app.py

import gradio as gr
import pandas as pd
import requests
from bs4 import BeautifulSoup
import re
import io

def get_scroll_depth_text_mapping(url):
    response = requests.get(url)
    soup = BeautifulSoup(response.text, 'html.parser')
    
    data = []
    elements = soup.select('p, table')
    total_height = len(soup.text)  # 簡易的な全体の高さの推定
    
    for i, element in enumerate(elements):
        text = element.text.strip().replace('\n', '').replace(',', '')
        if text:
            scroll_percent = (i / len(elements)) * 100
            data.append((scroll_percent, text))
    
    return data, total_height

def load_click_data(csv_content):
    return pd.read_csv(io.StringIO(csv_content), skiprows=9)

def load_exit_rate_data(csv_content):
    df = pd.read_csv(io.StringIO(csv_content), skiprows=9)
    exit_rate_data = df[['スクロールの奥行き', '% 人以上の訪問者']]
    exit_rate_data['スクロールの奥行き'] = exit_rate_data['スクロールの奥行き'].astype(int)
    exit_rate_data['離脱率'] = 100 - exit_rate_data['% 人以上の訪問者'].astype(float)
    return exit_rate_data[['スクロールの奥行き', '離脱率']]

def integrate_data(scroll_depth_text, click_data, exit_rate_data, total_height):
    integrated_data = []

    for scroll_percent, text in scroll_depth_text:
        clicks = 0
        click_rate = 0
        exit_rate = 0

        for index, row in click_data.iterrows():
            element_clicks = int(row['クリック:'])
            element_click_rate = float(row[' % のクリック'].strip('%'))
            
            # 簡易的なマッチング（実際のブラウザでの位置取得はできないため）
            if abs(scroll_percent - index / len(click_data) * 100) < 5:
                clicks += element_clicks
                click_rate += element_click_rate

        scroll_depth = int(scroll_percent // 5 * 5)
        exit_rate_row = exit_rate_data[exit_rate_data['スクロールの奥行き'] == scroll_depth]
        if not exit_rate_row.empty:
            exit_rate = exit_rate_row['離脱率'].values[0]

        integrated_data.append({
            'スクロール深度': f"{scroll_percent:.2f}%",
            'テキスト': text,
            'クリック数': clicks,
            'クリック率': click_rate,
            '離脱率': exit_rate
        })

    return pd.DataFrame(integrated_data)

def process_data(url, click_csv, exit_rate_csv):
    scroll_depth_text, total_height = get_scroll_depth_text_mapping(url)
    click_data = load_click_data(click_csv)
    exit_rate_data = load_exit_rate_data(exit_rate_csv)
    
    integrated_data = integrate_data(scroll_depth_text, click_data, exit_rate_data, total_height)
    
    output_csv = integrated_data.to_csv(index=False)
    return output_csv

iface = gr.Interface(
    fn=process_data,
    inputs=[
        gr.Textbox(label="URL"),
        gr.File(label="Click CSV"),
        gr.File(label="Exit Rate CSV")
    ],
    outputs=gr.File(label="Integrated Data CSV"),
    title="Web Page Data Integration",
    description="Integrate scroll depth, text, click, and exit rate data from a web page."
)

iface.launch()