Bodo Python Quick Start

This quickstart guide will walk you through the process of running a simple Python computation using Bodo on your local machine.

Installation

Install Bodo to get started (e.g., pip install -U bodo or conda install bodo -c bodo.ai -c conda-forge).

Drop-in Pandas Replacement with Bodo DataFrames

Bodo DataFrame library can be used as a drop-in replacement for Pandas by changing import pandas as pd with import bodo.pandas as pd. For example:

import bodo.pandas as pd
import numpy as np

n = 20_000_000
df = pd.DataFrame({"A": np.arange(n) % 30, "B": np.arange(n)})
df2 = df.groupby("A", as_index=False)["B"].max()
df2.to_parquet("my_data.pq")

Bodo DataFrame library will optimize and parallelize the code automatically when possible. It will fall back to Pandas seamlessly when some API isn't supported yet and throw a warning. See the DataFrame Library API reference for supported Pandas APIs.

Bodo JIT Compilation for Custom Code

JIT compilation converts Python functions to optimized parallel binaries, which can provide orders of magnitude performance boost for custom code like user defined functions (UDFs). For example:

import bodo
import bodo.pandas as pd
import numpy as np

@bodo.jit
def f(df):
    return df.apply(lambda r: 0 if r.A == 0 else (r.B // r.A), axis=1)

n = 20_000_000
df = pd.DataFrame({"A": np.arange(n) % 30, "B": np.arange(n)})
S = f(df)
pd.DataFrame({"C": S}).to_parquet("my_data.pq")

All the code in JIT functions has to be compilable by Bodo JIT (will throw appropriate errors otherwise). See JIT development guide and JIT API reference for supported Python features and APIs.