Facets Overview

1.1.1 · abandoned · verified Fri Apr 17

Facets Overview is a Python library that generates interactive visualizations for analyzing machine learning datasets. It processes data into protobufs, which are then rendered by front-end web components (Polymer). The library is currently at version 1.1.1, with the last release in 2017, indicating it is no longer actively maintained.

Common errors

```
ModuleNotFoundError: No module named 'facets_overview'
```
cause The `facets-overview` package is not installed in your current Python environment.

fix Run `pip install facets-overview` to install the library.
```
google.protobuf.message.DecodeError: Error parsing message
```
cause This error often indicates a version mismatch between the `protobuf` library installed in your environment and the `protobuf` version used to compile the `.proto` definitions within `facets-overview`.

fix Try installing an older version of the `protobuf` library: `pip install protobuf<3.20` (or `protobuf==3.19.0` for a specific version known to work with older Google libraries).
```
Javascript errors in browser console related to '<link rel="import">' or Polymer
```
cause Facets Overview relies on deprecated Web Components v0 APIs (Polymer 1/2 and HTML Imports). Modern browsers have removed support for these features.

fix This is a fundamental limitation of an abandoned library. There is no simple fix for modern browsers. You might need to use an older browser version or a specific environment where these legacy web components are still supported.
```
AttributeError: 'HTML' object has no attribute '_repr_html_' (or similar non-rendering behavior in JupyterLab)
```
cause The visualization relies on Jupyter Notebook nbextensions for rendering and does not have native support or an easy integration path for JupyterLab or other environments.

fix Ensure you are running in a classic Jupyter Notebook and that the `facets_overview` nbextension is enabled (`jupyter nbextension enable --py facets_overview`). For JupyterLab, consider using alternative libraries.

Warnings

breaking The v1.0.0 release updated the underlying web component technology from Polymer 1 to Polymer 2. This can break existing display mechanisms or require browser environment updates.
Fix: Ensure your Jupyter environment and browser are compatible with Polymer 2. Running old notebooks may require adjustments or specific browser versions.
deprecated Facets Overview is an abandoned library, with the last release in 2017. It relies on deprecated front-end technologies (Polymer 1/2) and older Python practices, making it largely incompatible with modern web browsers, JupyterLab, and newer Python versions (3.8+).
Fix: Consider migrating to actively maintained data exploration and visualization libraries such as `pandas-profiling`, `sweetviz`, `explori`, or `TensorFlow Data Validation` (the spiritual successor by Google).
gotcha The display mechanism is tightly coupled to Jupyter Notebooks via nbextensions. It does not natively support JupyterLab or other Python environments without significant workarounds.
Fix: Ensure you are running in a classic Jupyter Notebook environment with the `facets_overview` nbextension enabled. For JupyterLab, manual configuration or alternative rendering methods would be required, which are not officially supported.
gotcha Facets Overview uses Google Protobuf for data serialization. There are known compatibility issues with newer `protobuf` library versions, leading to `DecodeError` or `TypeError` exceptions.
Fix: Pin the `protobuf` library to an older, compatible version (e.g., `pip install protobuf<3.20`). Experimentation may be required to find a working version for your specific setup.

Install

pip install facets-overview Install with pip

Imports

GenericFeatureStatisticsGenerator

from facets_overview.generic_feature_statistics_generator import GenericFeatureStatisticsGenerator

Main class for generating feature statistics protobufs.

FeatureStatisticsList

from facets_overview.feature_statistics_pb2 import FeatureStatisticsList

Protobuf definition for feature statistics.

Quickstart

This quickstart demonstrates how to generate and display a Facets Overview visualization within a Jupyter Notebook. It uses `pandas` for data handling and `IPython.display` to render the interactive HTML. Note that this requires the `facets_overview` Jupyter nbextension to be enabled and an environment compatible with Polymer 2.

import pandas as pd
from facets_overview.generic_feature_statistics_generator import GenericFeatureStatisticsGenerator
from IPython.display import display, HTML

# Example data
data = [
    {'feature1': 1, 'feature2': 'cat'},
    {'feature1': 2, 'feature2': 'dog'},
    {'feature1': 3, 'feature2': 'cat'}
]
df = pd.DataFrame(data)

# Create the FeatureStatisticsList proto for Facets Overview.
gfsg = GenericFeatureStatisticsGenerator()
overview_proto = gfsg.CreateFeatureStatisticsListFromDataFrame(df)
protostr = overview_proto.SerializeToString().decode('utf-8')

# Display the Facets Overview UI (requires Jupyter Notebook and nbextension enabled)
HTML_TEMPLATE = """<link rel="import" href="/nbextensions/facets_overview/facets-overview.html">
        <facets-overview id="facets-overview"></facets-overview>
        <script>
          document.querySelector("#facets-overview").protoInput = "{protostr}";
        </script>"""
html_output = HTML_TEMPLATE.format(protostr=protostr)
display(HTML(html_output))

view raw JSON →