sankey.py

import pandas as pd
import plotly.graph_objects as go
from tkinter import filedialog, Tk
import re
from collections import defaultdict

def select_file():
    """
    Open a file dialog to select a log file.
    """
    root = Tk()
    root.withdraw()
    file_path = filedialog.askopenfilename()
    root.destroy()
    return file_path

def process_log_data(file_path):
    """
    Process the log data to prepare for the Sankey diagram.
    """
    # Read the log data
    df = pd.read_csv(file_path)

    # Filter out anonymous users
    # df = df[df['id'] != 'anonymous']

    # Replace hex-like segments in URLs
    # df['name'] = df['name'].apply(replace_hex_segments)

    # Map URLs to nodes and count transitions
    url_to_node = {url: idx for idx, url in enumerate(df['name'].unique())}
    transitions = defaultdict(int)

    for user_id in df['id'].unique():
        user_data = df[df['id'] == user_id].sort_values(by='time')
        for i in range(len(user_data) - 1):
            source = url_to_node[user_data.iloc[i]['name']]
            target = url_to_node[user_data.iloc[i + 1]['name']]
            transitions[(source, target)] += 1

    return url_to_node, transitions

def replace_hex_segments(url):
    """
    Replace hex-like segments in URLs with a placeholder, maintaining the original delimiters.
    """
    # Using regex to find all segments and delimiters
    pattern = r'([/\.,\?&=]+)'
    segments = re.split(pattern, url)

    for i in range(0, len(segments), 2): # Skipping delimiters
        segment = segments[i]
        if re.fullmatch(r'[0-9a-fA-F\-_]+', segment):
            segments[i] = '{param}'

    return ''.join(segments)  # Joining with the original delimiters

def create_sankey_diagram(url_to_node, transitions):
    """
    Create a Sankey diagram using Plotly.
    """
    nodes = list(url_to_node.keys())
    links_source = [pair[0] for pair in transitions.keys()]
    links_target = [pair[1] for pair in transitions.keys()]
    links_value = list(transitions.values())

    fig = go.Figure(data=[go.Sankey(
        node=dict(
            pad=15,
            thickness=20,
            line=dict(color="black", width=0.5),
            label=nodes
        ),
        link=dict(
            source=links_source,
            target=links_target,
            value=links_value
        ))])

    fig.update_layout(title_text="User Page Transitions", font_size=10)
    fig.show()

# Main execution
file_path = select_file()
url_to_node, transitions = process_log_data(file_path)
create_sankey_diagram(url_to_node, transitions)