app.py

import json
import gradio as gr
from pdfminer.high_level import extract_pages
from pdfminer.layout import LTTextBoxHorizontal, LTFigure, LTImage
import os
import io
from PIL import Image
import pandas as pd
import pdfplumber
import tempfile
import traceback
import re

def save_image(element, images):
    try:
        if hasattr(element, 'stream') and element.stream:
            image_data = element.stream.get_rawdata()
            image = Image.open(io.BytesIO(image_data))
            image_filename = f"extracted_image_{len(images)}.png"
            image.save(image_filename)
            images.append({"filename": image_filename})
        else:
            print("No stream data for image element")
    except Exception as e:
        print(f"Error extracting image: {e}")

def detect_headers(text):
    """Detect headers in the text and format them."""
    lines = text.split('\n')
    formatted_text = ""
    header_patterns = [r"^\d+\.\s", r"^[A-Z\s]+$", r"^[A-Z][a-z]+\s\d"]

    for line in lines:
        if any(re.match(pattern, line.strip()) for pattern in header_patterns):
            formatted_text += f"# {line.strip()}\n"
        else:
            formatted_text += f"{line.strip()}\n"

    return formatted_text

def parse_pdf(pdf_file, output_format, progress=gr.Progress()):
    """
    Parses a PDF file, extracts text, tables, and images, and formats the output.

    Args:
        pdf_file: Path to the uploaded PDF file.
        output_format: Desired output format ("JSON", "Markdown", or "HTML").
        progress: Gradio Progress object for displaying progress.

    Returns:
        tuple: Extracted text and download data in the specified format.
            Returns an empty string and None if there is an error.
    """
    try:
        with open(pdf_file, 'rb') as file:
            text = ""
            tables = []
            images = []

            for page in extract_pages(file):
                for element in page:
                    if isinstance(element, LTTextBoxHorizontal):
                        text += element.get_text()
                    elif isinstance(element, (LTFigure, LTImage)):
                        print(f"Processing element: {type(element)}")
                        save_image(element, images)

            formatted_text = detect_headers(text)

            with pdfplumber.open(pdf_file) as pdf:
                for page_num, page in enumerate(pdf.pages):
                    for table in page.extract_tables():
                        try:
                            if len(table) > 0 and len(set(table[0])) != len(table[0]):
                                unique_columns = []
                                for col in table[0]:
                                    if col in unique_columns:
                                        col = f"{col}_{unique_columns.count(col)}"
                                    unique_columns.append(col)
                                df = pd.DataFrame(table[1:], columns=unique_columns)
                            else:
                                df = pd.DataFrame(table[1:], columns=table[0] if table[0] else None)
                            tables.append(df)
                        except Exception as e:
                            print(f"Error processing table: {e}")

            with tempfile.NamedTemporaryFile(mode="w", delete=False, suffix="." + output_format.lower()) as tmp:
                if output_format == "JSON":
                    json_data = {
                        "text": formatted_text,
                        "tables": [table.to_dict(orient='records') for table in tables if not table.columns.duplicated().any()],
                        "images": images
                    }
                    json.dump(json_data, tmp, ensure_ascii=False, indent=4)
                elif output_format == "Markdown":
                    markdown_text = f"# Extracted Text\n\n{formatted_text}\n\n# Tables\n"
                    for i, table in enumerate(tables):
                        if not table.columns.duplicated().any():
                            markdown_text += f"## Table {i+1}\n"
                            markdown_text += table.to_markdown(index=False) + "\n\n"
                    markdown_text += "\n\n# Images\n\n"
                    for image in images:
                        image_path = os.path.join(os.getcwd(), image["filename"])
                        markdown_text += f'![Image]({image_path})\n'
                    tmp.write(markdown_text)
                elif output_format == "HTML":
                    html_text = f"<p>{formatted_text}</p>\n\n<h2>Tables</h2>\n"
                    for i, table in enumerate(tables):
                        if not table.columns.duplicated().any():
                            html_text += f"<h2>Table {i+1}</h2>\n"
                            html_text += table.to_html() + "<br>"
                    html_text += "\n\n<h2>Images</h2>\n\n"
                    for image in images:
                        image_path = os.path.join(os.getcwd(), image["filename"])
                        html_text += f'<img src="{image_path}" alt="Image"><br>\n'
                    tmp.write(html_text)
                download_path = tmp.name

            return formatted_text, download_path

    except Exception as main_e:
        traceback.print_exc()  # Print full traceback to console
        print(f"A main error occurred: {main_e}")
        return "", None

iface = gr.Interface(
    fn=parse_pdf,
    inputs=["file", gr.Dropdown(["JSON", "Markdown", "HTML"])],
    outputs=[
        gr.Text(label="Output Text"),
        gr.File(label="Download Output")
    ],
    title="PDF Parser",
    description="Parse a PDF and choose the output format."
)

if __name__ == "__main__":
    iface.launch()  # Temporarily disable sharing for debugging