dash_app_deploy_heroku.py

"""

This is an example of the app.py file for a dash app deployed on Heroku. 
It creates http://esg-trending-test.herokuapp.com/

Follow the steps on https://dash.plotly.com/deployment 
to delploy the app with Heroku.

Note: 
    Put your data into the same folder or subfolder where `app.py` lives.

    I used a conda virtual environment, but installed all required packages 
    with `pip install` instead of `conda install`. This ensures that the 
    requirements.txt is readable by the Heroku Dyno. 


Author: Philipp Schmalen
License: MIT

"""

import os
import pandas as pd

# plotly
import plotly.graph_objects as go
from plotly.subplots import make_subplots

# dash
import dash
import dash_core_components as dcc
import dash_html_components as html


#~ --------------- SETTINGS --------------- ~
DATA_DIR = './data/'                # <WHERE YOUR DATA LIVES>
FILENAME = 'related_queries_210218_2021.csv' # <NAME OF THE FILE>.csv
#~ ---------------------------------------- ~



#~ --------------- DATA PREPARATION --------------- ~
def load_data(data_dir, filename):
    df_raw = pd.read_csv(os.path.join(DATA_DIR, FILENAME))
    df = df_raw.copy()

    return df

def prepare_data(df):
    """ Feature engineering for Google trends 'related queries'
    
    Input
        df: dataframe
    Return
        dfr, dft: dataframes with rising search interest (r) and top search interest (t)  

    """
    # feature engineering: totals and normalize
    grouped = df.groupby(['ranking']).value # group values by ranking
    df['value_total'] = grouped.transform('sum') # total sum 
    df['value_normalized'] = (df.value-grouped.transform('min'))/(grouped.transform('max')-grouped.transform('min')) # normalize 
    df['value_normalized_total'] = df.groupby(['ranking']).value_normalized.transform('sum') # total sum of normalized values 
    
    # labelling
    df['labels'] = df['query'].apply(lambda x: x.replace(' ', '<br>'))
    df['ranking_label'] = df.ranking.replace({'top': 'Evergreens', 'rising': 'Trending'})
    
    # rankings: top (t) and rising (r)
    dfr, dft = df.query('ranking == "rising"'),  df.query('ranking == "top"')

    return dfr, dft

#~ --------------- PIPELINE --------------- ~
dfr, dft = (load_data(DATA_DIR, FILENAME)
                .pipe(prepare_data))

#~ --------------- PLOT --------------- ~
fig = make_subplots(2, 1,
                    specs=[[{"type": "domain"}], [ {"type": "domain"}]])

fig.add_trace(go.Treemap(
    labels = dfr['labels'],
    parents = dfr.ranking_label, 
    values = dfr.value_normalized, 
), 1, 1)

fig.add_trace(go.Treemap(
    
    labels = dft['labels'],
    parents = dft.ranking_label, 
    values = dft.value_normalized
), 2, 1)

fig.update_layout(margin=dict(t=10, b=10, r=10, l=10))

fig.update_traces(
    opacity=1, 
    textposition='middle center', 
    textfont={'family':"Arial", 'size': 18}, 
    hoverinfo= "label",
    tiling = {'squarifyratio': 1, 'pad': 0}, 
    marker = {'depthfade': False}
)


external_stylesheets = ['https://codepen.io/chriddyp/pen/bWLwgP.css']


app = dash.Dash(__name__, external_stylesheets=external_stylesheets)
server = app.server

app.layout = html.Div(children=[
    html.H1(children='Trending topics in sustainable finance'),

    html.Div(children='''
        Know what people search
    '''),

    dcc.Graph(
        id='example-graph',
        figure=fig
    )
])

if __name__ == '__main__':
    app.run_server(debug=True)