Clean Code in Python
/me • Python developer • Interests ◦ Linux ◦ Software
development ◦ Software Architecture / system design marianoanaya at gmail dot com /rmariano @rmarianoa
def “You know you are working on clean code when
each routine you read turns out to be pretty much what you expected. You can call it beautiful code when the code also makes it look like the language was made for the problem.” Ward Cunningham In Python: magic methods → “Pythonic” code
Introduction / __init__ • What is “clean code”? ◦ Does
one thing well ◦ Every f(x) does what you’d expect • Why is it important? ◦ Code quality => Software quality ◦ Readability ◦ Agile development ◦ Code: blueprint
What is not clean code • Complex, obfuscated code •
Duplicated code • Code that is not intention revealing ...Technical Debt
Meaning def elapse(year): days = 365 if year % 4
== 0 or (year % 100 == 0 and year % 400 == 0): days += 1 for day in range(1, days + 1): print("Day {} of {}".format(day, year))
Meaning and logic separation def elapse(year): days = 365 if
year % 4 == 0 or (year % 100 == 0 and year % 400 == 0): days += 1 for day in range(1, days + 1): print("Day {} of {}".format(day, year)) ? def elapse(year): days = 365 if is_leap(year): days += 1 ... def is_leap(year): ...
Duplicated code • Often caused by the lack of meaningful
abstractions • Unclear patterns usually drive to code duplication Problems: • Hard to maintain, change, adapt • Error prone
DRY principle Don’t Repeat Yourself! • Avoid code duplication at
all cost • Proposed solution: decorators
Duplicated code: decorators def decorator(original_function): def inner(*args, **kwargs): # modify
original function, or add extra logic return original_function(*args, **kwargs) return inner General idea: take a function and modify it, returning a new one with the changed logic.
def update_db_indexes(cursor): commands = ( """REINDEX DATABASE transactional""", ) try:
for command in commands: cursor.execute(command) except Exception as e: logger.exception("Error in update_db_indexes: %s", e) return -1 else: logger.info("update_db_indexes run successfully") return 0
def move_data_archives(cursor): commands = ( """INSERT INTO archive_orders SELECT *
from orders WHERE order_date < '2016-01-01' """, """DELETE from orders WHERE order_date < '2016-01-01' """,) try: for command in commands: cursor.execute(command) except Exception as e: logger.exception("Error in move_data_archives: %s", e) return -1 else: logger.info("move_data_archives run successfully") return 0
def db_status_handler(db_script_function): def inner(cursor): commands = db_script_function(cursor) function_name = db_script_function.__qualname__
try: for command in commands: cursor.execute(command) except Exception as e: logger.exception("Error in %s: %s", function_name, e) return -1 else: logger.info("%s run successfully", function_name) return 0 return inner
@db_status_handler def update_db_indexes(cursor): return ( """REINDEX DATABASE transactional""", ) @db_status_handler
def move_data_archives(cursor): return ( """INSERT INTO archive_orders SELECT * from orders WHERE order_date < '2016-01-01' """, """DELETE from orders WHERE order_date < '2016-01-01' """, )
Implementation details • Abstract implementation details • Separate them from
business logic • We could use: ◦ Properties ◦ Context managers ◦ Magic methods
class PlayerStatus: ... def accumulate_points(self, new_points): current_score = int(self.redis_connection.get(self.key) or
0) score = current_score + new_points self.redis_connection.set(self.key, score) . . . player_status = PlayerStatus() player_status.accumulate_points(20)
class PlayerStatus: ... def accumulate_points(self, new_points): current_score = int(self.redis_connection.get(self.key) or
0) score = current_score + new_points self.redis_connection.set(self.key, score) . . . -- implementation details -- business logic
player_status.accumulate_points(20) player_status.points += 20 ... print(player_status.points) player_status.points = 100 The
kind of access I’d like to have
class PlayerStatus: @property def points(self): return int(self.redis_connection.get(self.key) or 0) @points.setter
def points(self, new_points): self.redis_connection.set(self.key, new_points) How to achieve it
@property • Compute values for objects, based on other attributes
• Avoid writing methods like get_*(), set_*() • Use Python’s syntax instead
Looking for elements class Stock: def __init__(self, categories=None): self.categories =
categories or [] self._products_by_category = {}
def request_product_for_customer(customer, product, current_stock): product_available_in_stock = False for category in
current_stock.categories: for prod in category.products: if prod.count > 0 and prod.id == product.id: product_available_in_stock = True if product_available_in_stock: requested_product = current_stock.request(product) customer.assign_product(requested_product) else: return "Product not available"
def request_product_for_customer(customer, product, current_stock): product_available_in_stock = False for category in
current_stock.categories: for prod in category.products: if prod.count > 0 and prod.id == product.id: product_available_in_stock = True if product_available_in_stock: requested_product = current_stock.request(product) customer.assign_product(requested_product) else: return "Product not available"
Python was made for the problem def request_product_for_customer(customer, product, current_stock):
if product in current_stock: requested_product = current_stock.request(product) customer.assign_product(request_product) else: return "Product not available"
The magic method product in current_stock Translates into: current_stock.__contains__(product)
Looking for elements class Stock: ... def __contains__(self, product): self.products_by_category()
available = self.categories.get(product.category) ...
Maintaining state • Some functions might require certain pre-conditions to
be met before running • … and we might also want to make sure to run other tasks upon completion.
Context Managers class DBHandler: def __enter__(self): stop_database_service() return self def
__exit__(self, *exc): start_database_service() ... with DBHandler(): run_offline_db_backup()
Context Managers class db_status_handler(contextlib.ContextDecorator): def __enter__(self): stop_database_service() return self def
__exit__(self, *exc): start_database_service() @db_status_handler() def offline_db_backup(): ... • Import contextlib • Python 3.2+
Pythonic A more Pythonic code, should blend with Python’s words.
if product in current_stock: Python’s mine
Summary • Python’s magic methods help us write more pythonic
code. ◦ As well as context managers do. ◦ Use them to abstract the internal complexity and implementation details. • Properties can enable better readability. • Decorators can help to: ◦ Avoid duplication ◦ Separate logic
Achieving quality code • PEP 8 ◦ Define coding guidelines
for the project ◦ Check automatically (as part of the CI) • Docstrings (PEP 257)/ Function Annotations (PEP 3107) • Unit tests • Tools ◦ Pycodestyle, Flake8, pylint, radon ◦ coala
More info • Python Enhancement Proposals: PEP 8, PEP 257,
PEP 343 ◦ https://www.python.org/dev/peps/ • Clean Code, by Robert C. Martin • Code Complete, by Steve McConnell • Pycodestyle: https://github.com/PyCQA/pycodestyle • PyCQA: http://meta.pycqa.org/en/latest/
Questions?