app.py

import streamlit as st
import os
from streamlit_option_menu import option_menu
import pandas as pd
import plotly.express as px
from plotly.subplots import make_subplots
import plotly.graph_objects as go
from streamlit_ace import st_ace
from streamlit_pandas_profiling import st_profile_report
import pandas_profiling



def set_data_files_session_object(file_name, file_path):
    if 'data_files' not in st.session_state:
        files_dictionary = {}
        files_dictionary[file_name] = file_path
        st.session_state['data_files'] = files_dictionary
    else:
        files_dictionary = st.session_state['data_files']
        files_dictionary[file_name] = file_path
        st.session_state['data_files'] = files_dictionary

def set_filtered_data_session_object(df, file_name):
    if 'filtered_data' not in st.session_state:
        filtered_data_dictionary = {}
        filtered_data_dictionary[file_name] = df
        st.session_state['filtered_data'] = filtered_data_dictionary
    else:
        filtered_data_dictionary = st.session_state['filtered_data']
        filtered_data_dictionary[file_name] = df
        st.session_state['filtered_data'] = filtered_data_dictionary

def set_dataframe_session_object(file_name, file_path):
    if 'data_frames' not in st.session_state:
        data_frame_dictionary = {}
        data_frame_dictionary[file_name] = pd.read_csv(file_path)
        st.session_state['data_frames'] = data_frame_dictionary
    else:
        data_frame_dictionary = st.session_state['data_frames']
        data_frame_dictionary[file_name] = pd.read_csv(file_path)
        st.session_state['data_frames'] = data_frame_dictionary
        
def save_file(file_object):
    file_path = os.path.join(os.getcwd(), "uploaded_files", file_object.name)
    with open(file_path, "wb") as f:
        f.write(file_object.getbuffer())
    
    set_data_files_session_object(file_object.name, file_path)
    set_dataframe_session_object(file_object.name, file_path)
    



def create_upload_file_component():
    uploaded_files = st.file_uploader("Upload one file at a time.", type=['csv', 'xls', 'xlsx', 'pkl', 'pdf'],
                                      accept_multiple_files=True)

    if uploaded_files:

        os.makedirs(os.path.join(os.getcwd(), "uploaded_files"), mode=0o777, exist_ok=True)
        for uploaded_file in uploaded_files:
            save_file(uploaded_file)

def create_component_to_add_target_func(selected_files, dfs, i):
    target_var_name = st.text_input("Name of the target variable",key="target_var" + str(i))
#     content = st_ace(language="python")
#     if content:
    code= "def f1(x): return str(x * 3)"
    exec(code)
    st.write(f1(3))

#         st.write(len(content.splitlines()))
#         exec(content)
#         code= "def f1(x): return str(x * 3)"
   
#         exec(code)
#         st.text(content)
#         st.write(f1(3))

def set_filtered_data(df,selected_files,i):
    action = "data_filter"
    col_to_filter = st.selectbox("Select the field to Filter on ", df.columns.values,
                                 key= action + "_col_filter_" + str(i))
    filter_operation = st.selectbox("Operation ",
                                    ['Greater Than', 'Equals', 'Less Than', "In", "In Between"],
                                    key=action + "_col_filter_op_" + str(i))
    selected_filter_vals = None
    
    if filter_operation:
        if filter_operation == 'In':
            selected_filter_vals = st.multiselect("Select Values to Filter on ", df[col_to_filter].unique(),
                                                  key=action + "_col_filter_val_" + str(i))
            if selected_filter_vals:
                filtered_df = df[df[col_to_filter].isin(selected_filter_vals)]
        elif filter_operation == 'Equals':
            selected_filter_vals = st.text_input("Enter a numeric value",
                                                 key=action + "_col_filter_val_" + str(i))
            if selected_filter_vals:
                filtered_df = df[df[col_to_filter] == selected_filter_vals]
        elif filter_operation == 'Greater Than':
            selected_filter_vals = st.text_input("Enter a numeric value",
                                                 key=action + "_col_filter_val_" + str(i))
            if selected_filter_vals:
                filtered_df = df[df[col_to_filter] > selected_filter_vals]
        elif filter_operation == 'Less Than':
            selected_filter_vals = st.text_input("Enter a numeric value",
                                                 key=action + "_col_filter_val_" + str(i))
            if selected_filter_vals:
                filtered_df = df[df[col_to_filter] < selected_filter_vals]
        elif filter_operation == 'In Between':
            selected_filter_vals = st.select_slider("Select range",
                                                    (df[col_to_filter].min(), df[col_to_filter].max()),
                                                    key=action + "_col_filter_val_" + str(i))
            if selected_filter_vals:
                filtered_df = df[df[col_to_filter] < selected_filter_vals]
        
        if selected_filter_vals:
            set_filtered_data_session_object(filtered_df,selected_files[i])
            st.write('data filtered',st.session_state['filtered_data'][selected_files[i]].shape)
#                 st.write(df.shape)
#                 st.write( st.session_state['filtered_data'][selected_files[i]].shape)


def create_component_for_analysis_for_single_df(selected_files, dfs, i):
    st.subheader(selected_files[i])
    df = dfs[selected_files[i]]

    filter_data = st.checkbox("Analyse on Filtered Data",key="filter_data_check"+str(i))
    
    if filter_data:
        set_filtered_data(df,selected_files,i)

    analysis_actions = st.multiselect("What analysis do you wish to do?",
                                      ['Summary of Data', 'Sample Data','Get Profile' ,'Univariate Analysis',
                                       'Bivariate Analysis'], key='analysis_action_' + str(i))
    if analysis_actions:
        
        df_for_analysis = st.session_state['filtered_data'][selected_files[i]] if filter_data else df
                                                                                          
        for action in analysis_actions:

            if action == 'Sample Data':
                clear_chart_type_session_var()
                st.write(df_for_analysis.sample(10))
            elif action == 'Get Profile':
                clear_chart_type_session_var()
                full_data_check = st.checkbox("Report on all columns",key="filter_data_check"+str(i))

                if full_data_check:
                    st.warning("This might take a lot of time to generate the report depending on the size of the data.Select a subset of columns")
                    confirm_full_run = st.button("Run on full data")
                    if confirm_full_run:
                        pr = df_for_analysis.profile_report()
                        st_profile_report(pr)
                else:
                    col_subset = st.multiselect("Select subset of columns", df.columns.values,key='filter_subset_'+ str(i))

                    if col_subset: 
                        pr = df_for_analysis[col_subset].profile_report()
                        st_profile_report(pr)
                
            elif action == 'Summary of Data':
                clear_chart_type_session_var()
                st.write(df_for_analysis.describe())
                
            elif action == 'Univariate Analysis':
                clear_chart_type_session_var()
                cols_for_analysis = st.multiselect("Select Columns for Univariate Analysis",options= df_for_analysis.columns.values)
                for col in cols_for_analysis:
                    if str(df_for_analysis[col].dtype) in ['int64','float64'] and df_for_analysis[col].nunique() > 10 :
                        
                        fig = px.scatter(x=df_for_analysis.index, y=df_for_analysis[col],labels=dict(x="Index", y=col))
                        st.plotly_chart(fig, use_container_width=True)

                    elif str(df_for_analysis[col].dtype) in ['object','category'] or df_for_analysis[col].nunique() <= 10:
                        
                        value_dist_df = df_for_analysis[col].value_counts(normalize=True)[:20].reset_index()
                        value_dist_df.columns = [col,'% Distribution']
                        
                        value_dist_df_counts = df_for_analysis[col].value_counts()[:20].reset_index()
                        value_dist_df_counts.columns = [col,'Count']
                        value_dist_df = value_dist_df.merge(value_dist_df_counts,on=col)
                        
                        trace1 = go.Bar(x=value_dist_df[col],y=value_dist_df['Count'],name='Count',marker=dict(color='rgb(34,163,192)'))
                        trace2 = go.Scatter(x=value_dist_df[col],y=value_dist_df['% Distribution'],name='% Distribution',yaxis='y2')

                        fig = make_subplots(specs=[[{"secondary_y": True}]])
                        fig.add_trace(trace1)
                        fig.add_trace(trace2,secondary_y=True)
                        
                        fig['layout'].update(height = 600, width = 800, title = f"{col} data distribution",xaxis=dict(tickangle=-90))
                        
#                         fig.update_layout(height=200, width=400, title_text=f"{col} data distribution")
                        
                        st.plotly_chart(fig, use_container_width=True)
            elif action == "Bivariate Analysis":
                add_chart_options_to_sidebar()
                create_for_bivariate_analysis(selected_files, df, i)                        

def clear_chart_type_session_var():
    if 'chart_type' in st.session_state:
        del st.session_state[chart_type]
    
def add_chart_options_to_sidebar():
    if 'chart_type' not in st.session_state :
        with st.sidebar:
            viz_type = st.radio("Graph Type",('None','Cross Tab','Pivot Table','Box Plot'))
        if viz_type and viz_type != 'None':
            st.session_state['chart_type'] == viz_type
    
def create_for_bivariate_analysis(selected_files, df, i):
        
    target_column = st.selectbox("Select the target column ", df.columns.values,
                                 key= "bivariate_target_column_" + str(i))
    bivariate_columns = st.multiselect("Select the columns to analyse ", df.columns.values,
                                 key= "bivariate_analysis_columns_" + str(i))
    
    col_vals = []
    
    if bivariate_columns:
        for col in bivariate_columns:
            col_vals.append(df[col])
            
        if st.session_state['chart_type'] == 'Cross Tab':
            if len(col_vals) > 3 :
                st.warning("Too many columns to split on. Please consider reducing the no of columns")
            crosstab_df = pd.crosstab(df[target_column], col_vals, margins=True)
            st.write(crosstab_df.to_html(),unsafe_allow_html=True)
# 3 any other aggregation function can be used based on column type
        
    
    
def create_component_for_data_analysis():
    if 'data_files' in st.session_state:

        selected_files = st.multiselect("Select the File(S) to analyze", st.session_state['data_files'].keys())

        if selected_files:
            cols = st.columns(len(selected_files))

            dfs = {}

            for selected_file in selected_files:
                if selected_file in  st.session_state['data_frames']:
                    dfs[selected_file] =  st.session_state['data_frames'][selected_file]
                else:
                    st.session_state['data_frames'][selected_file] = pd.read_csv(st.session_state['data_files'][selected_file])
                    dfs[selected_file] = st.session_state['data_frames'][selected_file]

            for i, col in enumerate(cols):
                with col:
                    create_component_for_analysis_for_single_df(selected_files, dfs, i)

    else:
        st.write("Upload a file to start analysis")


def main():
    
    st.title("Model Results Analyzer")
    with st.sidebar:

        selected_menu = option_menu(None, ["Home", "Upload Data", "Add Features","Analyze Data","Iframe"],
                                    icons=['house', 'cloud-upload', "list-task", 'gear'],
                                    menu_icon="cast", default_index=0, orientation="vertical",
                                    styles={
                                        "container": {"padding": "0!important", "background-color": "#fafafa"},
                                        "icon": {"color": "orange", "font-size": "15px"},
                                        "nav-link": {"font-size": "15px", "text-align": "left", "margin": "0px",
                                                     "--hover-color": "#eee"},
                                        "nav-link-selected": {"background-color": "green"},
                                    })

    if selected_menu == "Home":
        st.markdown('**This is to analyse models performance.**')

    elif selected_menu == "Upload Data":

        create_upload_file_component()

        if 'data_files' in st.session_state:
            st.write(pd.DataFrame(
                data={"File Name": pd.DataFrame.from_dict(st.session_state['data_files'], orient='index').index}))

    elif selected_menu == "Analyze Data":
        create_component_for_data_analysis()

    elif selected_menu == "Add Features":
        if 'data_files' in st.session_state:
            selected_file = st.selectbox("Select the File(S) to analyze", st.session_state['data_files'].keys())

            if selected_file:
                df =  st.session_state['data_frames'][selected_file]
                st.header("Enter the function definiton to create a new feature")
                feature_name = st.text_input("Enter the New Feature Name")
                st.warning("please retain the function signature as 'add_feature(row)'")

                content = st_ace(language="python",value="def add_feature(row):")

                if content != 'def add_feature(row):':
                    exec(content)
                    df[feature_name] = df.apply(lambda x:add_feature(x),axis=1)

                    st.session_state['data_frames'][selected_file] = df
                    st.write(df.columns.values)
    elif selected_menu == "Iframe":
        # st.components.v1.iframe("https://huggingface.co/spaces/Sasidhar/information-extraction-demo", width=None, height=None, scrolling=False)
        st.components.v1.iframe("https://docs.streamlit.io/en/latest", width=None, height=None, scrolling=False)

main()