Copies/splitting_int.py

import fitz  # PyMuPDF
from pypdf import PdfWriter
from pypdf import PdfReader
import os
import sys
import json
import shutil
from pathlib import Path
from collections import defaultdict

def decode_json(pdf_file):
    file_path = Path(pdf_file)
    with open(file_path.with_suffix(".json"), "r") as f:
        json_result = json.load(f)

    nb_pages = len(PdfReader(file_path).pages)

    bb_list = json_result["list"]
    name = json_result["name"]
    column_width = 1000 // nb_pages

    def page_number(b):
        return ((b[1] + b[3]) // 2) // column_width

    result = []
    for d in bb_list:
        (b, label) = d["box_2d"], d["label"]
        pn = page_number(b)
        carreau = 1000 // 38
        result.append((label, pn, b[0] - int(carreau), b[2]-int(carreau)))
    result.sort(key=lambda x: (x[1], x[2]))
    return (name, result)


def split_an_interro(base_dir, input_pdf, coords_list):
    doc = fitz.open(input_pdf)

    output_dir = base_dir / input_pdf.stem
    generated_files = set()
    parts_by_label = defaultdict(list)

    # Filter consecutive duplicate labels
    filtered_coords = []
    if coords_list:
        filtered_coords.append(coords_list[0])
        for item in coords_list[1:]:
            if item[0] != filtered_coords[-1][0]:
                filtered_coords.append(item)
    coords_list = filtered_coords

    def scale_coord(y, page):
        """Scale y from 0–1000 range to PDF points."""
        page_height = page.rect.height
        return (y / 1000) * page_height

    def save_cropped_page(doc, page_num, y0, y1, out_path):
        """Saves a cropped portion of a page as a new PDF."""
        page = doc[page_num]
        rotated_rect = page.rect * page.transformation_matrix
        visual_crop_rect = fitz.Rect(rotated_rect.x0, y0, rotated_rect.x1, y1)
        unrotated_clip_rect = visual_crop_rect * page.derotation_matrix

        temp_doc = fitz.open()
        temp_page = temp_doc.new_page(
            width=visual_crop_rect.width,
            height=visual_crop_rect.height
        )
        temp_page.show_pdf_page(
            temp_page.rect,
            doc,
            page_num,
            rotate=-page.rotation,
            clip=unrotated_clip_rect
        )
        temp_doc.save(out_path)
        temp_doc.close()

    # Iterate through all labels
    for idx, (title, start_page, y_start_raw, _) in enumerate(coords_list):
        temp_parts = []

        # Determine the stopping point for this label
        if idx + 1 < len(coords_list):
            # Normal case: stop at the next label
            _, end_page, _, y_end_raw = coords_list[idx + 1]
            end_y_target_raw = y_end_raw
        else:
            # FIX BUG 1: Last label extends to the very end of the document
            end_page = doc.page_count - 1
            end_y_target_raw = 1000  # 1000 represents full height

        # FIX BUG 2: Iterate through EVERY page from start to end
        # This handles cases where start_page == end_page, start_page + 1 == end_page,
        # AND start_page + N == end_page (gaps)
        current_p = start_page
        while current_p <= end_page:

            # Determine Top Cut (y0)
            if current_p == start_page:
                y0 = scale_coord(y_start_raw, doc[current_p])
            else:
                y0 = 0 # Start from top of page for intermediate/last pages

            # Determine Bottom Cut (y1)
            if current_p == end_page:
                y1 = scale_coord(end_y_target_raw, doc[current_p])
            else:
                y1 = doc[current_p].rect.height # Go to bottom of intermediate pages

            # Only save if the slice has height (avoid empty files)
            if y1 > y0 + 1:
                temp_path = f"_part_{idx}_{current_p}.pdf"
                save_cropped_page(doc, current_p, y0, y1, temp_path)
                temp_parts.append(temp_path)

            current_p += 1

        parts_by_label[title].extend(temp_parts)

    output_dir.mkdir(parents=True, exist_ok=True)

    # Process aggregated parts by label
    for title, parts in parts_by_label.items():
        merger = PdfWriter()
        for part in parts:
            if os.path.exists(part):
                merger.append(part)

        filename = f"{title}.pdf"
        merger.write(output_dir / filename)
        merger.close()
        generated_files.add(filename)

        # Cleanup
        for part in parts:
            if os.path.exists(part):
                os.remove(part)

    doc.close()

    # Move files not generated in this run to 'Missing' folder
    if output_dir.exists():
        missing_dir = output_dir / "Missing"
        for item in output_dir.iterdir():
            if item.is_file() and item.name not in generated_files:
                print(f"ALERT: File '{item.name}' not generated. Moving to {missing_dir}")
                missing_dir.mkdir(exist_ok=True)
                item.rename(missing_dir / item.name)


if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python script.py <directory or pdf_file>")
        sys.exit(1)

    input_arg = Path(sys.argv[1])

    if input_arg.is_file():
        base_dir = input_arg.parent
        pdf_files = [input_arg]
    elif input_arg.is_dir():
        base_dir = input_arg
        pdf_files = sorted(base_dir.glob("*.pdf"))
    else:
        print(f"Error: {input_arg} is not a valid file or directory.")
        sys.exit(1)

    for pdf_path in pdf_files:
            json_path = pdf_path.with_suffix(".json")
            # print("Debug :", json_path)
            if json_path.exists():
                (name, coords) = decode_json(pdf_path)
                print("Decoded name : ", name)
                split_an_interro(base_dir, pdf_path, coords)
            else:
                print(f"Warning: No JSON found for {pdf_path.name}")