Catalog usage: Always prefer using an Iceberg Catalog so StarRocks can automatically resolve the

FIXED_LEN_BYTE_ARRAY

as a UUID. References • sql-reference/sql-functions/binary-functions/to_binary.md • sql-reference/sql-functions/bitmap-functions/base64_to_bitmap.md • sql-reference/sql-functions/string-functions/hex_decode_string.md • sql-reference/sql-functions/bitmap-functions/bitmap_to_binary.md • sql-reference/sql-functions/string-functions/hex_decode_binary.md

Jaya Kasa

03/12/2026, 1:57 AM

@Rocky given this query to read a table from iceberg with a uuid column stored as

FIXED_LEN_BYTE_ARRAY

, fix this query to work from starrocks:

select * from datalake.events_table;

Rocky

03/12/2026, 1:58 AM

To read an Iceberg table containing a

UUID

column (stored as

FIXED_LEN_BYTE_ARRAY

) in StarRocks, you should be aware that from version 3.3.11 onwards, StarRocks officially maps the Iceberg

UUID

type to the BINARY
data type. If your query

SELECT * FROM datalake.events_table;

is currently failing or returning unreadable binary data, here is how to fix it: 1. Version Requirement Ensure you are using StarRocks v3.3.11 or later. In earlier versions,

FIXED_LEN_BYTE_ARRAY

columns representing UUIDs were often unsupported in the Iceberg Catalog and would cause the query to fail. 2. Read as a Human-Readable String Since the column is mapped to

BINARY

, it will appear as a hexadecimal blob (e.g.,

0x123E...

) in most clients. To view it as a standard UUID string (

8-4-4-4-12

format), use the following SQL:

Copy code

sql
SELECT 
    lower(concat_ws('-', 
        substr(hex(uuid_column), 1, 8), 
        substr(hex(uuid_column), 9, 4), 
        substr(hex(uuid_column), 13, 4), 
        substr(hex(uuid_column), 17, 4), 
        substr(hex(uuid_column), 21, 12)
    )) AS uuid_string,
    * 
FROM datalake.events_table;

3. How to Filter by UUID If you need to filter the table using a UUID string, you must convert the string back to binary using

hex_decode_binary

and remove the dashes:

Copy code

sql
SELECT * 
FROM datalake.events_table 
WHERE uuid_column = hex_decode_binary(replace('550e8400-e29b-41d4-a716-446655440000', '-', ''));

Summary of Tips: * Check Mapping: Run

DESC datalake.events_table;

to verify that the column type is indeed

VARBINARY

BINARY

. * Session Variable: If you are on a version older than 3.3.11 and the query fails due to "Unsupported type," you can try setting

SET force_external_table_query_to_binary = true;

to force the engine to treat the complex/unsupported types as binary blobs, though upgrading is highly recommended for stable Iceberg support. References • sql-reference/sql-functions/binary-functions/to_binary.md • sql-reference/sql-functions/string-functions/hex_decode_string.md • sql-reference/sql-functions/string-functions/hex_decode_binary.md • sql-reference/data-types/string-type/BINARY.md • sql-reference/sql-functions/bitmap-functions/bitmap_from_binary.md • sql-reference/sql-functions/bitmap-functions/bitmap_to_binary.md

2 Views

Open in Slack

Previous Next