src/streamlit_filter.py

# https://blog.streamlit.io/auto-generate-a-dataframe-filtering-ui-in-streamlit-with-filter_dataframe/

from pandas.api.types import (
    is_categorical_dtype,
    is_datetime64_any_dtype,
    is_numeric_dtype,
    is_object_dtype,
)
import pandas as pd
import streamlit as st


def filter_dataframe(df: pd.DataFrame, force_on: bool = False, force_on_columns: list[str] = []) -> pd.DataFrame:
    """
    Adds a UI on top of a dataframe to let viewers filter columns

    Args:
        df (pd.DataFrame): Original dataframe

    Returns:
        pd.DataFrame: Filtered dataframe
    """
    if force_on:
        modify = True
    else:
        modify = st.checkbox("Add more filters")

    if not modify:
        return df

    df = df.copy()

    # Try to convert datetimes into a standard format (datetime, no timezone)
    for col in df.columns:
        if is_object_dtype(df[col]):
            try:
                df[col] = pd.to_datetime(df[col])
            except Exception:
                pass

        if is_datetime64_any_dtype(df[col]):
            df[col] = df[col].dt.tz_localize(None)

    modification_container = st.container()

    with modification_container:
        to_filter_columns = st.multiselect("Filter dataframe on", df.columns) + force_on_columns
        for column in to_filter_columns:
            left, right = st.columns((1, 20))
            # Treat columns with < 17 unique values as categorical
            if is_categorical_dtype(df[column]) or df[column].nunique() < 50:
                user_cat_input = right.multiselect(
                    f"Values for {column}",
                    df[column].unique(),
                    default=list(df[column].unique()),
                )
                df = df[df[column].isin(user_cat_input)]
            elif is_numeric_dtype(df[column]):
                _min = float(df[column].min())
                _max = float(df[column].max())
                step = (_max - _min) / 100
                user_num_input = right.slider(
                    f"Values for {column}",
                    min_value=_min,
                    max_value=_max,
                    value=(_min, _max),
                    step=step,
                )
                df = df[df[column].between(*user_num_input)]
            elif is_datetime64_any_dtype(df[column]):
                user_date_input = right.date_input(
                    f"Values for {column}",
                    value=(
                        df[column].min(),
                        df[column].max(),
                    ),
                )
                if isinstance(user_date_input, tuple):
                    if len(user_date_input) == 2:
                        user_date_input_dt = tuple(map(pd.to_datetime, user_date_input))
                        start_date, end_date = user_date_input_dt
                        df = df.loc[df[column].between(start_date, end_date)]
            else:
                user_text_input = right.text_input(
                    f"Substring or regex in {column}",
                )
                if user_text_input:
                    df = df[df[column].astype(str).str.contains(user_text_input)]

    return df


def get_multiselect_for_df_column(df: pd.DataFrame, column_name: str) -> list:
    options_list = sorted(df[column_name].unique().tolist())
    if len(options_list) > 1:
        selected = (
            st.multiselect(column_name.title(), options_list, placeholder=f"Select a {column_name} to filter")
            or options_list
        )
    else:
        selected = options_list
    return selected