Help Center/ DataArts Fabric/ Developer Guide/ Multimodal AI Data Lake/ Multimodal Data Processing/ User-Defined Functions/ Vectorized UDF
Updated on 2025-12-19 GMT+08:00
View PDF
Share

Vectorized UDF

Designed to overcome performance limitations of traditional row-based UDFs, vectorized UDFs execute operations in batches. They typically accept and return data in PyArrow or Pandas formats.

Example

The two examples below demonstrate how to use vectorized UDFs.

  • Example 1: Use PyArrow to accelerate vectorization. 
    import fabric_data as fabric
from fabric_data.udf import RegisterType

import pyarrow.compute as pc

# Implicitly register a UDF.
@fabric.udf.pyarrow(database="your-database", register_type=RegisterType.STAGED)
def calculate_product(prices: fabric.PyarrowVector[float], quantities: fabric.PyarrowVector[int]) -> fabric.PyarrowVector[float]:    
    return fabric.PyarrowVector[float](pc.multiply(prices, quantities))

# Use the UDF.
con = ibis.fabric.connect(...)
t = con.table("your-table", database="your-database")
expression = t.select(calculate_product(t.price, t.quantity).name("product column"))
print(expression.execute())
  • Example 2: Use Pandas to accelerate vectorization. 
    import fabric_data as fabric
from fabric_data.udf import RegisterType

import pandas as pd

# Implicitly register a UDF.
@fabric.udf.pandas(database="your-database", register_type=RegisterType.STAGED)
def calculate_product(prices: fabric.PandasVector[float], quantities: fabric.PandasVector[int]) -> fabric.PandasVector[float]:    
    return fabric.PandasVector[float](prices * quantities, dtype=pd.Float64Dtype())

# Use the UDF.
con = ibis.fabric.connect(...)
t = con.table("your-table", database="your-database")
expression = t.select(calculate_product(t.price, t.quantity).name("product column"))
print(expression.execute())
Parent topic: User-Defined Functions