Datatypes Conversion

When FastBCP exports data from a source database, each native column type is mapped to an appropriate type in the target output format (Parquet, CSV, JSON, etc.).

This section documents the exact type mappings produced by FastBCP for every supported source database and output format combination. Each page includes the Parquet schema generated from a comprehensive test table covering all native datatypes — both in NOT NULL and NULLABLE variants.

Supported Sources

Source	Connection Type	Page
SQL Server	mssql	Parquet
MySQL	mysql	Parquet
Oracle	oracle	Parquet
PostgreSQL (Standard)	pgsql	Parquet
PostgreSQL (COPY)	pgcopy	Parquet
SAP HANA	saphana	Parquet
Teradata	teradata	Parquet

How to Read the Tables

Each conversion page contains two Parquet schema tables:

• testdatatypes_notnull — all columns defined as NOT NULL
• testdatatypes_nullable — all columns allow NULL values

The columns in each table are:

Column	Description
name	The column name in the source table
physical_type	The Parquet physical storage type (e.g. INT32, INT64, BYTE_ARRAY, BOOLEAN, DOUBLE, FLOAT, FIXED_LEN_BYTE_ARRAY)
converted_type	The Parquet converted type annotation (legacy, e.g. UTF8, DECIMAL, DATE, TIMESTAMP_MILLIS)
logical_type	The Parquet logical type (modern annotation, e.g. StringType(), DecimalType(scale=2, precision=18), IntType(...))

tip

The Parquet schemas for NOT NULL and NULLABLE variants are identical in terms of type mapping. The difference is only in the column's repetition level (REQUIRED vs OPTIONAL), which is not shown in these tables.