Hive

An HMS or Glue table is treated as Hive when its properties match none of the markers used by the other table formats — Hive is the fallback format. Lakelet reads the schema from the metastore and lists data files from the table and partition locations.

Supported Data File Formats

Input format	SerDe	Status
TextFile	`LazySimpleSerDe`	Supported
Parquet	`ParquetHiveSerDe`	Supported
ORC	`OrcSerde`	Supported
Avro	`AvroSerDe`	Supported

TextFile

These table and SerDe properties are honored:

Property	Behavior
`field.delim`, `serialization.format`	Field delimiter, default `\001`
`serialization.null.format`	Text that reads as `NULL`, default `\N`
`skip.header.line.count`	`0` or `1`

A value that cannot be parsed as its column type is read as NULL, as in Hive. Empty fields are also read as NULL.

Complex type columns (array, map, struct and uniontype) are currently read as NULL. The other columns of the table are read as usual.

Parquet

INT96 timestamps, as written by Hive, Impala and older Spark versions, are read as microsecond timestamps. Their values are taken as UTC, with no timezone conversion.

ORC

Hive ACID tables are not supported.

Avro

Everything is good.

Data Types

Hive type	Arrow type
`tinyint`	`Int8`
`smallint`	`Int16`
`int`	`Int32`
`bigint`	`Int64`
`float`	`Float32`
`double`	`Float64`
`boolean`	`Boolean`
`string`	`Utf8`
`varchar(n)`, `char(n)`	`Utf8`
`binary`	`Binary`
`date`	`Date32`
`timestamp`	Microsecond timestamp without a timezone
`timestamp with local time zone`	Microsecond timestamp in UTC
`decimal(p,s)`	`Decimal128(p,s)`
`decimal(p)`	`Decimal128(p,0)`
`decimal`	`Decimal128(10,0)`
`void`	`Null`
`array<T>`	`List`
`map<K,V>`	`Map`
`struct<name:T,...>`	`Struct`
`uniontype<T,...>`	Sparse `Union`

Metadata Table

data_files

The data_files metadata table lists visible, non-empty data files:

SELECT * FROM `table_name$data_files`;

Column	Description
`file_path`	Full path of the data file, including the storage scheme
`file_size`	Size of the data file in bytes

partitions

The partitions metadata table returns one row for each metastore partition:

SELECT * FROM `table_name$partitions`;

An unpartitioned table returns no rows.

Column	Description
`partition`	Partition values as a string, such as `dt=2026-01-01/country=CN`
`data_file_count`	Number of data files in the partition, counted the same way as `data_files`
`total_data_file_size`	Combined size in bytes of those data files