pdf-processing-server/app/pdf/processor.py

from pymupdf import TEXTFLAGS_DICT, TEXT_PRESERVE_IMAGES
from typing import Tuple, Dict, List

import pymupdf

BBox = Tuple[float, float, float, float] # (x0, y0, x1, y1) or top-left and bottom-right corners
Origin = Tuple[float, float] # (x, y) or (x1, y0) or bottom-left corner

def extract_text(pdf_path: str):
    flags = TEXTFLAGS_DICT & ~TEXT_PRESERVE_IMAGES
    extracted_text = {}
    with pymupdf.open(pdf_path) as pdf_file:
        for page in pdf_file:
            text_dict = page.get_text('dict', flags=flags, sort=True)
            for block in text_dict['blocks']:
                for line in block['lines']:
                    for span in line['spans']:
                        original_text = span['text']
                        stripped_text = original_text.strip()
                        if not stripped_text:
                            continue
                        origin = span['origin']
                        bbox = span['bbox']
                        corrected_bbox = _correct_bbox(bbox, origin)
                        text_with_metadata = {
                            'original_text': span['text'],
                            'stripped_text': stripped_text,
                            'bbox': corrected_bbox,
                            'origin': origin,
                            'size': span['size'],
                            'font': span['font'],
                            'color': _int_to_rgbf(span['color']),
                        }
                        extracted_text.setdefault(page.number, []).append(text_with_metadata)
    return extracted_text

def replace_texts(pdf_path: str, output_path: str, replacement_data: Dict, preserve_original_fonts: bool = True):
    with pymupdf.open(pdf_path) as pdf_file:
        for page_number, page_replacements in replacement_data.items():
            page = pdf_file[int(page_number)]
            page_fonts = page.get_fonts()

            # Add redact annotation to remove the original text
            for replacement in page_replacements:
                page.add_redact_annot(replacement['bbox'])

            page.apply_redactions()

            # Insert the corrected text
            for replacement in page_replacements:
                original_text = replacement['original_text']
                stripped_text = replacement['stripped_text']
                corrected_text = replacement['corrected_text']
                replacement_text = original_text.replace(stripped_text, corrected_text)

                page.insert_text(
                    replacement['origin'],
                    replacement_text,
                    fontsize=replacement['size'],
                    fontname=_get_font_name(replacement['font'], page_fonts) if preserve_original_fonts else 'helvetica',
                    color=replacement['color'],
                )
        pdf_file.save(output_path, garbage=3, deflate=True, clean=True)

def _correct_bbox(bbox: BBox, origin: Origin) -> BBox:
    corrected_bbox = list(bbox)
    if bbox[3] != origin[1]:
        corrected_bbox[1] = origin[1] - (bbox[3] - bbox[1])
        corrected_bbox[3] = origin[1]
    return tuple(corrected_bbox)

def _int_to_rgbf(color: int) -> Tuple[float, float, float]:
    return (
        ((color >> 16) & 0xFF) / 255,
        ((color >> 8) & 0xFF) / 255,
        (color & 0xFF) / 255,
    )

def _get_font_name(keyword: str, fonts: List):
    for font in fonts:
        if keyword in font[3]:
            return font[4]
    return 'helvetica'