Top Python package picks

xlwings

xlwings is one of the most valuable packages anyone working with spreadsheets can install. With xlwings, spreadsheets (excel and google) can be created, loaded, altered, and automated. The package has additional features allowing Python to be called from excel, workbooks to be exposed through API, and array formulas to be built.

DuckDB

DuckDB is an in-process SQL OLAP database management system with a fully featured Python API. One of the best features of DuckDB is allowing users to write SQL queries directly on a variety of dataframes, including both Polars and Pandas. Other features include creating persistent tables from dataframes and registering dataframes as virtual tables.

python-fsutil

python-fsutil is a great package for anyone that works with file systems. This package contains a ton of functions that make working with directories, paths, and files a breeze. Many of the mundane file system operations can be easily automated. Things like cleaning directories, creating zip files, getting file creation dates, and more are quickly achievable with this package.

Pins

Pins is a terrific way to store, share, version, and publish Python data products (datasets, models, etc.). Pins is compatible with tools most of us already use, like network drives, Amazon S3, Posit Connect, DropBox, Google Cloud Storage, and more. Pins makes tracking changes and re-running analysis on historical data efficient.

from pins import board_temp
from pins.data import mtcars

# Create a temporary board and write pin containing CSV file to it
board = board_temp()

# Save dataset with detailed information
board.pin_write(
    mtcars,
    name="mtcars2",
    type="csv",
    description = "Data extracted from the 1974 Motor Trend US magazine, \
        and comprises fuel consumption and 10 aspects of automobile design \
        and performance for 32 automobiles (1973–74 models).",
    metadata = {
         "source": "Henderson and Velleman (1981), \
         Building multiple regression models interactively. \
         Biometrics, 37, 391–411."
    }
)

# Read pin from board
board.pin_read("mtcars")

Loguru

Making logs in Python is made more pleasant by Loguru, which makes configuring logs incredibly simple without sacrificing customizability.

Pendulum

Pendelum improves on the Python datetime library with added functionality like time zone management, enhanced parsing, easier calculations, and additional attributes. This package is a must for anyone working with datetimes across different time zones.

DataPrep

DataPrep is my go-to for doing exploratory analysis on a dataset. In less than three lines of code, you can generate a full report on a dataset, covering everything from missing data to correlations. DataPrep also includes functions to clean and standardize datasets. This package saves a ton of time on some of the traditionally least interesting work.

from dataprep.datasets import load_dataset
from dataprep.eda import create_report
df = load_dataset("titanic")
create_report(df).show()

ydata-profiling

ydata-profiling has a stated goal of providing one-line exploratory data analysis. To that end, it largely succeeds. A single line of code offers in-depth dataset analysis, helping you understand you’re data. ydata-profiling has great functionality for exploring time series and text data.

Altair

Altair is a simple but powerful plotting library capable of producing elegant and interactive charts. Altair is handy when data transformations and filtering are required to generate a plot, thanks to its built-in support for these actions.

import altair as alt
from vega_datasets import data

source = data.cars()

alt.Chart(source).mark_circle(size=60).encode(
    x='Horsepower',
    y='Miles_per_Gallon',
    color='Origin',
    tooltip=['Name', 'Origin', 'Horsepower', 'Miles_per_Gallon']
).interactive()

Pygal

Pygal is another great option for those needing to easily create beautiful interactive plots. Pygal has a deep roster of chart types compared to some alternatives.

Shiny for Python

Shiny, R’s primary web app framework, is now available for Python. Shiny enables Python users to build interactive and reactive web applications that remain performant. This is an excellent choice for anyone who needs to allow user input and response without refreshing the entire application.

Beautiful Soup 4

Beautiful Soup is a package that retrieves data from HTML and XML files. Beautiful Soup is commonly used to parse data retrieved through the aforementioned Requests package. With Beautiful Soup you can easily ingest data directly from the internet.

import pandas as pd
from bs4 import BeautifulSoup
import requests

response = requests.get(
'https://en.wikipedia.org/wiki/List_of_U.S._states_by_median_home_price'
)
tbl = BeautifulSoup(response.text, 'html.parser').find('table', {'class':'wikitable'})
tbl_df = pd.DataFrame(pd.read_html(str(tbl))[0])

FastAPI

FastAPI allows users to build fully functional and deployable APIs with minimal code. FastAPI helps automate much of the work in creating an API, like documentation and data validation.

vetiver

Vetiver is a package for individuals and teams that want to instill good MLOps practices. Versioning, deploying, and monitoring ML models is painless with vetiver. Several model frameworks are supported, including sci-kit-learn and PyTorch.

from vetiver import VetiverModel
from vetiver.data import mtcars
from sklearn.linear_model import LinearRegression

model = LinearRegression().fit(mtcars.drop(columns="mpg"), mtcars["mpg"])
v = VetiverModel(model, 
                 model_name = "cars_linear", 
                 prototype_data = mtcars.drop(columns="mpg")
                )
v.description

ELI5

ELI5 helps to visualize and debug ML models, helping to explain model results. ELI5 helps identify model issues and explain model behaviors. For those building and debugging ML models, this package can help you save a ton of time.

FlashText

FlashText is a package for anyone needing to perform regular expression operations on large datasets. The primary benefit of FlashText is its speed at scale. The benefits are muted on smaller datasets.

NLTK

Natural Language Toolkit, or NLTK for short, is Python’s standard natural language processing framework. Anyone planning to do something involving text analysis will likely be using NLTK, thanks to its ubiquity, diverse feature set, and documentation. NLTK works excellent with other packages making creating things like sentiment models as easy as possible.

import numpy as np
import nltk
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from scipy.special import softmax

# Use Roberta Model
model = f"cardiffnlp/twitter-roberta-base-sentiment"
tokenizer = AutoTokenizer.from_pretrained(model)
model = AutoModelForSequenceClassification.from_pretrained(model)

sentence = 'I really love Post Package Manager. \
            It is a great tool for managing all of \
            my R and Python packages.'

# Run Model on Sentence to get sentiment
encoded_text = tokenizer(example, return_tensors='pt')
output = model(**encoded_text)
scores = output[0][0].detach().numpy()
scores = softmax(scores)
scores_dict = {
    'roberta_neg' : scores[0],
    'roberta_neu' : scores[1],
    'roberta_pos' : scores[2]
}

fbs

fbs makes creating a desktop application easy. It makes application deployment simple, sometimes taking seconds to generate an installer for a PyQt6 application. Anyone who desires to build and deploy Python-based applications should take a look at pairing PyQt6 with fbs.