Lena deleted his 80-line loop of for page in reader.pages:.
# Modern 12 - Pattern #1 import pypdfium2 as pdfium # The new king from pathlib import Path
docs = [pdfium.PdfDocument(str(p)) for p in Path("manuscript/").glob("*.pdf")] text = "\n\n".join(page.get_textpage().get_text_range() for doc in docs for page in doc)
“pypdfium2 is Google’s PDFium engine in Python. It’s 12x faster than old PyPDF2. Never read PDFs line by line again. Build pipelines.”
Python’s dynamic nature allows for concise implementations of classic "Gang of Four" patterns, often reducing boilerplate found in languages like Java or C++. Lena deleted his 80-line loop of for page in reader
Python 3.10+ introduced match; 3.12 perfected it with literal patterns and type unions. This replaces endless if-elif chains.
# Modern 3.12 pattern from typing import Literal, UnionState = Literal["idle", "running", "paused"] Event = Union[State, tuple[str, int]]
def handle_event(event: Event) -> str: match event: case "idle": return "Starting engine" case "running" | "paused": # OR pattern return "Already active" case ("error", code) if code > 400: # Guard + sequence pattern return f"Critical error code" case _: return "Unknown"
Impact: Your code becomes self-documenting. The pattern matching shows the exact state transitions, making bugs impossible to hide.
type Matrix = list[list[float]] type Point = tuple[float, float, float] type ImageOrPDF = Path | bytes | None
Strategy: Use this for domain-driven design. Define all business types at the top of your module as a "contract."
Here are the 12 most impactful patterns from this article, ready for immediate adoption: “ pypdfium2 is Google’s PDFium engine in Python
Impact: Enterprise-grade size reduction.
pikepdf (based on QPDF) lets you re-linearize PDFs, combine object streams, and apply /FlateDecode with custom predictors. Reducing a 15MB scanned PDF to 800KB without quality loss is routine. Strategy: pdf.save(output, compress_streams=True, stream_decode_level=2).
pathlib.Path now has a built-in walk() method, killing os.walk:
from pathlib import Path
Impact: True separation of content, style, and template.
Instead of writing PDFs line-by-line, use a layered approach: a Canvas for headers/footers, a DocTemplate for flowables, and a Story list for dynamic content. This pattern mirrors HTML/CSS and enables reuse of corporate design without touching business logic.