Python code reference

D.Hub Python code transforms and analyzes data. The primary data-processing library is Polars.

Where it is used

• Create and manage Python code under Code in a collection.
• Add a Python code node to a pipeline, then select an existing code asset or write new code.
• A code node passes its result to the next node in the pipeline.

Execution contract

A Python code node uses the run function as its entry point. Its input is a Polars DataFrame, and it returns a dictionary containing an "output" key.

import polars as pl

def run(input, *, options=None, contexts=None):
    # input   : Polars DataFrame (input data)
    # options : dict (custom options)
    # contexts: dict (values shared between steps)
    output = input.with_columns(
        (pl.col("value") * 2).alias("doubled_value")
    )
    return {"output": output}

Input input

• A single input is passed as the first parameter, input (a Polars DataFrame).
• A node with multiple inputs receives each input connection's alias as a parameter name. See Accept multiple inputs for an example.

Options options

• Values entered in the code node's Step options are passed in the options dictionary.
• Use options for dates, filter conditions, and other values that change between runs.

Shared values contexts

• Values shared between steps are passed in the contexts dictionary.
• contexts is always passed at runtime. Declare it as a keyword-only parameter even if the code does not use it.

Output output

• Return a dictionary containing an "output" key whose value is a Polars DataFrame.

Return format

Returning the DataFrame directly, as in return output, causes an error. Return it in the form {"output": output}.

Available packages

The pipeline Python environment provides the following commonly used data-processing and geospatial-analysis libraries.

Package	Purpose
polars	DataFrame processing (primary library)
pandas	General-purpose DataFrame processing
pyarrow	Apache Arrow columnar-data processing
numpy	Numerical computing
geopandas	Geospatial data processing
h3	H3 hexagonal spatial indexing
pyproj	Coordinate-system transformations
rdflib	RDF and ontology data processing

Additional packages

Enter packages that are not included in the default environment in the code's Packages field. Ask your administrator which packages are permitted.

Code examples

Filter data

import polars as pl

def run(input, *, options=None, contexts=None):
    return {"output": input.filter(
        (pl.col("status") == "active") & (pl.col("age") >= 18)
    )}

Accept multiple inputs

A node with multiple input connections receives each input's alias as a parameter name. Declare options and contexts as keyword-only parameters after the input parameters.

import polars as pl

def run(orders, customers, *, options=None, contexts=None):
    output = orders.join(customers, on="customer_id", how="left")
    return {"output": output}

Use options

import polars as pl

def run(input, *, options=None, contexts=None):
    threshold = float(options.get("threshold", 100))
    return {"output": input.filter(pl.col("value") > threshold)}

See the Polars user guide for general DataFrame transformations and functions. The D.Hub run input and output contract remains the same when you use Pandas or GIS packages.

Debugging tips

• View print() output in the pipeline run logs.
• Inspect intermediate results with print(input.head(5)) and the schema with print(input.schema).
• When an error occurs, its traceback appears in the run logs.

Related documentation

• See SQL reference for SQL code nodes and the input table.
• See Add nodes to connect a code node's inputs and outputs.