run1_skill-1

[SKILL]

name: pymupdf-visual-calendar-parsing description: Parse visual PDF calendars by extracting visual elements (horizontal lines, colored event blocks) and mapping their vertical Y-coordinates to time values using PyMuPDF (fitz).

When parsing visual calendars in PDF format, events and grid lines are typically drawn as vector graphics rather than plain text. You can use PyMuPDF (fitz) to extract text (for the timeline axis labels) and vector drawings (rectangles representing events, and lines representing time intervals).

1. Extracting Drawings (Lines and Rectangles)

Use page.get_drawings() to extract all vector shapes. You can identify horizontal grid lines and colored event blocks based on the drawing item properties.

import fitz

doc = fitz.open("calendar.pdf")
page = doc[0]
drawings = page.get_drawings()

horizontal_lines = set()
event_blocks = []

for d in drawings:
    # Colors are typically represented as RGB tuples/lists like [0.0, 0.0, 1.0] for blue
    fill_color = d.get("fill") 
    
    for item in d["items"]:
        if item[0] == "l": # It's a line
            p1, p2 = item[1], item[2]
            if p1.y == p2.y: # It's a horizontal line
                horizontal_lines.add(p1.y)
        elif item[0] == "re": # It's a rectangle
            rect = item[1] # A fitz.Rect object
            event_blocks.append({
                "y0": rect.y0,
                "y1": rect.y1,
                "color": fill_color
            })

# Sort lines from top to bottom
sorted_y_lines = sorted(list(horizontal_lines))

2. Extracting Timeline Text Labels

You must parse text from the PDF to identify the time values corresponding to the axis, and find their vertical coordinates.

text_dict = page.get_text("dict