Update app.py

app.py

@@ -1,3 +1,336 @@
-import streamlit as st 
+import streamlit as st
+import re
+import altair as alt
+st.set_page_config('Product Recommendation Application', layout='wide')
 
-st.write('A Streamlit Application')
+# Functions/Pages
+def intro():
+    import streamlit as st
+
+    st.markdown("## Welcome to the Product Recommendation Application! 👋")
+    st.sidebar.success("Select a page from above.")
+
+    st.markdown(
+        """
+        **👈 Select a page from the navigation bar to the left**
+
+###  Functionality of Pages
+#### Generate Recommendations Page
+- This page allows you to generate recommendations for a particular keyword/publisherid combo.
+- The recommendations for each keyword/publisherid combo, are stored in a separate SQLite Database Table.
+- Ideally, for each keyword/publisherid combo, the recommendations should only be generated once per day. To mock this, you are only allowed to generate recommendations for a specific keyword/publisherid combo once.
+#### Product Recommendations Page
+- This page displays the Product Recommendations for each keyword/publisherid combo for which the recommendations were generated in the previous page.
+- You can currently select the available keywords from the dropdown provided in the side bar.
+- For the time being, the Publisher ID is restricted to a single known value. This can be easily updated later.
+- Note: Since the click tracker I was using had issues, as a workaround I had to use a Button instead. So now, to register a click against a product, you need to click on the 'Select' button provided below each product. 
+- For each session, once a product is clicked on, it is no longer displayed.
+- You can choose the number of recommendations to be displayed, between 1 & 5.
+- The maximum number of recommendations that are available for display for each keyword/publisherid combo is restricted to 50.
+- The minumum number depends on the products available from bizrate.
+- If/when all available recommendations for a keyword/publisherid combo has been clicked on, you can either choose a different keyword, reload the application or navigate to a different page. The last two options creates a new session.
+#### Analytics Page
+- Displays a Line Chart and a Bar Chart, to show the distribution of clicks across Session ID, Keyword, Publisher ID, SKU or Date.
+- Displays a table of the most clicked SKU's
+    """
+    )
+
+def generate_recommendations():
+    import os
+    import string
+    import random
+    import streamlit as st
+    import warnings
+    warnings.filterwarnings('ignore')
+
+    # custom module
+    #from clickcounter import clickcounter
+    from folder_management import create_folder, remove_files_folder
+    from sqlite_database import create_insert_table, query_table, insert_clickdata_table, check_for_table
+    from recommend import query_bizrate, recommend
+
+    # Functions
+    def random_string(N=7):
+        res = ''.join(random.choices(string.ascii_lowercase + string.digits, k=N))
+        return str(res)
+
+    def query_and_recommend(keyword, publisherid):
+        query_bizrate(keyword, publisherid)
+        recommend(keyword, publisherid)
+
+    # Main Program
+    # st.set_page_config('Generate Recommendations', layout='wide')
+
+    st.title('Generate Recommendations')
+    st.markdown('''<pre style="text-align:center">
+    <strong><span style="color:#6a8759">
+    Generate Recommendations for any Keyword - Publisher ID Combo. <br><br> </span></strong></pre>''',
+                unsafe_allow_html=True)
+
+    col1, col2 = st.columns(2)
+
+    selected_keyword = ''
+    publisher_id = ''
+
+    with col1:
+        selected_keyword = st.text_input('Enter a single word as keyword:', 'aquaman')
+        selected_keyword = selected_keyword.lower()
+        selected_keyword = ''.join(re.split(r"[ \|\\\/,.-]", selected_keyword))
+        #selected_keyword = ''.join(selected_keyword.split())
+        st.write(selected_keyword)
+
+    with col2:
+        publisher_id = st.text_input("Enter a Publisher ID: ", '725895')
+        # st.write(publisher_id)
+
+    keyword_pubid_list = []
+    for file in os.listdir('bizrate'):
+        keyword_pubid_list.append(file.replace(".xml", ""))
+
+    if selected_keyword + "_" + publisher_id in keyword_pubid_list:
+        st.error('Keyword - Publisher ID Combo Exists!')
+    else:
+        st.info("Keyword - Publisher ID Combo doesn't exist. Must be queried")
+        st.button('Generate Recommendations', key=random_string(), on_click=query_and_recommend,
+                  args=([selected_keyword, publisher_id]))
+
+def display_recommendations():
+    import pandas as pd
+    import numpy as np
+    import requests
+    from bs4 import BeautifulSoup
+    import urllib.parse
+    import os
+    import string
+    import random
+
+    import time
+
+    import streamlit as st
+    import uuid
+
+    import warnings
+    import sys
+
+    warnings.filterwarnings('ignore')
+
+    # custom module
+    # from clickcounter import clickcounter
+    from folder_management import create_folder, remove_files_folder
+    from sqlite_database import create_insert_table, query_table, insert_clickdata_table, check_for_table
+
+    # Functions
+    def random_string(N=7):
+        res = ''.join(random.choices(string.ascii_lowercase + string.digits, k=N))
+        return str(res)
+
+    # def display_products(dfp):
+    #     product_list = list(dfp['title'])
+    #     image_list = list(dfp['Image'])
+    #     price_list = list(dfp['price'])
+    #     org_price = list(dfp['originalPrice'])
+    #     disc_list = list(dfp['markdownPercent'])
+    #     dfp['Skus'] = dfp['Skus'].astype('object')
+    #     sku_list = list(dfp['Skus'])
+    #     url_list = list(dfp['url'])
+    #     clicked_sku, counter_dict = clickcounter(image_list, sku_list, price_list, disc_list, org_price, url_list, product_list)
+    #     return clicked_sku, counter_dict
+
+    # Main Program
+    #st.set_page_config('Product Recommender System', layout='wide')
+    st.title('Product Recommender System')
+    st.markdown('''<pre style="text-align:center">
+    <strong><span style="color:#6a8759">
+    It will take as inputs: Keyword and Publisher ID. 
+    Displays (upto) Top 5 Recommendations &amp; Collects Click Information. <br><br> </span></strong></pre>''',
+                unsafe_allow_html=True)
+
+    # Generating Session ID:
+    if 'store' not in st.session_state:
+        st.session_state.store = False
+
+        try:
+            df_session = pd.read_excel('df_session.xlsx')
+            df_row = pd.DataFrame({'ID': uuid.uuid4()}, index=[0])
+            df_session = pd.concat([df_session, df_row], ignore_index=True)
+            df_session.to_excel('df_session.xlsx', index=False)
+            df_sku = pd.DataFrame({'SessionID': pd.Series(dtype='object'), 'Keyword': pd.Series(dtype='object'),
+                                   'Skus': pd.Series(dtype='object'),
+                                   'Count': pd.Series(dtype='int')})  # Initializing the SKU-Count DataFrame
+            df_sku.to_excel('df_sku.xlsx', index=False)
+
+        except FileNotFoundError:
+            df_session = pd.DataFrame({'ID': uuid.uuid4()}, index=[0])  # Initializing the SKU-Count DataFrame
+            df_session.to_excel('df_session.xlsx', index=False)
+
+    df_session = pd.read_excel('df_session.xlsx')
+    session_id = list(df_session['ID'])[-1]
+    # st.sidebar.write(session_id) # Uncomment to display the Session ID
+
+    # Inputs
+
+    # selected_keyword = st.sidebar.text_input("Enter a single word as Keyword: ", 'aquaman')
+    # publisher_id = st.sidebar.text_input("Enter Publisher ID: ", '725895')
+    list_keywords = []
+    list_publisherid = []
+    for file in os.listdir('bizrate'):
+        keyword_publisherid = file.replace(".xml", "")
+        list_keywords.append(keyword_publisherid.split("_")[0])
+        list_publisherid.append(keyword_publisherid.split("_")[1])
+    # list_keywords = ['aquaman', 'superman', 'batman', 'shoes', 'electronics', 'wallet', 'movies', 'books']  # Temporary
+    list_publisherid = ['725895'] # For the time being Publisher ID is being HardCoded.
+    selected_keyword = st.sidebar.selectbox("Select a Keyword:", set(list_keywords))
+    publisher_id = st.sidebar.selectbox("Select a Publisher ID:", set(list_publisherid))
+
+    # st.write(selected_keyword)
+    # st.write(publisher_id)
+
+    # Query Recommended Data as per Keyword and Publisher ID
+
+    try:
+        rec_df = query_table('RecSysData', selected_keyword + "_" + publisher_id)
+    except pd.errors.DatabaseError:
+        # st.error('This Keyword Publisher ID Combo Doesnt Exist!')
+        rec_df = pd.DataFrame({}, columns=['title', 'Brand', 'url', 'Image', 'Skus', 'price', 'originalPrice',
+                                           'markdownPercent', 'totalPrice', 'condition', 'stock', 'relevancy'])
+
+    # Logic to Decide which SKU's to Display
+
+    # 1. If a SKU has already been clicked on. It cannot be displayed again.
+
+    try:
+        click_data_df = query_table('session_data', 'session_data')
+
+        if click_data_df.empty:
+            df_top_rec = rec_df.head(5)
+        else:
+            displayed_skus = list(click_data_df[click_data_df['session_id'] == session_id][
+                                      'Skus'])  # List of SKU's already displayed in the current session
+            # st.write(displayed_skus)
+            rec_df = rec_df[~rec_df['Skus'].isin(displayed_skus)]
+    except pd.errors.DatabaseError as e:
+        print(e)
+
+    # Top 5 Recommendations
+    #top_df = rec_df.head(5).reset_index()
+    top_df = rec_df.sample(5).reset_index() # random 5 recommendations
+
+    recs_to_display = 0
+    if len(top_df) > 5:
+        recs_to_display = st.sidebar.slider('Recommendations to Display', 1, 5)
+    elif len(top_df) > 1:
+        recs_to_display = st.sidebar.slider('Recommendations to Display', 1, len(top_df), len(top_df))
+    else:
+        recs_to_display = 1
+
+    # recs_to_display image width dictionary
+    image_width_dictionary = {
+        5: 200,
+        4: 240,
+        3: 280,
+        2: 320,
+        1: 400}
+
+    if recs_to_display > 0:
+        idx = 0
+        cols = st.columns(recs_to_display)
+        for col in cols:
+            with col:
+                try:
+                    title = top_df['title'][idx]
+                    img_link = top_df['url'][idx]
+                    sku = top_df['Skus'][idx]
+                    list_price = top_df['originalPrice'][idx]
+                    selling_price = top_df['price'][idx]
+                    discount = top_df['markdownPercent'][idx]
+                    st.image(top_df['Image'][idx], width=image_width_dictionary[recs_to_display])
+                    # st.write('SKU: {}'.format(sku))
+                    # st.write('SKU: {}'.format(sku))
+                    # st.write('SKU: {}'.format(sku))
+                    content = '''<p><strong> <a href={}>{}</a> </strong> <br>
+                                    <strong>SKU:</strong> {} <br>
+                                    <strong>S.P:</strong> $ {}<br>
+                                    <strong>Discount:</strong> % {} <br>
+                                    <strong>L.P:</strong> $ {} <br>
+                                    </p>'''.format(img_link, title, sku, selling_price, discount, list_price)
+                    st.markdown(content, unsafe_allow_html=True)
+                    st.button('Select', key=random_string(), on_click=insert_clickdata_table,
+                              args=([session_id, selected_keyword, publisher_id, sku, 1]))
+                except KeyError:
+                    st.info('No more recommendations to display for this Keyword-PublisherID Combo!')
+                    st.info('You can search for another keyword or reload the page!')
+            idx += 1
+    else:
+        st.text(
+            'No more recommendations to display for this Keyword-PublisherID Combo! You can search for another keyword or reload the page!')
+
+
+def display_analytics():
+    import pandas as pd
+    import numpy as np
+    import requests
+    from bs4 import BeautifulSoup
+    import urllib.parse
+    import os
+    import string
+    import random
+
+    import time
+
+    import streamlit as st
+    import uuid
+
+    import warnings
+
+    warnings.filterwarnings('ignore')
+
+    # custom module
+    # from clickcounter import clickcounter
+    from folder_management import create_folder, remove_files_folder
+    from sqlite_database import create_insert_table, query_table, insert_clickdata_table, check_for_table
+
+    # Main Program
+    #st.set_page_config('Analytics', layout='wide')
+    st.title('Analytics')
+
+    df = query_table('session_data', 'session_data')
+    # print(pd.Timestamp(df['clicked_at'][0]).date())
+    df['date'] = df['clicked_at'].map(lambda x: pd.Timestamp(x).date())
+    x_col_list = list(df.columns)
+    x_col_list.remove('count')
+    x_col_list.remove('clicked_at')
+    print(x_col_list)
+    column = st.selectbox('Select a Dimension:', x_col_list)
+
+    # Grouping the DataFrame
+    if column:
+        grouped_df = df.groupby([column]).sum().reset_index()
+        grouped_df.rename(columns={'count': 'clicks'}, inplace=True)
+        # st.dataframe(grouped_df.head())
+        col1, col2 = st.columns(2)
+        with col1:
+            # st.line_chart(grouped_df, x=column, y='clicks')
+            alt_line_chart = alt.Chart(grouped_df).mark_line(color='#3ac81e').encode(x=column, y='clicks')
+            st.altair_chart(alt_line_chart, use_container_width=True)
+            
+        with col2:
+            #st.bar_chart(grouped_df, x=column, y='clicks')
+            alt_bar_chart = alt.Chart(grouped_df).mark_bar(color='#3ac81e').encode(x=column, y='clicks')
+            st.altair_chart(alt_bar_chart, use_container_width=True)
+
+    st.title("Top Products")
+    agg_df = df[['keyword', 'publisherid', 'Skus', 'count']]
+    agg_df.rename(columns={'keyword': 'Keywords', 'publisherid': 'PublisherID', 'count': 'Clicks'}, inplace=True)
+    agg_df = agg_df.groupby(['Skus', 'Keywords', 'PublisherID']).sum().reset_index()
+    agg_df = agg_df.sort_values(by='Clicks', ascending=False).reset_index(drop=True)
+    st.dataframe(agg_df.head(20), width=1000)
+
+page_names_to_funcs = {
+    "—": intro,
+    "Generate Recommendations": generate_recommendations,
+    "Product Recommendations": display_recommendations,
+    "Analytics": display_analytics
+}
+
+page_name = st.sidebar.selectbox("Choose a Page", page_names_to_funcs.keys())
+page_names_to_funcs[page_name]()