Highlight non-M rows in Hilldale report card PDF

highlight_report_card.py

#!/usr/bin/env uv run
# /// script
# dependencies = [
#   "pymupdf",
# ]
# ///

import fitz

def highlight_rows(input_pdf, output_pdf):
    doc = fitz.open(input_pdf)
    
    # Define colors (RGB values 0-1)
    # E -> Light Green
    color_e = (0.56, 0.93, 0.56)
    # R -> Yellow
    color_r = (1, 1, 0)
    # P -> Pink
    color_p = (1, 0.75, 0.8)

    rating_colors = {
        "E": color_e,
        "R": color_r,
        "P": color_p
    }

    for page in doc:
        # Get text lines
        # "dict" gives blocks -> lines -> spans
        # We want to flatten this to a list of lines with (bbox, text)
        
        all_lines = []
        blocks = page.get_text("dict")["blocks"]
        for block in blocks:
            if "lines" not in block:
                continue
            for line in block["lines"]:
                # Reconstruct text for the line
                text = " ".join([span["text"] for span in line["spans"]]).strip()
                all_lines.append({
                    "bbox": fitz.Rect(line["bbox"]),
                    "text": text
                })
        
        # Identify rating lines and content lines
        # Rating lines: Text is exactly E, R, P (or M, N, etc but we only care about E, R, P)
        # and positioned on the right side (x0 > 400 is a safe heuristic based on inspection)
        
        rating_candidates = []
        content_lines = []
        
        for line_obj in all_lines:
            text = line_obj["text"]
            bbox = line_obj["bbox"]
            
            # Check if it looks like a rating column entry
            # The inspection showed x0 around 525 for ratings. 450 is a safe cutoff.
            if bbox.x0 > 450 and len(text) <= 2: 
                if text in rating_colors:
                    rating_candidates.append(line_obj)
                # We ignore M, -, etc. for candidates but treat them as 'non-content' effectively?
                # Actually, if we have a rating 'M', we shouldn't treat it as content for another rating.
                # So we just don't add it to content_lines if it looks like a rating (short, right side).
            else:
                content_lines.append(line_obj)
                
        # For each target rating, find overlapping content
        for rating_line in rating_candidates:
            rating_text = rating_line["text"]
            r_bbox = rating_line["bbox"]
            
            # Find vertically overlapping content lines
            # Overlap: max(y0_1, y0_2) < min(y1_1, y1_2)
            
            overlapping_bboxes = [r_bbox]
            
            for content in content_lines:
                c_bbox = content["bbox"]
                
                # Check vertical overlap
                if max(r_bbox.y0, c_bbox.y0) < min(r_bbox.y1, c_bbox.y1):
                    overlapping_bboxes.append(c_bbox)
            
            # Calculate union bbox
            if overlapping_bboxes:
                # Union of all rectangles
                x0 = min([b.x0 for b in overlapping_bboxes])
                y0 = min([b.y0 for b in overlapping_bboxes])
                x1 = max([b.x1 for b in overlapping_bboxes])
                y1 = max([b.y1 for b in overlapping_bboxes])
                
                # Expand slightly for aesthetics or use page margins?
                # Using a fixed left margin (e.g. 36) to the rating's right edge looks neat.
                # Inspection showed left margin starts around 36 or 60.
                final_rect = fitz.Rect(36, y0, r_bbox.x1 + 5, y1)
                
                # Add highlight
                annot = page.add_highlight_annot(final_rect)
                annot.set_colors(stroke=rating_colors[rating_text])
                annot.update()

    doc.save(output_pdf)
    print(f"Processed PDF saved as: {output_pdf}")

if __name__ == "__main__":
    highlight_rows("Report Card.pdf", "Report Card_Highlighted.pdf")