Python Cheat Sheet

A concise reference for modern Python 3 (3.11+) syntax, built-in types, and the most-used standard library modules — covers about 80% of everyday scenarios.

Python Python 3.11 (LTS-style)

CPython · Multi-paradigm (OOP · functional · procedural) · Dynamic · strong typing · duck typing

Recommended Learning Path

Start by running python3 and using the REPL -> master variables, built-in types, and control flow -> go deeper into functions, classes, and exceptions -> use the standard library for file and network I/O -> understand references and mutability (shallow vs deep copy) -> pick up asyncio, processes, and regex as needed -> finally learn venv/pip/pyproject for engineering workflows. Revisit the FAQ section whenever you hit a pitfall.

1.Hello World & Build Environment

Run Python, the interactive REPL, virtual environments, and package management.

Minimal Program

print writes to the console. Python organizes blocks with indentation — no braces or semicolons.

# hello.py

print('Hello, world!')

# Run: python3 hello.py

# Or interactively: launch python3 to enter the REPL and type directly

print(f'1 + 1 = {1 + 1}')

# Print to stderr:

import sys

print('error', file=sys.stderr)

Ways to Run

python3 runs a script, -c executes a string, -m runs a module. The interactive REPL evaluates immediately.

# Run a script:

# python3 hello.py

# Execute a string:

# python3 -c "print('hi')"

# Run a module (with __main__):

# python3 -m http.server 8000

# Check the version:

# python3 --version

# Enter the REPL:

# python3

# At the >>> prompt, type code and press Enter to run it

Virtual Environments

venv isolates project dependencies. After activation, pip installs affect only that environment. Always use a virtual environment per project.

# Create a virtual environment:

python3 -m venv .venv

# Activate (Windows):

# .venv\Scripts\activate

# Activate (macOS/Linux):

# source .venv/bin/activate

# Deactivate: deactivate

# Once activated:

# python -m pip install requests

# Check the interpreter:

# which python # Linux/macOS

# The venv name appears as a prompt prefix

pip Package Management

pip installs, uninstalls, and freezes dependencies. Pin with ==, range with >=,<y.

# Install:

pip install requests

# Pin a version:

pip install 'requests==2.31.0'

# Version range:

pip install 'requests>=2,<3'

# Upgrade: pip install -U requests

# Uninstall: pip uninstall requests

# Export dependencies:

pip freeze > requirements.txt

# Install from a file:

pip install -r requirements.txt

# List: pip list

# Upgrade pip: python3 -m pip install -U pip

pyproject.toml

The modern project config: dependencies, build backend, and tool config all live in pyproject.toml.

# pyproject.toml (PEP 621):

# [project]

# name = "my-app"

# version = "0.1.0"

# requires-python = ">=3.11"

# dependencies = [

# "requests>=2.31",

# "httpx>=0.24",

# ]

# [build-system]

# requires = ["setuptools>=68"]

# build-backend = "setuptools.build_meta"

# [tool.ruff]

# line-length = 88

# Install dependencies: pip install -e .

Shebang Scripts

#!/usr/bin/env python3 makes a script executable. On POSIX, chmod +x it and run it directly with ./script.py.

#!/usr/bin/env python3

# script.py

# Grant execute permission:

# chmod +x script.py

# Run it directly:

# ./script.py

print('可执行脚本')

# No need to hardcode a python path:

# env locates python3 for you

# On Windows a double-clicked .py uses the associated interpreter

# Without a shebang, run it as python3 script.py

REPL Interaction

Run python3 to enter the REPL for interactive debugging. _ holds the last result. Call exit() to leave.

>>> 1 + 2

>>> _

>>> def f(x): return x * 2

>>> f(21)

# Exit the REPL:

# exit() or Ctrl+D

# Built-in help:

>>> help(str)

>>> help('str'.upper)

# Inspect object attributes:

>>> dir('hello')

# History: browse with the arrow keys

main & Modules

if __name__ == '__main__' distinguishes running-as-a-script from being imported. This is the standard idiom.

# greetings.py

def hello(name='world'):

return f'Hello, {name}!'

if __name__ == '__main__':

# Runs only when executed directly

print(hello())

# Run directly: python3 greetings.py

# Import and use:

# from greetings import hello

# hello('Nick')

# __name__ is '__main__' when run directly, the module name when imported

2.Variables & Type Annotations

Dynamic typing, type annotations, multiple assignment, scope, and the walrus operator.

Assignment & Inference

No variable declarations — assignment creates the binding. Type is inferred dynamically and can change at any time.

x = 42 # inferred int

y = 3.14 # float

name = 'Rex' # str

ok = True # bool

empty = None # NoneType

# The type can change:

x = '现在是字符串'

# Several variables at once:

a = b = c = 0

# A variable is a name, not a box:

# it binds to an object; reassigning rebinds it

Type Annotations

PEP 484 style annotations with a colon to mark variable types. Annotations don't affect runtime — they're hints and enable checking.

count: int = 0

items: list[str] = []

mapping: dict[str, int] = {}

# PEP 604 union syntax (3.10+):

optional: int | None = None

# Function annotations:

def add(a: int, b: int) -> int:

return a + b

# Annotations are not enforced:

add('a', 'b') # no error at runtime

# Checkers: mypy / pyright

# Generics: list[int], not list[int, ...]

Multiple Assignment

Swap with a, b = b, a. Extended unpacking uses *rest. Counts must match.

a, b, c = 1, 2, 3

# Swap:

a, b = b, a

# Extended unpacking:

first, *rest = [1, 2, 3, 4]

# first=1, rest=[2,3,4]

*head, last = [1, 2, 3, 4]

# head=[1,2,3], last=4

# Unpack a tuple:

x, y = (10, 20)

# Unpack dict keys:

keys = 'ab'

d = dict(zip(keys, [1, 2]))

# A count mismatch raises ValueError

Constant Conventions

Python has no constant keyword. ALL_CAPS is the convention for module-level constants.

PI = 3.14159

MAX_RETRIES = 3

DATABASE_URL = 'postgres://...'

# Convention only, still mutable:

# PI = 99 allowed, but do not do it

# Why:

# values that never change stand out at a glance

# configuration stays in one place

# Use Enum instead of magic numbers:

# from enum import Enum

# class Color(Enum): RED = 1

LEGB Scope

Lookup order: Local -> Enclosing -> Global -> Built-in. Assigning in a function requires global or nonlocal.

counter = 0

def inc():

global counter # read/write module level

counter += 1

def outer():

x = 1

def inner():

nonlocal x # read/write the enclosing function's variable

x += 1

inner()

print(x) # 2

# Reading an outer variable needs no nonlocal

# Globals are defined at the top of the module

# Built-ins (len/print) live in the Built-in scope

Walrus Operator

The := operator assigns inside an expression (Python 3.8+). Avoids repeating a computation.

# Classic form:

line = input('> ')

while line != 'quit':

print(line)

line = input('> ')

# Walrus form:

while (line := input('> ')) != 'quit':

print(line)

# Reuse it inside a comprehension:

# [y for x in data if (y := f(x)) > 0]

# Mind the parentheses: an assignment expression must be wrapped

# Use case: initialise and test in a single condition

None & Booleans

None means "no value". Test truthiness with if x, not if x == True. Empty containers are falsy.

value = None

if value is None: # test None with is

print('无值')

# Truthiness:

if [1, 2]: pass # a non-empty list is true

if []: pass # an empty list is false

# Falsy: False/0/''/[]/{}/(,)/None

if not value: # when value is falsy

pass

# Ternary:

status = 'ok' if value else 'empty'

# Null-coalescing alternative:

name = value or '默认值'

del & Garbage Collection

del removes a name or a container element. The name becomes unbound and the object's reference count drops by one.

x = 42

del x # delete the name

# accessing x again raises NameError

items = [1, 2, 3]

del items[0] # delete an element -> [2, 3]

d = {'a': 1}

del d['a'] # delete a key

# pop deletes a dict key and takes a default:

val = d.pop('a', None)

# Pair it with try to catch KeyError

# del on a local often releases a large object early

3.Data Types

Built-in types: numeric, boolean, sequence, mapping, and set, plus advanced type annotations.

Numeric Types

int has arbitrary precision, float is IEEE-754, complex for complex numbers. bool is a subclass of int. Division has two semantics.

n = 10 # int

g = 3.14 # float

c = 1 + 2j # complex

big = 2 ** 100 # int, arbitrary precision

# Division:

5 / 2 # 2.5 (true division)

5 // 2 # 2 (floor division)

5 % 2 # 1 (remainder)

# Bases:

0b1010 # binary 10

0o12 # octal 10

0xff # hex 255

# Conversion:

int('42'); float(1); bool(0)

Booleans & None

True/False are capitalized. None is a singleton. and/or short-circuit and return the deciding operand.

t, f = True, False

# Short-circuit logic:

print(0 and 'x') # 0 (short-circuits to the left value)

print('' or '默认') # '默认'

print(None or []) # []

# Ternary:

age = 17

msg = '成年' if age >= 18 else '未成年'

# Chained comparison:

if 0 < x < 10: pass

# Testing None:

if x is None: pass

# bool() conversion:

bool([]) # False

bool([1]) # True

Sequence Types

str/list/tuple are all sequences: indexed, sliced, and iterable. str/tuple are immutable.

s = 'abc'

# Indexing:

s[0] # 'a'

s[-1] # 'c'

# Slicing:

s[1:] # 'bc'

s[::-1] # 'cba' reversed

# Common to all sequences:

len(s), min([3,1]), max([3,1])

sum([1,2,3])

list('abc') # ['a','b','c']

# Membership:

'a' in s # True

# Immutable sequences cannot be changed:

# s[0] = 'x' error

# An out-of-range index raises IndexError

Mapping dict

dict stores key/value pairs and preserves insertion order. Keys must be hashable. .get safely fetches a value.

d = {'a': 1, 'b': 2}

# Lookup:

d['a'] # 1, a missing key raises KeyError

d.get('c', 0) # 0, returns the default when missing

# Add / update:

d['c'] = 3

d.setdefault('x', 0)

# Delete:

d.pop('a', None)

# Merge:

d.update({'y': 9})

# Views:

d.keys(); d.values(); d.items()

# Dict comprehension:

{n: n**2 for n in range(3)}

# Keys must be hashable: numbers/strings/tuples work

Set Types

set is an unordered, deduplicating collection; frozenset is its immutable variant. Supports union/intersection/difference.

s = {1, 2, 3}

s.add(4)

s.remove(2) # a missing key raises KeyError

s.discard(99) # a missing key is ignored

# Deduplicate:

list(set([1, 1, 2])) # [1, 2]

# Operations:

{1, 2} | {2, 3} # union {1,2,3}

{1, 2} & {2, 3} # intersection {2}

{1, 2} - {2} # difference {1}

# Immutable set:

frozenset([1, 2])

# Use set() for an empty set; {} is an empty dict

Bytes Types

bytes is immutable, bytearray is mutable. Convert between text and bytes with encode/decode.

b = b'hello' # bytes

ba = bytearray(b'abc')

ba[0] = 120 # mutable

# Encoding:

text = '你好'.encode('utf-8')

# b'\xe4\xbd\xa0\xe5\xa5\xbd'

text.decode('utf-8') # '你好'

# Decoding errors:

text.decode('utf-8', errors='ignore')

# Converting between str and bytes always needs an encoding

# Binary data:

import struct

struct.pack('>i', 42)

# Byte order: > big-endian, < little-endian

Advanced Type Annotations

The typing module: Optional, Union, TypeVar, generic containers. Python 3.10+ supports the | syntax.

from typing import Optional, Union, TypeVar, Callable

# Optional:

x: Optional[int] = None # same as int | None

# Union:

y: Union[int, str] # same as int | str

# Type variable:

T = TypeVar('T')

def first(items: list[T]) -> T:

return items[0]

# Callable:

def apply(f: Callable[[int], int], n: int) -> int:

return f(n)

# Type alias (3.12+):

type Vector = list[float]

# Checkers: mypy / pyright

Duck Typing

"If it walks like a duck…": care about behavior, not type. typing.Protocol defines structural interfaces.

# Duck typing: any object with len will do

def size(obj):

return len(obj)

size([1, 2]) # 2

size('abc') # 3

# Protocols (typing.Protocol):

from typing import Protocol

class Sizeable(Protocol):

def __len__(self) -> int: ...

def total_len(objs: list[Sizeable]) -> int:

return sum(len(o) for o in objs)

# Structural subtyping: a matching shape is enough

# No explicit inheritance required

4.References & Mutability

Python's object reference semantics: mutable/immutable, shallow vs deep copy, and shared reference pitfalls.

Reference Semantics

Variables bind to objects, not boxes. Multiple names can point to the same object. Assignment rebinds, it does not copy.

a = [1, 2, 3]

b = a # b and a point at the same list

b.append(4)

print(a) # [1, 2, 3, 4]

# Rebinding does not touch the original object:

a = [9]

print(b) # still [1,2,3,4]

# Immutable objects such as ints:

x = 10

y = x

x = 20

print(y) # 10, the value did not change

# Key idea: separate mutating an object from rebinding a name

Mutable vs Immutable

Mutable types can be changed in place: list/dict/set. Immutable: int/str/tuple/frozenset.

# Mutable: modified in place

lst = [1, 2]

lst.append(3) # the same object changes

# Immutable: operations build a new object

s = 'ab'

s2 = s.upper() # a new string

# s is unchanged

# Immutable values can be dict keys:

d = {('a', 1): 'x'}

# A list cannot be a key:

# {[1]: 'x'} error

# A tuple holding mutable items is unhashable too

# Immutable objects are safe to share

Shallow Copy

copy.copy creates a shallow copy: a new outer container, but inner elements are still shared. Slicing a list is also shallow.

import copy

original = [[1, 2], [3, 4]]

shallow = copy.copy(original)

# The outer object differs:

shallow is original # False

# The inner objects are shared:

shallow[0] is original[0] # True

# A list slice behaves the same:

sub = original[:]

# Changing an inner item affects both:

shallow[0].append(99)

print(original) # [[1,2,99], [3,4]]

# A shallow copy suits one-level structures

Deep Copy

copy.deepcopy copies recursively. Edits to nested mutable structures don't leak across the copy. Watch the cost.

import copy

original = [[1, 2], [3, 4]]

deep = copy.deepcopy(original)

deep[0].append(99)

print(original) # [[1, 2], [3, 4]]

print(deep) # [[1, 2, 99], [3, 4]]

# A deep copy recurses:

# every nested object is brand new

# Cost grows with depth

# Cyclic references are handled

# Self-referencing objects: copy.deepcopy(x)

# Use with care on big data; rethink the structure

Shared Reference Pitfalls

When references share a mutable object, a mutation is visible everywhere. Argument passing is also by reference.

# Mutating a list passed into a function:

def add_item(items):

items.append('x') # affects the caller!

my_list = []

add_item(my_list)

print(my_list) # ['x']

# Pass a copy to avoid in-place changes:

add_item(my_list.copy())

# Or return a new list from the function

# See the FAQ for the default-argument trap

# Ask: does the function promise a read-only parameter?

is vs ==

== compares values; is compares identity (same object). Use is None to compare with None.

a = [1, 2]

b = [1, 2]

a == b # True, equal values

a is b # False, different objects

# Small-int interning:

x = 256

y = 256

x is y # True (-5..256 are cached)

# Large ints:

x, y = 10**6, 10**6

x is y # not guaranteed, do not rely on it

# Correct usage:

if x is None: pass

# Compare values with ==:

if x == 42: pass

Default Argument Pitfall

A mutable default is evaluated once at definition and shared across calls. Use None as a sentinel and create it inside the function.

# BAD: a shared default list

def add(x, acc=[]):

acc.append(x)

return acc

add(1) # [1]

add(2) # [1, 2] !

# GOOD: None + create it inside the function

def add(x, acc=None):

if acc is None:

acc = []

acc.append(x)

return acc

# A default value is evaluated once:

# created at definition time, reused ever after

Interning & Caching

Small ints and short strings are interned for memory savings. Don't rely on is to compare contents.

# Small-int cache range:

# -5 to 256 are singletons

x, y = 256, 256

x is y # True

# Outside that range:

a, b = 257, 257

a is b # unspecified (implementation detail)

# String interning:

'hello' is 'hello' # True (compile-time constant)

# Built at runtime it differs:

'he' + 'llo' is 'hello' # may be False

# Conclusion:

# always compare content with ==

# use is only for None and singletons

5.Control Flow

Conditional branches, loops, pattern matching, and comprehensions.

if / elif / else

Indentation delimits blocks. Use elif for multiple conditions. The condition can be any boolean expression.

score = 85

if score >= 90:

grade = 'A'

elif score >= 80:

grade = 'B'

elif score >= 70:

grade = 'C'

else:

grade = 'D'

# One-liner (not for complex logic):

if x > 0: print('正数')

# Multiple conditions:

if a and not b: pass

if x in (1, 2, 3): pass

Pattern Matching

match-case is structured pattern matching (Python 3.10+). Match and destructure by shape.

def handle(command):

match command.split():

case ['quit']:

print('退出')

case ['open', path]:

print(f'打开 {path}')

case ['open', path, *rest]:

print(f'打开 {path} 附加 {rest}')

case _:

print('未知命令')

# Guards:

def f(point):

match point:

case (0, 0): print('原点')

case (x, y) if x == y: print('对角')

case (x, y): print(f'({x}, {y})')

# The _ branch is the catch-all

for Loop

for iterates any iterable. enumerate adds the index, zip runs in parallel, dict.items() yields key/value pairs.

for i in range(5):

print(i) # 0..4

for idx, item in enumerate(['a', 'b']):

print(idx, item)

for k, v in {'a': 1}.items():

print(k, v)

for a, b in zip([1, 2], ['x', 'y']):

print(a, b) # (1,'x') (2,'y')

# range step:

range(0, 10, 2) # 0 2 4 6 8

# Reverse: reversed(lst)

# Reverse with index: range(len(x)-1, -1, -1)

while Loop

while loops while the condition is true. Use break to exit early — and avoid infinite loops.

n = 0

while n < 5:

print(n)

n += 1

# Infinite loop + break:

while True:

line = input('> ')

if line == 'quit':

break

print(line)

# while-else:

# else runs when the loop ends without break

attempts = 0

while attempts < 3:

if try_connect():

break

attempts += 1

else:

print('连接失败')

break & continue

break exits the whole loop; continue skips to the next iteration. The else clause runs when no break occurred.

for i in range(10):

if i % 2 == 0:

continue # skip even numbers

if i > 7:

break # stop early

print(i) # 1 3 5 7

# for-else:

for n in range(2, 10):

for d in range(2, n):

if n % d == 0:

break

else:

print(n, '是质数')

# else runs when the loop finishes normally

# Note: the else belongs to for, not to if

Comprehensions

List/dict/set comprehensions build collections in a single line — filter with conditions, transform with expressions.

# List comprehension:

squares = [n ** 2 for n in range(5)]

# [0, 1, 4, 9, 16]

# With a condition:

evens = [n for n in range(10) if n % 2 == 0]

# Dict comprehension:

{n: n * 2 for n in range(3)}

# {0:0, 1:2, 2:4}

# Set comprehension:

{n % 3 for n in range(10)}

# Nested:

flat = [y for xs in matrix for y in xs]

# Generator expression (lazy):

total = sum(n * n for n in range(100))

Truthiness

Use if x for truthiness. 0, '', [], and None are falsy. Avoid comparing explicitly to True/False.

# Testing empty containers:

items = []

if not items: # GOOD: empty is falsy

print('空列表')

# BAD: do not write this

if len(items) == 0: pass

# Testing None:

if value is not None:

print(value)

# Testing numbers:

if count: pass # 0 is falsy

# Strings:

if name: pass # '' is falsy

# Explicit bool:

if bool(x): pass

# Priority: readability first, keep conditions short

Iteration Idioms

itertools provides efficient iteration primitives: groupby, chain, product, and more.

from itertools import chain, groupby, product

# Concatenate:

list(chain([1, 2], [3, 4])) # [1,2,3,4]

# Cartesian product:

list(product('ab', [1, 2]))

# [('a',1),('a',2),('b',1),('b',2)]

# Group (input must be sorted):

rows = sorted(rows, key=lambda r: r['type'])

for key, group in groupby(rows, key=lambda r: r['type']):

print(key, list(group))

# Infinite iterator: itertools.count()

# Permutations/combinations: itertools.permutations / combinations

6.Functions

Function definitions, parameter passing, decorators, generators, and closures.

Function Definition

def defines a function with an indented body. return yields a value; omitting it returns None.

def greet(name):

return f'Hello, {name}!'

print(greet('Nick'))

# No return value:

def log(msg):

print(msg) # returns None

# Multiple return values (a tuple):

def min_max(nums):

return min(nums), max(nums)

lo, hi = min_max([3, 1, 4])

# Type annotations:

def add(a: int, b: int) -> int:

return a + b

# Docstring:

def f():

"""Describe what it does"""

pass

Parameter Types

Positional, keyword, and default arguments. A call can mix positional and keyword arguments.

def show(a, b, c):

print(a, b, c)

# Positional call:

show(1, 2, 3)

# Keyword call:

show(c=3, a=1, b=2)

# Mixed (positional first):

show(1, c=3, b=2)

# Default arguments:

def greet(name, greeting='Hi'):

return f'{greeting}, {name}!'

# Defaults must come last:

# def f(a=1, b) error

# Keyword-only parameters:

def f(*, strict=False): pass

# Positional-only parameters (3.8+):

def f(a, /): pass

*args & **kwargs

*args collects extra positional args into a tuple; **kwargs collects extra keyword args into a dict.

def log_all(*args, **kwargs):

print('位置:', args)

print('关键字:', kwargs)

log_all(1, 2, x=3, y=4)

# Forwarding a call:

def wrapper(*args, **kwargs):

return original(*args, **kwargs)

# Unpacking into a call:

items = [1, 2, 3]

print(*items) # 1 2 3

config = {'a': 1}

f(**config)

# Order: positional -> *args -> keyword -> **kwargs

# Do not overuse; prefer an object when there are many parameters

Lambda

lambda is a single-expression anonymous function, often used as a key for sort/map/filter.

square = lambda x: x * x

square(5) # 25

# Sort key:

users.sort(key=lambda u: u['age'])

# Higher-order functions:

list(map(lambda x: x * 2, [1, 2]))

list(filter(lambda x: x > 1, [1, 2]))

# Immediately invoked:

(lambda x: x + 1)(41) # 42

# Use def for complex logic; keep lambda to simple expressions

# A lambda cannot contain statements (if/return)

Closures

An inner function captures variables from the enclosing scope. Returning such a function is the closure pattern used in factories.

def make_multiplier(n):

def multiply(x):

return x * n # captures n

return multiply

double = make_multiplier(2)

triple = make_multiplier(3)

double(5) # 10

triple(5) # 15

# A closure that modifies an outer variable:

def counter():

count = 0

def inc():

nonlocal count

count += 1

return count

return inc

# Each closure keeps its own n/count

Decorators

@decorator wraps a function to enhance its behavior without changing its source — useful for timing and logging.

import functools, time

def timing(func):

@functools.wraps(func) # keep the metadata

def wrapper(*args, **kwargs):

start = time.perf_counter()

result = func(*args, **kwargs)

print(f'{func.__name__}: {time.perf_counter()-start:.4f}s')

return result

return wrapper

@timing

def slow():

time.sleep(0.1)

# Decorator with arguments:

def repeat(times):

def deco(func):

@functools.wraps(func)

def wrapper(*a, **kw):

for _ in range(times): func(*a, **kw)

return wrapper

return deco

# Built-in decorators: @staticmethod/@classmethod/@property

Generators

yield produces values lazily. Generators stream values one at a time — memory-friendly for large sequences.

def count_up(n):

i = 0

while i < n:

yield i # pause and hand back a value

i += 1

for x in count_up(3):

print(x) # 0 1 2

# Generator expression:

squares = (n * n for n in range(3))

# Infinite sequence:

def fib():

a, b = 0, 1

while True:

yield a

a, b = b, a + b

# Lazy: not everything is held in memory

# Single-use: rebuild it once exhausted

# Pipelines: chain generators to process a stream

Functional Tools

map/filter/reduce/sorted for functional style. functools adds partial and other tools.

from functools import partial, reduce

# map transforms:

doubled = list(map(lambda x: x * 2, [1, 2, 3]))

# filter selects:

evens = list(filter(lambda x: x % 2 == 0, range(6)))

# reduce aggregates:

total = reduce(lambda a, b: a + b, [1, 2, 3])

# Partial application: pin an argument

def power(base, exp): return base ** exp

square = partial(power, exp=2)

square(5) # 25

# Generators work with map/filter:

list(map(str, range(3))) # ['0','1','2']

# Modern style prefers comprehensions for readability

Recursion

A function that calls itself — always with a base case. CPython caps depth near 1000 (tweakable via sys.setrecursionlimit).

def factorial(n):

if n <= 1:

return 1

return n * factorial(n - 1)

factorial(5) # 120

# Recursion depth limit:

import sys

sys.getrecursionlimit() # 1000 by default

# Deep recursion: use iteration or an explicit stack

def fib_iter(n):

a, b = 0, 1

for _ in range(n):

a, b = b, a + b

return a

# Tree traversal / divide and conquer suit recursion

# Memoisation: @functools.lru_cache

7.Strings

Immutable str, f-string formatting, common methods, and encoding.

String Basics

Strings are created with single, double, or triple quotes. Escape sequences apply. str is immutable — operations return new strings.

s1 = '单引号'

s2 = "双引号"

s3 = '''多行

字符串'''

# Raw string (paths/regex):

r'C:\\path'

# Escapes:

'\n' # newline

'\t' # tab

'\\' # backslash

# Strings are sequences:

s = 'hello'

s[0] # 'h'

s[1:4] # 'ell'

# Immutable:

# s[0] = 'H' error

# Concatenation:

first + last

' '.join(['a', 'b'])

Common Methods

upper/lower for case, strip for whitespace, split/join for splitting and joining, startswith for prefix checks.

s = ' Hello World '

s.upper() # ' HELLO WORLD '

s.lower()

s.strip() # strip whitespace

s.startswith('He') # True

s.endswith('ld') # True

s.replace('World', 'Python')

# Search:

s.find('World') # index or -1

s.index('World') # ValueError if not found

s.count('l')

# Split / join:

'1,2,3'.split(',') # ['1','2','3']

'-'.join(['a', 'b']) # 'a-b'

# Tests:

'abc'.isalpha(); '123'.isdigit()

f-string Formatting

f"{var}" interpolates expressions. Specs control alignment, precision, and thousands separators. Python 3.12+ allows reusing quotes.

name = 'Nick'

age = 30

f'{name} 今年 {age} 岁'

# Expressions:

f'{age + 1}'

# Alignment width:

f'{name:>10}' # right-align, width 10

f'{name:<10}' # left-align

# Number formatting:

f'{3.14159:.2f}' # '3.14'

f'{1000000:,}' # '1,000,000'

f'{0.5:.0%}' # '50%'

# Base:

f'{255:x}' # 'ff'

# Date:

import datetime

f'{datetime.date.today():%Y-%m-%d}'

# Dict lookup: f'{d["key"]}'

format & % Formatting

str.format uses positional/named placeholders. The % operator is the legacy style. Prefer f-strings for new code.

# format method:

'{0} {1}'.format('a', 'b')

'{name} 今年 {age}'.format(name='Nick', age=30)

# Alignment:

'{:>10}'.format('hi') # right-align

'{:.2f}'.format(3.14159)

# Old-style % formatting:

'%s 今年 %d' % ('Nick', 30)

'%.2f' % 3.14159

'%x' % 255

# Three styles compared:

# f-string: f'{x:.2f}' (3.6+, recommended)

# format: '{:.2f}'.format(x)

# Old-style: '%.2f' % x

# Use f-string in new code

Split & Join

split on a separator, rsplit from the right, partition into a 3-tuple, join for efficient concatenation.

'a,b,c'.split(',') # ['a','b','c']

'a,b,c'.split(',', 1) # ['a', 'b,c']

'a b c'.split() # whitespace auto-split

'a-b-c'.rsplit('-', 1) # ['a-b', 'c']

'key=value'.partition('=') # ('key','=','value')

# Join (performance):

parts = ['a', 'b', 'c']

', '.join(parts)

# Per-character split:

list('abc')

# Multiple delimiters:

import re

re.split(r'[;,|]', 'a;b,c|d')

# Avoid + in loops (O(n^2))

Strip & Pad

strip/lstrip/rstrip trim whitespace. zfill zero-pads, center aligns, removeprefix/removesuffix strip known strings (3.9+).

s = ' hello '

s.strip() # 'hello'

s.lstrip() # 'hello '

s.rstrip() # ' hello'

' x \n'.strip() # 'x'

# Specific characters:

'xxhelloxx'.strip('x') # 'hello'

# Padding:

'42'.zfill(5) # '00042'

'hi'.center(7, '-') # '--hi---'

# Remove prefix/suffix (3.9+):

'/usr/bin'.removeprefix('/usr/') # 'bin'

'/usr/bin'.removesuffix('/bin') # '/usr'

# Newline cleanup:

text.rstrip('\n')

Unicode & Encoding

Python str is a sequence of Unicode code points. encode to bytes, decode back. len counts code points.

s = '你好👋'

len(s) # 3 (emoji is 1 codepoint)

# Encode:

b = s.encode('utf-8')

# Decode:

b.decode('utf-8')

# Codepoint:

ord('中') # 20013

chr(20013) # '中'

# Normalize:

import unicodedata

unicodedata.normalize('NFKC', s)

# Case:

s.casefold() # stricter lowercase

# Encoding error handling:

b.decode('utf-8', errors='replace')

# Iterate codepoints: for ch in s is already codepoint-level

Search & Replace

find/index locate substrings, replace substitutes, translate maps a translation table, count tallies occurrences.

s = 'the quick brown fox'

s.find('quick') # 4

s.find('z') # -1

s.index('quick') # 4 (raises if missing)

s.rfind('o') # search from right

s.count('o') # 4

# Replace:

s.replace('fox', 'dog')

s.replace('o', '0', 1) # replace only once

# Replace many (translate):

table = str.maketrans({'a': 'A', 'e': 'E'})

'apple'.translate(table) # 'ApplE'

# Case-insensitive replace:

import re

re.sub('(?i)the', 'THE', s)

# Extract substring: s[4:9]

Text Wrapping

textwrap formats text: wrap to a width, dedent common leading whitespace, fill to a paragraph.

import textwrap

long_text = '这是一个比较长的文本，需要按照指定宽度断行显示。'

# Wrap by width:

lines = textwrap.wrap(long_text, width=20)

# Fill into a paragraph:

textwrap.fill(long_text, width=20)

# Remove common indent:

block = '''\

line1

line2'''

textwrap.dedent(block)

# Indent:

textwrap.indent('a\nb', ' ')

# Use case: CLI help text layout

# Console output alignment

8.Collections & Containers

list/dict/set/tuple operations, the collections module, and sorting.

List Operations

append to the tail, insert anywhere, remove by value, pop the tail or an index, index to find, count to tally.

nums = [3, 1, 2]

nums.append(4) # [3,1,2,4]

nums.insert(0, 0) # [0,3,1,2,4]

nums.remove(1) # remove first 1

last = nums.pop() # pop tail

first = nums.pop(0) # pop at index

nums.index(2) # index

nums.count(1)

nums.reverse() # reverse in place

nums.clear() # clear

# Copy:

nums[:] or nums.copy()

# Extend:

nums.extend([5, 6])

nums += [7, 8]

Dict Operations

Fetch with .get, merge with update or |, iterate items, sort by key/value, and drill into nested dicts.

d = {'a': 1, 'b': 2}

d['c'] = 3 # add/update

d.get('x', 0) # default 0

val = d.setdefault('y', 0)

# Merge (3.9+):

merged = d | {'z': 9}

d.update({'z': 9})

# Iterate:

for k, v in d.items(): pass

# Sort by key:

sorted(d.items())

# Sort by value:

sorted(d.items(), key=lambda kv: kv[1])

# Nested:

d['user']['name']

# Safe chain access:

d.get('user', {}).get('name')

# defaultdict: collections.defaultdict

Set Operations

Union |, intersection &, difference -, symmetric difference ^. Membership test is O(1). Great for dedup.

a = {1, 2, 3}

b = {2, 3, 4}

a | b # {1,2,3,4} union

a & b # {2,3} intersection

a - b # {1} difference

b - a # {4}

a ^ b # {1,4} symmetric difference

# Test:

2 in a # True O(1)

# Subset:

a <= b; a < b

# Dedupe preserving order:

list(dict.fromkeys([3, 1, 3, 2]))

# Find duplicates:

[x for x in lst if lst.count(x) > 1]

# Immutable: frozenset

Tuple Usage

tuple is an immutable sequence — handy for multiple return values, records, and dict keys. namedtuple is a typed variant.

t = (1, 2, 3)

# Unpack:

a, b, c = t

# Single element needs comma:

single = (1,)

# Record use:

point = (10, 20)

x, y = point

# Named tuple:

from collections import namedtuple

Point = namedtuple('Point', ['x', 'y'])

p = Point(1, 2)

p.x; p[0] # two access styles

# Named tuple with type annotations:

# from typing import NamedTuple

# Immutable: t[0] = 9 raises

# As dict key (hashable)

collections Utilities

Counter for tallies, deque for double-ended queues, defaultdict for missing-key defaults, OrderedDict for ordered mappings.

from collections import Counter, deque, defaultdict

# Count:

Counter('abracadabra').most_common(2)

# [('a', 5), ('r', 2)]

# Double-ended queue:

q = deque([1, 2])

q.appendleft(0); q.pop(); q.popleft()

# Fixed length (drop oldest):

last5 = deque(maxlen=5)

# defaultdict:

dd = defaultdict(list)

dd['k'].append(1) # auto-create empty list

# Ordered dict:

# Python 3.7+ dict is ordered by default

# Counter arithmetic:

Counter('aab') + Counter('abb')

# Counter('aab') - Counter('b')

Sorting

sorted returns a new list; list.sort sorts in place. Provide a key callable. Set reverse=True to flip the order.

nums = [3, 1, 2]

sorted(nums) # [1,2,3] new list

nums.sort() # in-place sort

# Reverse:

sorted(nums, reverse=True)

# Sort by key:

words = ['banana', 'apple', 'cherry']

sorted(words, key=len)

# List of dicts:

users.sort(key=lambda u: u['age'])

# Multi-key sort:

sorted(users, key=lambda u: (u['age'], u['name']))

# Stable sort:

sorted(users, key=lambda u: u['age'])

# Descending key:

sorted(users, key=lambda u: -u['age'])

Slicing

seq[start:stop:step] slices. Negative indices count from the end. Reverse with step -1, copy with [:].

s = list(range(10))

s[2:5] # [2, 3, 4]

s[:3] # from start

s[7:] # to end

s[-3:] # last 3

s[::2] # every other

s[::-1] # reverse

s[1:8:2] # step

# Assign:

s[1:3] = [9] # replace slice

# Delete:

del s[1:3]

# Copy:

copy = s[:]

# Same for strings:

'hello'[::-1] # 'olleh'

# Note: slicing out of range does not raise

Heaps & Queues

heapq implements heaps; queue provides thread-safe queues — including priority queues.

import heapq

nums = [3, 1, 4, 1, 5]

heapq.heapify(nums) # build min-heap

heapq.heappop(nums) # pop smallest

heapq.heappush(nums, 0)

# Get largest/smallest n:

heapq.nlargest(2, [3, 1, 4])

heapq.nsmallest(2, [3, 1, 4])

# Priority queue:

import queue

pq = queue.PriorityQueue()

pq.put((1, '低'))

pq.put((0, '高'))

pq.get() # (0, '高')

# Thread-safe: queue.Queue()

# Task queue: queue.Queue(maxsize)

9.Memory & Performance

Garbage collection, reference counting, memory tools, and performance tuning.

Garbage Collection

Reference counting handles most reclamation; a generational GC breaks reference cycles. No manual free needed.

# Refcount hits zero -> collected

x = [1, 2]

y = x

del x # count -1, still alive

del y # count hits zero, collected

# Cyclic references:

import gc

gc.collect() # manually triggered

# Object finalizer hook:

class Temp:

def __del__(self):

print('回收')

# Generally no need to call gc manually

# Refcount can't handle cycles; generational GC catches them

memoryview & Zero-Copy

memoryview exposes a buffer without copying — efficient for large binary data.

data = bytearray(b'hello world')

view = memoryview(data)

view[0] # 104

view[6:11] # b'world'

# Slice is a view, not a copy:

sub = view[6:11]

sub[0] = 87 # modifications affect original

# Format:

mem = memoryview(b'\x00\x01')

mem.cast('H') # read as 16-bit unsigned

# Use cases:

# 1. Chunked large-file processing

# 2. Protocol parsing without copies

# 3. struct combined views

# Faster than slice-copy in perf-sensitive paths

Weak References

weakref.ref does not keep its referent alive. Useful for caches of large objects and for breaking reference cycles.

import weakref

class Node:

pass

node = Node()

ref = weakref.ref(node)

print(ref()) # the object

node = None # drop strong reference

gc.collect()

print(ref()) # None (already collected)

# WeakKeyDictionary/WeakValueDictionary:

# key/value weakly referenced, removed when collected

cache = weakref.WeakValueDictionary()

# Prevent cycles:

# replace one edge with a weak reference

# Use cases: cache, observer, singleton

slots for Memory

__slots__ declares fixed attributes and removes per-instance __dict__ — saving memory and speeding access for millions of instances.

# By default each instance has __dict__:

class Point:

def __init__(self, x, y):

self.x = x

self.y = y

# With __slots__:

class Point:

__slots__ = ('x', 'y')

def __init__(self, x, y):

self.x = x

self.y = y

p = Point(1, 2)

# p.z = 3 Error: undeclared attribute

# Saves: one dict per instance

# Significant for millions of small objects

# Tradeoffs: can't add attrs dynamically, no __dict__

Large-Data Processing

Process large data in chunks, iterate lazily, and never load it all at once. Stream files line by line.

# Stream a large file line-by-line:

with open('big.log') as f:

for line in f: # lazy per-line

process(line)

# Use generators for large lists:

def read_all_lines(path):

with open(path) as f:

for line in f:

yield line

# Process in batches:

def chunks(iterable, size):

from itertools import islice

it = iter(iterable)

while batch := list(islice(it, size)):

yield batch

for batch in chunks(range(100), 10):

process_batch(batch)

# Avoid list(map(...)) materializing everything

# numpy handles large numeric data more efficiently

Profiling

cProfile profiles by function and call count; timeit runs micro-benchmarks. Profile before optimizing.

import timeit, cProfile

# Microbenchmark:

timeit.timeit("'x' * 100", number=10000)

# Or CLI:

# python3 -m timeit "'-'.join(str(n) for n in range(100))"

# Full profiling:

# python3 -m cProfile -s cumulative script.py

# In-code profiling:

cProfile.run('my_func()', sort='cumulative')

# Timing decorator:

import time

def timed(f):

def w(*a, **k):

t = time.perf_counter()

r = f(*a, **k)

print(f.__name__, time.perf_counter() - t)

return r

return w

# Rule: profile first, then optimize

Performance Tips

Locals beat globals, comprehensions beat loops, set lookups are O(1). Don't recompute in tight loops.

# Local-variable caching:

import math

def f(n):

m = math.sqrt # local reference

return m(n)

# Set membership is O(1):

allowed = set(['a', 'b'])

if x in allowed: pass

# Comprehension vs loop:

# comprehensions are usually faster and cleaner

squares = [n * n for n in range(1000)]

# Avoid repeated computation:

# hoist invariants out of loops

# join for string concatenation

# numpy for large numeric data

# C extension / numba for hot paths

Common Memory Errors

MemoryError for out-of-memory, RecursionError for stack blow-ups, and reference cycles — diagnostic tools included.

# Stack overflow:

def infinite():

return infinite()

# RecursionError: maximum recursion depth

# Out of memory:

# MemoryError (loading too much at once)

# Diagnostics:

import tracemalloc

tracemalloc.start()

# ... code runs ...

snapshot = tracemalloc.take_snapshot()

top = snapshot.statistics('lineno')

print(top[:5])

# Object reference graph:

import objgraph # third-party

# objgraph.show_refs([obj])

# Fix:

# chunking, releasing refs, generators

10.Classes & Objects

class definition, inheritance, dunder methods, dataclass, and enum.

Class Basics

class defines a type. __init__ initializes instances. self refers to the instance; instance methods bind to it.

class Person:

def __init__(self, name, age):

self.name = name

self.age = age

def greet(self):

return f'Hi, {self.name}'

p = Person('Nick', 30)

p.greet()

# Class attribute (shared):

class Counter:

total = 0 # class-level

def __init__(self):

Counter.total += 1

# Instance attribute (per instance):

p.name

# Dynamic attribute:

p.email = '[email protected]'

# Type check: isinstance(p, Person)

Attribute Access

Instance attrs shadow class attrs. setattr/getattr/hasattr do dynamic access. property turns a method into a computed attribute.

class User:

default_role = 'user' # class attribute

def __init__(self, name):

self.name = name

# Lookup: instance -> class -> base

u = User('a')

u.default_role # 'user'

# Dynamic access:

hasattr(u, 'name') # True

getattr(u, 'name', None)

setattr(u, 'role', 'admin')

# property:

class Circle:

def __init__(self, r):

self._r = r

@property

def area(self):

return 3.14 * self._r ** 2

c = Circle(2)

c.area # method accessed like attribute

# property setter for validation

Three Kinds of Methods

Instance methods take self, class methods take cls, static methods take neither. Decorate with @staticmethod/@classmethod.

class Tool:

count = 0

def __init__(self, name):

self.name = name

Tool.count += 1

def instance_method(self):

return self.name # access instance

@classmethod

def class_method(cls):

return cls.count # access class

@staticmethod

def helper(x):

return x * 2 # no self/cls

# Call:

t = Tool('锤子')

t.instance_method()

Tool.class_method()

Tool.helper(4)

# classmethod for factories / counters

# staticmethod for utility functions

Inheritance

A subclass inherits methods from its parent. super() reaches the parent. Override methods as needed. Multiple inheritance follows MRO.

class Animal:

def __init__(self, name):

self.name = name

def speak(self):

return f'{self.name} 发出声音'

class Dog(Animal):

def speak(self):

return f'{self.name} 汪汪'

def fetch(self):

return f'{self.name} 捡球'

d = Dog('Rex')

d.speak() # override

# super() calls parent:

class Cat(Animal):

def __init__(self, name, color):

super().__init__(name)

self.color = color

# isinstance and issubclass

# Multiple inheritance uses MRO order

# Duck typing: prefer protocols over inheritance

Dunder Methods

Dunder methods customize built-in behavior: __repr__ for display, __eq__ for comparison, __len__ for length, __add__ for +.

class Vector:

def __init__(self, x, y):

self.x, self.y = x, y

def __repr__(self):

return f'Vector({self.x}, {self.y})'

def __add__(self, other):

return Vector(self.x + other.x, self.y + other.y)

def __eq__(self, other):

return (self.x, self.y) == (other.x, other.y)

def __len__(self):

return 2

v = Vector(1, 2)

str(v); v + v; v == Vector(1, 2)

# Common ones:

# __str__ for end users

# __getitem__ subscript access

# __call__ callable objects

# __bool__ truthiness

# __enter__/__exit__ context manager

dataclass

@dataclass auto-generates __init__, __repr__, __eq__, and more — a declarative way to write data classes.

from dataclasses import dataclass, field

@dataclass

class Point:

x: int

y: int

label: str = 'origin' # default

tags: list = field(default_factory=list)

p = Point(1, 2)

p2 = Point(1, 2)

p == p2 # True (auto __eq__)

repr(p) # Point(x=1, y=2, label='origin')

# Frozen (immutable):

@dataclass(frozen=True)

class Config:

debug: bool = False

# Orderable:

@dataclass(order=True)

class Item:

price: int

# Convert to dict:

from dataclasses import asdict, astuple

asdict(p)

# Inherited fields handled too

Enum

Enum defines a set of named constants. Supports str enums, auto numbering, and iteration. Replaces magic numbers.

from enum import Enum, auto

class Color(Enum):

RED = 1

GREEN = 2

BLUE = 3

class Status(Enum):

ACTIVE = auto() # auto-numbered

INACTIVE = auto()

# Access:

Color.RED

Color(1) # lookup by value

Color.RED.value # 1

Color.RED.name # 'RED'

# Iterate:

for c in Color: pass

# String enum:

class Mode(str, Enum):

FAST = 'fast'

SLOW = 'slow'

# Compare:

Color.RED is Color.RED # True (singleton)

# With description:

class Kind(Enum):

A = ('a', '描述')

def __init__(self, code, desc):

self.code = code

self.desc = desc

Context Managers

with manages resources. Implement __enter__/__exit__, or use @contextlib.contextmanager to write one as a generator.

class Timer:

def __enter__(self):

import time

self.start = time.perf_counter()

return self

def __exit__(self, exc_type, exc, tb):

import time

print(f'耗时 {time.perf_counter()-self.start:.4f}s')

return False # False: don't swallow exception

with Timer():

pass

# Simplified:

from contextlib import contextmanager

@contextmanager

def timer():

import time

t = time.perf_counter()

yield

print(f'耗时 {time.perf_counter()-t:.4f}s')

with timer():

pass

# with open(...) is a built-in context manager

# contextlib.suppress / ExitStack

11.Exception Handling

try-except, exception hierarchy, custom exceptions, and the with statement for resource management.

try / except

Catch and handle errors. except filters by type. Use as e to grab the exception object.

try:

num = int('abc')

except ValueError:

print('无法转换')

# Catch and access exception:

except ValueError as e:

print(e)

# Multiple types:

except (ValueError, TypeError):

pass

# Order: more specific first

try:

result = risky()

except (ValueError, KeyError) as e:

print(f'可预期错误: {e}')

except Exception as e:

print(f'未知错误: {e}')

# Uncaught exceptions propagate until crash

else & finally

else runs only when no exception was raised. finally runs no matter what — perfect for cleanup.

try:

data = parse(input_data)

except ValueError as e:

print(f'解析失败: {e}')

else:

print('解析成功') # runs only if no exception

finally:

cleanup() # always runs

# Combined use:

# try: code that may raise

# else: logic only when successful

# finally: cleanup (close/release)

# finally doesn't suppress propagation

# else or finally can be omitted

Exception Hierarchy

Exception is the base for most errors. Common ones: ValueError, TypeError, KeyError, IndexError.

try:

d['missing'] # KeyError

except KeyError:

pass

try:

[1][5] # IndexError

except IndexError:

pass

try:

int('x') # ValueError

except ValueError:

pass

try:

1 / 0 # ZeroDivisionError

except ZeroDivisionError:

pass

# Hierarchy:

# BaseException > Exception > specific errors

# KeyboardInterrupt/SystemExit are not Exception

# except Exception catches business errors

# Avoid bare except: (swallows even exits)

Custom Exceptions

Subclass Exception to define domain errors. Add fields and messages. Callers catch by type.

class ValidationError(Exception):

pass

class NotFoundError(Exception):

def __init__(self, entity, key):

self.entity = entity

self.key = key

super().__init__(f'{entity} 不存在: {key}')

# Raise:

def get_user(user_id):

if user_id == 0:

raise NotFoundError('用户', user_id)

return {'id': user_id}

# Catch:

try:

get_user(0)

except NotFoundError as e:

print(e.entity, e.key)

# Common base class:

class AppError(Exception): pass

# SpecificError(AppError)

raise & Chaining

raise throws deliberately. raise ... from ... chains the cause. A bare raise inside except re-throws the current exception.

def validate(age):

if age < 0:

raise ValueError('年龄不能为负')

# Reraise:

try:

risky()

except SomeError:

raise # reraise as-is

# Exception chaining (preserve cause):

try:

parse_file(path)

except FileNotFoundError as e:

raise RuntimeError(f'读取失败: {path}') from e

# Suppress chain:

raise NewError('x') from None

# Traceback (most recent call last):

# cause first, current after

# Preserve __cause__ for debugging

assert

assert checks invariants during development. Running with -O strips asserts — never use them for input validation.

def process(data):

assert data is not None, 'data 不能为空'

assert isinstance(data, list), '期望列表'

return len(data)

# Failed assert raises AssertionError

# Disable asserts:

# python3 -O script.py

# Use for:

# 1. Internal invariants

# 2. Debug-time argument checks

# 3. Type preconditions

# Don't use assert for:

# user input validation

# security boundary checks

# critical business logic

Exceptions & Logging

Use logging to capture exceptions with stack traces. logger.exception in an except block records the full traceback.

import logging

logging.basicConfig(level=logging.INFO)

try:

x = 1 / 0

except ZeroDivisionError:

logging.exception('除法出错') # with stack trace

# Or:

logger.error('出错', exc_info=True)

# Levels:

logging.debug/info/warning/error

# Config:

logging.basicConfig(

level=logging.DEBUG,

format='%(asctime)s %(levelname)s %(message)s',

)

# Structured:

logger.info('请求', extra={'id': req_id})

# Don't use print for logging (no level, no timestamp)

Error Handling Patterns

EAFP ("easier to ask forgiveness than permission") vs LBYL ("look before you leap"). Prefer guard clauses for early returns.

# EAFP: try first, catch on error

def parse_eafp(s):

try:

return int(s)

except ValueError:

return None

# LBYL: check before acting

def parse_lbyl(s):

if not s.isdigit():

return None

return int(s)

# Broad vs specific catch:

# Specific: except ValueError

# Fallback: except Exception

# Guard clauses for early return:

def handle(user):

if user is None:

return

if not user.active:

return

process(user)

# Reduces nesting, flatter logic

12.Input / Output

print/input, file I/O, pathlib, JSON/CSV, and standard streams.

Output & Input

print accepts many args with sep/end control. input reads one line as str. Format outputs with f-strings.

print('a', 'b') # a b

print('a', 'b', sep=', ') # a, b

print('end', end='!\n')

# Print to a file:

with open('out.txt', 'w') as f:

print('text', file=f)

# Input:

name = input('你的名字? ') # always str

# Numeric input:

age = int(input('年龄? '))

# Multiple inputs at once:

a, b = input().split()

# Output variables:

print(f'Hi {name}')

# Console progress:

print('\r进度 50%', end='')

# Flush buffer: print(..., flush=True)

File Read/Write

open modes: r read, w write, a append, b binary. read/readline/readlines for reading.

# Read:

f = open('data.txt', encoding='utf-8')

text = f.read() # all

line = f.readline() # one line

lines = f.readlines() # list of lines

f.close()

# Write:

f = open('out.txt', 'w', encoding='utf-8')

f.write('内容\n')

f.writelines(['a', 'b'])

f.close()

# Modes:

# 'r' read, 'w' overwrite write, 'a' append

# 'rb'/'wb' binary

# 'r+' read/write

# Always use with to auto-close

with & Files

with closes the file automatically — even on exceptions. The idiomatic way to read a file.

# Recommended:

with open('data.txt', encoding='utf-8') as f:

text = f.read() # auto-closes after use

# Line-by-line:

with open('data.txt', encoding='utf-8') as f:

for line in f:

print(line.rstrip())

# Write:

with open('out.txt', 'w', encoding='utf-8') as f:

f.write('hello')

# Append:

with open('log.txt', 'a') as f:

f.write('更多')

# Error handling:

try:

with open('missing.txt') as f:

pass

except FileNotFoundError:

print('文件不存在')

# Always pass encoding for cross-platform consistency

pathlib Paths

Path provides cross-platform paths with /, exists, mkdir, read_text, etc. The modern replacement for os.path.

from pathlib import Path

p = Path('data') / 'sub' / 'file.txt'

p.exists() # exists?

p.is_file(); p.is_dir()

p.mkdir(parents=True, exist_ok=True)

# Read/write:

text = Path('data.txt').read_text(encoding='utf-8')

Path('out.txt').write_text('hi', encoding='utf-8')

# Iterate:

for f in Path('.').glob('*.py'):

print(f.name)

# rglob recursive:

Path('.').rglob('*.json')

# Attributes:

p.name; p.suffix; p.stem

p.parent; p.absolute()

# Common dirs:

Path.home()

Path.cwd()

# Cleaner syntax, cross-platform

JSON Read/Write

json.loads/dumps serialize. Convert dict/list to/from JSON. Pass ensure_ascii=False to keep non-ASCII characters readable.

import json

data = {'name': 'Nick', 'tags': ['a', 'b']}

# Serialize:

text = json.dumps(data, ensure_ascii=False)

# Pretty:

json.dumps(data, indent=2, ensure_ascii=False)

# Deserialize:

obj = json.loads(text)

# File I/O:

with open('data.json', 'w', encoding='utf-8') as f:

json.dump(data, f, ensure_ascii=False, indent=2)

with open('data.json', encoding='utf-8') as f:

obj = json.load(f)

# Error:

try:

json.loads('invalid')

except json.JSONDecodeError as e:

print(e)

# Types: JSON object <-> dict

CSV Read/Write

The csv module handles comma-separated values. writer writes, reader and DictReader read.

import csv

rows = [['name', 'age'], ['Nick', 30], ['Anna', 25]]

# Write:

with open('data.csv', 'w', newline='', encoding='utf-8') as f:

writer = csv.writer(f)

writer.writerows(rows)

# Read:

with open('data.csv', newline='', encoding='utf-8') as f:

reader = csv.reader(f)

for row in reader:

print(row)

# Dict read/write:

with open('data.csv', encoding='utf-8') as f:

for row in csv.DictReader(f):

print(row['name'])

# Write dicts:

with open('o.csv', 'w', newline='') as f:

w = csv.DictWriter(f, fieldnames=['name'])

w.writeheader()

w.writerow({'name': 'x'})

# Watch quoting / newline handling

Standard Streams

sys.stdin/stdout/stderr handle pipe data. Read line by line from stdin.

import sys

# Read one line:

line = sys.stdin.readline()

# Read all lines (piped):

for line in sys.stdin:

sys.stdout.write(line.upper())

# Error stream:

print('error', file=sys.stderr)

# Pipe usage:

# cat data.txt | python3 filter.py

# Redirect:

# python3 filter.py > out.txt

# Encoding:

# PYTHONIOENCODING=utf-8 controls stream encoding

# input() is equivalent to sys.stdin.readline()

# Progress to stderr to avoid polluting stdout

Binary I/O

Binary mode reads and writes bytes. struct packs and unpacks. seek/tell provide random access.

# Read/write bytes:

with open('img.bin', 'rb') as f:

data = f.read()

with open('out.bin', 'wb') as f:

f.write(b'\x00\x01\x02')

# Fixed-size chunks:

with open('file.bin', 'rb') as f:

while chunk := f.read(1024):

process(chunk)

# struct pack:

import struct

packed = struct.pack('>I', 1024) # 4 bytes

value = struct.unpack('>I', packed)[0]

# Random access:

with open('db.bin', 'rb') as f:

f.seek(100) # seek to offset

data = f.read(10)

# Append: 'ab' mode

13.Common Pitfalls

The most common pitfalls in everyday Python development and the correct idioms.

Mutable Default Argument

A mutable default is created once and shared across calls. Use None as a sentinel.

# BAD: shared default list

def add(item, basket=[]):

basket.append(item)

return basket

add('a') # ['a']

add('b') # ['a', 'b'] !

# GOOD: None + create inside

def add(item, basket=None):

if basket is None:

basket = []

basket.append(item)

return basket

# Why: default is evaluated once at def time

# Impact: hits list/dict/set

is vs == Confusion

is compares identity, == compares values. Small ints / short strings may be interned — never use is for content.

# BAD: use is to compare values

a = [1, 2]

b = [1, 2]

a is b # False, but values match

# GOOD: == compares values

a == b # True

# Use is for None:

x = None

x is None # True

# Beware interning:

# 257 is 257 may be True in CPython

# but that's an implementation detail, don't rely on it

# Rule: == for values, is for None/singletons

Loop Closure Pitfall

A lambda in a loop captures the loop variable by reference (late binding). Bind the current value if you need it now.

# BAD: they all print 9

funcs = [lambda: i for i in range(10)]

funcs[0]() # 9

# GOOD: bind the current value

funcs = [lambda i=i: i for i in range(10)]

funcs[0]() # 0

# Or capture it through a factory:

funcs = [(lambda x: lambda: x)(i) for i in range(10)]

# Why: a closure captures the name, not the value

# Fix: default arguments are evaluated at definition time

# Note: the for variable still exists after the loop ends

String Concatenation

Loop + concat is O(n^2). Collect into a list and join — or build with io.StringIO for huge inputs.

# BAD: concatenating in a loop

s = ''

for i in range(10000):

s += str(i) # builds a new string every time

# GOOD: collect, then join

parts = []

for i in range(10000):

parts.append(str(i))

s = ''.join(parts)

# Or with a generator:

s = ''.join(str(i) for i in range(10000))

# A couple of + joins are fine:

name = first + ' ' + last

# Rule: use join when concatenating in a loop

# f-strings suit fragments you already know

Mutating a List While Iterating

Mutating a list while iterating it skips/repeats elements. Iterate over a copy, or build a new list.

# BAD: removing while iterating skips items

items = [1, 2, 3, 4]

for x in items:

if x % 2 == 0:

items.remove(x) # the indices shift

# GOOD: build a new list

items = [x for x in items if x % 2 == 1]

# Or iterate over a copy:

for x in items[:]:

if x % 2 == 0:

items.remove(x)

# Changing dict keys while iterating is just as dangerous

# Principle: the iterator is invalidated -> iterate a copy

# Or build a new container instead

Shadowing Built-ins

Naming a variable after a built-in shadows it. Avoid overwriting list, str, input, sum, etc.

# BAD: shadowing a built-in

def total(data):

sum = 0 # hides the built-in sum

for x in data:

sum += x

return sum

# Calling sum() later fails

# GOOD: pick a different name

total = 0

def total(data):

return sum(data)

# Common traps:

# list = [] shadows list()

# input = ... shadows input()

# same story for type/str/dict

# Comprehension variables do not leak (own scope since 3.x)

Swallowing Exceptions

A bare except catches everything — including KeyboardInterrupt. Narrow it to specific types, or at least log it.

# BAD: a bare except silently swallows everything

import os

try:

os.remove('f.txt')

except:

pass # you now know nothing

try:

value = int(x)

except Exception:

value = 0 # hides type errors

try:

risky()

except Exception:

print('出错了') # not even clear what failed

# GOOD: catch precisely + log it

import logging

try:

value = int(x)

except (ValueError, TypeError) as e:

logging.warning('转换失败 %s', e)

value = 0

# Never swallow KeyboardInterrupt/SystemExit

Misusing Globals

Assigning to a global inside a function creates a local. Add global to rebind it.

# BAD: looks like a global update but raises

count = 0

def inc():

count += 1 # UnboundLocalError

# Why: += is an assignment, so a local count is created

# GOOD: declare it global

def inc():

global count

count += 1

# Reading a global needs no declaration:

def show():

print(count)

# Do not overuse globals:

# pass arguments and return values

# or wrap the state in a class

# keep module-level config read-only

Copy & Sort

list.sort sorts in place; sorted returns a new list. Shared references can lead to accidental mutation.

# BAD: sort mutates in place

original = [3, 1, 2]

sorted_list = original.sort() # returns None

# and the original ends up sorted too

# GOOD: sorted returns a new list

original = [3, 1, 2]

sorted_list = sorted(original)

# Or copy first, then sort:

c = original[:]; c.sort()

# Shared references:

b = original

b.append(9) # original changes as well

# Copy when you need independence:

new_list = original[:]

# Mutating a parameter affects the caller:

# pass a copy instead

14.Concurrency & Async

threading, asyncio, the GIL, and choosing the right concurrency model.

Thread Basics

threading.Thread creates a thread. start launches it, join waits, daemon=True makes it a background thread.

import threading

def work(n):

print(f'线程 {n} 工作')

threads = []

for i in range(3):

t = threading.Thread(target=work, args=(i,))

threads.append(t)

t.start()

for t in threads:

t.join() # wait for it to finish

# Thread object with arguments:

threading.Thread(target=work, args=(1,), daemon=True)

# A daemon thread exits with the main thread

# Keep the thread count modest to avoid context-switch overhead

# Good enough for simple concurrency

Thread Class

Subclass Thread and override run. Use locks to share data safely — a clean way to package worker threads.

import threading

class Worker(threading.Thread):

def __init__(self, name):

super().__init__(name=name)

self.result = None

def run(self):

self.result = heavy_compute()

w = Worker('job-1')

w.start()

w.join()

print(w.result)

# Override run, never start

# Return data through instance attributes

# Exceptions in a child thread are silent:

# wrap the body of run in try-except and log

# Prefer ThreadPoolExecutor when you need return values

Locks & Synchronization

Lock guards shared state. with lock acquires and releases automatically. RLock is reentrant.

import threading

lock = threading.Lock()

counter = 0

def increment():

global counter

for _ in range(1000):

with lock: # acquires and releases for you

counter += 1

threads = [threading.Thread(target=increment) for _ in range(10)]

for t in threads: t.start()

for t in threads: t.join()

print(counter) # 10000 (the lock protects it)

# RLock: the same thread may acquire it again

# Condition variable: Condition

# Event: Event

# Semaphore: caps the level of concurrency

# Deadlock: inconsistent lock order, waiting while holding a lock

# Prefer a Queue to hand data over without locks

GIL & Parallelism

CPython's GIL prevents CPU-bound threading from going parallel. Threads still help when the workload is I/O-bound.

# GIL: the global interpreter lock

# Only one thread executes bytecode at any moment

# CPU-bound work gains nothing from threads:

# a compute task can even get slower

# I/O-bound work does benefit:

# the GIL is released while waiting on network/disk

# Real parallelism for CPU-bound work:

# 1. multiprocessing, separate processes

# 2. C extensions that release the GIL (numpy and friends)

# 3. asyncio, single-threaded async

# Deciding:

# bottleneck is I/O -> threads/asyncio

# bottleneck is CPU -> processes

Multiprocessing

multiprocessing parallelizes CPU-bound work with a process pool. ProcessPoolExecutor is the high-level wrapper.

from multiprocessing import Pool

def square(n):

return n * n

with Pool(4) as pool:

results = pool.map(square, range(10))

# ProcessPoolExecutor:

from concurrent.futures import ProcessPoolExecutor

with ProcessPoolExecutor(max_workers=4) as ex:

results = list(ex.map(square, range(10)))

# Separate memory: processes share nothing

# Communication: Queue/Pipe

# Startup costs more than a thread

# Watch the serialisation cost with big data

# Pair it with if __name__ == '__main__'

asyncio Basics

async def defines a coroutine, await suspends it, asyncio.run drives the event loop.

import asyncio

async def say_hi():

await asyncio.sleep(1)

print('hi')

asyncio.run(say_hi())

# await suspends the coroutine and yields control

# Running concurrently:

async def main():

task = asyncio.create_task(say_hi())

await task

asyncio.run(main())

# The event loop:

# schedules coroutines inside a single thread

# Blocking code stalls the loop:

# time.sleep -> await asyncio.sleep

# swap blocking libraries for async ones: httpx/aiohttp

# 3.11+ TaskGroup manages tasks

async / await

async declares a coroutine. await waits on a result. gather runs coroutines in parallel. asyncio.timeout sets deadlines.

import asyncio

async def fetch(url):

await asyncio.sleep(0.1)

return f'data from {url}'

# Sequential:

async def main():

a = await fetch('a')

b = await fetch('b')

# Parallel:

results = await asyncio.gather(

fetch('a'), fetch('b'), fetch('c'))

# Timeout:

try:

result = await asyncio.wait_for(fetch('a'), timeout=1)

except asyncio.TimeoutError:

print('超时')

# Creating tasks in bulk:

tasks = [fetch(u) for u in urls]

await asyncio.gather(*tasks)

# Results keep the input order; exceptions propagate

Async I/O

aiohttp / httpx make async HTTP requests. For very high-throughput I/O, async wins handily.

# httpx async client:

# pip install httpx

# import httpx

async def main():

async with httpx.AsyncClient() as client:

r = await client.get('https://example.com')

return r.status_code

# Mixing in synchronous code:

async def worker():

loop = asyncio.get_event_loop()

result = await loop.run_in_executor(None, blocking_fn)

# Async files:

import aiofiles

# async with aiofiles.open('f', 'r') as f:

# text = await f.read()

# Throughput improves markedly with heavy concurrent I/O

# Databases: asyncpg/aiomysql

Executors

ThreadPoolExecutor pools threads; ProcessPoolExecutor pools processes. Both expose submit() and map().

from concurrent.futures import ThreadPoolExecutor, as_completed

def work(n):

return n * 2

with ThreadPoolExecutor(max_workers=4) as ex:

futures = [ex.submit(work, i) for i in range(10)]

for f in as_completed(futures):

print(f.result()) # completion order

# map keeps the input order:

with ThreadPoolExecutor(4) as ex:

results = list(ex.map(work, range(10)))

# Exceptions:

# f.result() re-raises the original exception

# shutdown(wait=True) waits for the pool

# A thread pool suits I/O-bound work

# A process pool suits CPU-bound work

# 3.9+ cancellation: f.cancel()

15.Networking & Modules

Module system, HTTP requests, URL handling, and servers.

Modules & Imports

import loads a module. from x import y pulls specific names. A package is a directory with an __init__.py.

# Importing:

import os

import os.path

import json as js # alias

from datetime import datetime, timedelta

from collections import * # not recommended

# Relative imports (inside a package):

# from . import utils

# from ..other import x

# How importing works:

# sys.path is the search path

# cached: a module is loaded only once

# __main__ is the entry point

# Your own modules:

# utils.py in the same directory

# import utils

# Lazy import: import inside a function to cut startup time

HTTP Requests

The requests library: GET/POST, query params, headers, JSON bodies, response status.

import requests

# GET:

r = requests.get('https://api.example.com/users',

params={'page': 2}, timeout=5)

# Status:

r.status_code == 200

r.ok

# Data:

r.json() # dict

r.text # text

r.content # bytes

# POST:

resp = requests.post('https://api.example.com/login',

json={'user': 'nick'},

headers={'Authorization': 'Bearer x'},

timeout=10)

# Errors:

try:

r.raise_for_status()

except requests.HTTPError:

print('请求失败')

# Session (keeps cookies):

s = requests.Session()

urllib Standard Library

urllib.request makes HTTP requests without third-party libs. urllib.parse splits URLs.

import urllib.request

with urllib.request.urlopen('https://example.com') as resp:

data = resp.read()

print(resp.status)

# Building a request with parameters:

req = urllib.request.Request('https://api.com',

headers={'User-Agent': 'my-app'},

data=b'body')

with urllib.request.urlopen(req) as resp:

print(resp.read().decode('utf-8'))

# URL parsing:

from urllib.parse import urlparse, urlencode, quote

urlparse('https://a.com/p?q=1#f')

# ParseResult(scheme='https', netloc='a.com', path='/p', query='q=1')

urlencode({'a': 1, 'b': 'x y'}) # 'a=1&b=x+y'

quote('中文') # percent-encoding

# Reach for requests first; urllib is the stdlib fallback

Simple HTTP Server

http.server serves static files or quick prototypes. Subclass BaseHTTPRequestHandler to customize.

# Static file server:

# python3 -m http.server 8000

# Browse http://localhost:8000

# Bind an address:

# python3 -m http.server 8000 --bind 0.0.0.0

# Custom handler:

from http.server import HTTPServer, BaseHTTPRequestHandler

class Handler(BaseHTTPRequestHandler):

def do_GET(self):

body = b'{"ok": true}'

self.send_response(200)

self.send_header('Content-Type', 'application/json')

self.send_header('Content-Length', str(len(body)))

self.end_headers()

self.wfile.write(body)

def log_message(self, fmt, *args):

pass # stay quiet

HTTPServer(('', 8000), Handler).serve_forever()

# Use a framework in production: FastAPI/Flask

Socket Networking

socket provides raw TCP/UDP. Hand-roll protocols or inspect traffic here.

import socket

# TCP client:

with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:

s.connect(('example.com', 80))

s.sendall(b'GET / HTTP/1.1\r\nHost: example.com\r\n\r\n')

data = s.recv(4096)

print(data[:200])

# TCP server:

with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as server:

server.bind(('0.0.0.0', 9000))

server.listen(5)

conn, addr = server.accept()

with conn:

data = conn.recv(1024)

conn.sendall(b'pong')

# UDP: SOCK_DGRAM with sendto/recvfrom

# Timeout: s.settimeout(3)

# Prefer a framework for complex protocols

URLs & Parameters

urlparse parses URLs, urlencode builds query strings, quote percent-encodes, urljoin resolves relative paths.

from urllib.parse import urlparse, urlencode, quote, urljoin

# Parse:

p = urlparse('https://[email protected]:8080/path?q=1#sec')

p.scheme; p.netloc; p.path

p.query; p.fragment

# Build a query string:

params = urlencode({'q': 'python 教程', 'page': 2})

# 'q=python+%E6%95%99%E7%A8%8B&page=2'

url = f'https://api.com/search?{params}'

# Join:

urljoin('https://a.com/docs/', '../about')

# Decode:

from urllib.parse import unquote

unquote('%E4%B8%AD') # '中'

# Path safety:

quote('a/b', safe='') # escapes the slash

# Parse the parameters:

from urllib.parse import parse_qs

parse_qs('a=1&a=2&b=3') # {'a': ['1','2'], 'b': ['3']}

Config & Environment

os.environ exposes environment variables; python-dotenv loads a .env file. Keep config layered.

import os

# Read an environment variable:

port = int(os.environ.get('PORT', '3000'))

secret = os.environ['SECRET_KEY'] # KeyError when missing

# Set one:

os.environ['DEBUG'] = '1'

# Test one:

if os.environ.get('DEBUG') == '1':

print('调试模式')

# dotenv:

# pip install python-dotenv

from dotenv import load_dotenv

load_dotenv() # loads .env

# The .env file:

# PORT=3000

# SECRET_KEY=xxx

# Why:

# deployment config stays out of the code

# secrets never reach version control

# Validate config: fail loudly when something is missing

# Layers: default/dev/prod

API Call Patterns

Wrap requests with retries and rate limiting. Validate responses with pydantic for typed data.

import requests

from time import sleep

def api_get(url, retries=3):

for attempt in range(retries):

try:

r = requests.get(url, timeout=10)

if r.status_code == 429: # rate limited

sleep(2 ** attempt) # back off

continue

r.raise_for_status()

return r.json()

except requests.RequestException:

if attempt == retries - 1:

raise

sleep(2 ** attempt)

# Shared headers:

HEADERS = {'User-Agent': 'my-app/1.0'}

# Response validation (pydantic):

# pip install pydantic

# from pydantic import BaseModel

# class User(BaseModel):

# name: str

# age: int

# u = User.model_validate(r.json())

# Caching: lru_cache / diskcache

16.Date & Time

datetime, formatting, timedeltas, and time zones.

datetime Basics

datetime construction, date/time subclasses, attribute access, and comparison.

from datetime import datetime, date, time

# Current time:

now = datetime.now() # local

now_utc = datetime.utcnow() # deprecated, use timezone

# Construct:

d = datetime(2026, 8, 2, 12, 30, 0)

# Access:

d.year; d.month; d.day

d.hour; d.minute; d.second

# Weekday:

d.weekday() # 0=Monday

d.isoweekday() # 1=Monday

# Date or time only:

date(2026, 8, 2)

time(12, 30)

# Compare:

d > datetime(2026, 1, 1)

# An invalid month raises ValueError

Format & Parse

strftime formats, strptime parses. Codes: %Y year, %m month, %d day, %H hour.

from datetime import datetime

now = datetime.now()

# Formatting:

now.strftime('%Y-%m-%d') # '2026-08-02'

now.strftime('%Y-%m-%d %H:%M:%S')

now.strftime('%H:%M') # hour and minute

now.strftime('%A') # weekday name

# Parsing a string:

d = datetime.strptime('2026-08-02', '%Y-%m-%d')

d = datetime.fromisoformat('2026-08-02T12:00:00')

# Common directives:

# %Y 4-digit year, %y 2-digit year

# %m month, %d day, %H 24-hour, %I 12-hour

# %M minute, %S second, %f microsecond

# Localised weekday: map it yourself

# Timezone safety: attach the tz before formatting

timedelta

timedelta expresses duration. Add/subtract days and hours. Compute differences between dates.

from datetime import datetime, timedelta

now = datetime.now()

# Plus 7 days:

now + timedelta(days=7)

# Plus 2 hours:

now + timedelta(hours=2)

# Minus 30 minutes:

now - timedelta(minutes=30)

# Difference between two moments:

d1 = datetime(2026, 8, 2)

d2 = datetime(2026, 1, 1)

delta = d1 - d2

delta.days # 213

delta.total_seconds() # seconds

# Available units:

# weeks/days/hours/minutes/seconds

# Deltas compare directly:

if delta > timedelta(days=100):

print('超过百天')

# Month differences need your own maths:

# timedelta has no month unit

Time Zones

timezone handles fixed offsets; zoneinfo (3.9+) loads IANA names. Store in UTC, render in local.

from datetime import datetime, timezone, timedelta

# Fixed offset:

utc8 = timezone(timedelta(hours=8))

now_utc8 = datetime.now(utc8)

# Named timezone (3.9+):

from zoneinfo import ZoneInfo

shanghai = ZoneInfo('Asia/Shanghai')

now_sh = datetime.now(shanghai)

# Conversion:

utc_time = datetime.now(timezone.utc)

local_time = utc_time.astimezone(shanghai)

# Aware vs naive:

# only a value carrying tzinfo is aware

# comparisons need the same zone, or both aware

# In practice:

# store UTC in the database

# convert to the user's zone for display

# the tzdata package ships the zone database

Timestamps

Timestamps are Unix seconds. Convert with timestamp()/fromtimestamp(). Ideal for storage and comparison.

import time

from datetime import datetime, timezone

# Unix timestamp (seconds):

now_ts = time.time()

datetime.now().timestamp()

# datetime to timestamp:

dt = datetime(2026, 8, 2, tzinfo=timezone.utc)

dt.timestamp()

# Timestamp to datetime:

datetime.fromtimestamp(now_ts) # local

datetime.fromtimestamp(now_ts, timezone.utc) # UTC

# Milliseconds:

int(time.time() * 1000)

# Comparing timestamps:

if now_ts > last_ts: pass

# Storage:

# an integer count of unix seconds is the common standard

# Display: convert the timestamp and format it locally

# Precision: the float carries fractional seconds

Timing & Sleeping

time.sleep pauses, time.perf_counter measures. For cron-style jobs use the schedule library.

import time

# Wait:

time.sleep(1.5) # seconds

# Timing:

start = time.perf_counter()

# run the code...

elapsed = time.perf_counter() - start

print(f'{elapsed:.4f}s')

# Precision:

time.time() # wall clock, can be adjusted

time.perf_counter() # high-resolution and monotonic

# Timed loop:

def run_every(interval, fn):

while True:

fn()

time.sleep(interval)

# Scheduling libraries:

# pip install schedule

# import schedule

# schedule.every(10).minutes.do(job)

# while True: schedule.run_pending()

# Blocking vs async waiting:

# inside an event loop use asyncio.sleep

Date Utilities

Weekday detection, month start/end, relative dates, date ranges. The calendar module helps.

from datetime import datetime, timedelta

import calendar

d = datetime.now()

# Weekday test:

d.weekday() < 5 # a working day

# First day of the month:

d.replace(day=1)

# Last day of the month:

last_day = calendar.monthrange(d.year, d.month)[1]

d.replace(day=last_day)

# A sequence of dates:

start = datetime(2026, 8, 1)

for i in range(7):

day = start + timedelta(days=i)

print(day.strftime('%m-%d %a'))

# Natural-language parsing:

# pip install python-dateutil

# from dateutil.parser import parse

# parse('2026-08-02 10:30')

# Month arithmetic:

# dateutil.relativedelta

# from dateutil.relativedelta import relativedelta

# d + relativedelta(months=1)

Time Interval Checks

Check whether a time falls in a window, detect overlapping intervals, compute remaining time.

from datetime import datetime, timedelta

# Range test:

now = datetime.now()

start = datetime(2026, 1, 1)

end = datetime(2027, 1, 1)

if start <= now < end:

print('在窗口内')

# Overlapping ranges:

def overlaps(a1, a2, b1, b2):

return a1 < b2 and b1 < a2

# Time remaining:

dealine = now + timedelta(hours=2)

remaining = deadline - now

if remaining > timedelta(minutes=30):

print('时间充足')

# A scheduled window:

def in_working_hours(dt):

return 9 <= dt.hour < 18

# Boundaries:

# half-open ranges [start, end) avoid double counting

# normalise when a range crosses midnight

# Rate limiting: compare against the previous timestamp

17.Processes & System

subprocess, the sys/os modules, command-line arguments, and signals.

subprocess

subprocess.run launches external commands. Capture output, check the return code, and set a timeout.

import subprocess

# Run and capture the output:

result = subprocess.run(

['ls', '-l'],

capture_output=True, text=True)

print(result.returncode) # 0 means success

print(result.stdout)

print(result.stderr)

# Check for failure:

result.check_returncode() # non-zero raises CalledProcessError

# Timeout:

subprocess.run(['sleep', '10'], timeout=2)

# On timeout: TimeoutExpired

# Safety:

# pass an argument list, do not build a shell string

# shell=True carries an injection risk

# Live output: read line by line from subprocess.Popen

sys Module

sys.argv holds CLI arguments, sys.exit sets the exit code, sys.path the module search path.

import sys

# Command-line arguments:

sys.argv # ['script.py', 'arg1', 'arg2']

# Exit:

sys.exit(0) # success

sys.exit(1) # failure

# Module search path:

sys.path # a list

sys.path.append('/custom') # add an entry

# Version:

sys.version_info # (3, 11, ...)

sys.version

# Streams:

sys.stdin; sys.stdout; sys.stderr

# Platform:

sys.platform # 'win32' / 'darwin' / 'linux'

# Encoding:

sys.getdefaultencoding() # 'utf-8'

# Recursion limit:

sys.getrecursionlimit()

sys.setrecursionlimit(5000)

os Module

os covers environment, paths, and directory operations. os.getcwd for the current directory, os.mkdir to create one.

import os

# Environment variables:

os.environ['HOME']

# Directories:

os.getcwd() # current directory

os.chdir('/tmp') # change directory

os.mkdir('dir') # create a directory

os.makedirs('a/b/c', exist_ok=True)

# Files:

os.remove('file')

os.rename('a', 'b')

# Paths (os.path):

os.path.join('a', 'b') # cross-platform join

os.path.exists('f')

os.path.isdir('d')

os.path.dirname('/a/b/c') # '/a/b'

os.path.basename('/a/b/c') # 'c'

os.path.splitext('f.txt') # ('f', '.txt')

# Walking a tree:

for root, dirs, files in os.walk('.'):

print(root, files)

# Prefer pathlib over os.path

argparse

argparse parses arguments, generates help, supports defaults. The standard way to write CLI tools.

import argparse

parser = argparse.ArgumentParser(description='示例工具')

parser.add_argument('input', help='输入文件')

parser.add_argument('-v', '--verbose', action='store_true')

parser.add_argument('-n', '--count', type=int, default=1)

args = parser.parse_args()

print(args.input, args.verbose, args.count)

# Usage:

# python3 tool.py data.txt -v -n 3

# Help:

# python3 tool.py --help

# Options:

parser.add_argument('--out', default='out.txt')

# Mutually exclusive arguments:

# parser.add_mutually_exclusive_group()

# Subcommands:

# subparsers = parser.add_subparsers()

# Type validation: type=int converts automatically

Exit Codes

0 means success, non-zero failure. CI/scripts rely on exit codes. An unhandled exception exits with 1.

import sys

# Normal exit:

sys.exit(0)

# Failure:

sys.exit(1)

sys.exit('错误信息') # prints to stderr and exits with 1

# An uncaught exception exits with 1 automatically

# Custom codes:

sys.exit(2) # conventionally a usage error

# Inspecting the return code:

# shell: echo $? (POSIX)

# python3 script.py; echo $?

# From the caller:

import subprocess

r = subprocess.run(['python3', 'x.py'])

if r.returncode != 0:

print('失败')

# In CI any non-zero code fails the build

Files & Processes

Large file handling, file locks, temp files, atomic writes.

# Temporary file:

import tempfile, os

with tempfile.NamedTemporaryFile(mode='w', delete=True) as f:

f.write('data')

name = f.name

# Temporary directory:

with tempfile.TemporaryDirectory() as d:

print(d) # cleaned up automatically

# Atomic write (write a temp file, then rename):

import os

def atomic_write(path, content):

tmp = path + '.tmp'

with open(tmp, 'w') as f:

f.write(content)

os.replace(tmp, path) # atomic replacement

# File locks:

# pip install filelock

# from filelock import FileLock

# with FileLock('app.lock'):

# mutually exclusive work

# Large files: read line by line or in fixed blocks

Signal Handling

signal handles Ctrl+C and SIGTERM for graceful shutdown. POSIX mainly — Windows has limited support.

import signal, sys, time

# Graceful shutdown flag:

running = True

def handle_stop(signum, frame):

global running

print(f'收到信号 {signum}，正在退出...')

running = False

signal.signal(signal.SIGINT, handle_stop) # Ctrl+C

signal.signal(signal.SIGTERM, handle_stop) # kill

# Main loop:

while running:

work()

time.sleep(1)

print('清理完成')

# Ignore a signal:

signal.signal(signal.SIGPIPE, signal.SIG_IGN)

# Timeout signal (POSIX):

# signal.alarm(5) raises SIGALRM after 5 seconds

# Windows supports only some signals

# Graceful shutdown: save state, close connections

Daemons & Backgrounding

Run in the background, rotate logs, supervise with supervisor or systemd.

# Run in the background:

# python3 script.py &

# Redirect the output:

# python3 script.py > out.log 2>&1 &

# nohup ignores hangups:

# nohup python3 script.py &

# Log rotation:

import logging

from logging.handlers import RotatingFileHandler

handler = RotatingFileHandler(

'app.log', maxBytes=1_000_000, backupCount=5)

logging.basicConfig(handlers=[handler])

logging.info('运行中')

# Supervision in production:

# a systemd unit / supervisor restart policy

# Enforcing a single instance:

# check a pid lock file

# Resource limits: ulimit

18.Regular Expressions & Text

The re module: matching, searching, replacing, groups, and compilation.

match & search

re.match anchors at the start; re.search scans the whole string. Returns a Match or None.

import re

# search: anywhere in the string

m = re.search(r'\d+', '订单号 123')

if m:

print(m.group()) # '123'

# match: only from the start

m2 = re.match(r'\d+', '123abc')

# m2.group() == '123'

# No match returns None:

if re.search(r'xyz', 'abc'): pass

# Position info:

m.start(); m.end(); m.span()

# Word boundaries:

re.search(r'\bcat\b', 'a cat eats')

# Metacharacters must be escaped:

r'\.' matches a literal dot

findall & finditer

findall returns all matches as a list. finditer yields them one at a time (memory-friendly for large inputs).

import re

# findall: every match

nums = re.findall(r'\d+', 'a1b22c333')

# ['1', '22', '333']

# With groups it returns tuples:

re.findall(r'(\d+)-(\d+)', '1-2,3-4')

# [('1','2'), ('3','4')]

# finditer: iterate over Match objects

for m in re.finditer(r'\d+', 'a1b22'):

print(m.group(), m.span())

# Use finditer when you need positions

# finditer saves memory on large text

# No match gives an empty list / empty iterator

sub Replacement

re.sub performs regex replacement. Backreferences \1 … \9 in the replacement; pass a function for dynamic logic.

import re

# Simple replacement:

re.sub(r'\s+', ' ', 'a b c') # 'a b c'

# Group references:

re.sub(r'(\d{4})-(\d{2})-(\d{2})', r'\3/\2/\1',

'2026-08-02')

# '02/08/2026'

# Replacement function:

re.sub(r'\d+', lambda m: str(int(m.group()) * 2),

'1 and 2')

# '2 and 4'

# Limit the count:

re.sub(r'a', 'x', 'aaa', count=1) # 'xaa'

# Case-insensitive:

re.sub(r'hello', 'hi', 'HELLO', flags=re.I)

split

re.split cuts on a regex. Supports multiple delimiters, retains captures, and caps splits.

import re

# Split on comma/semicolon/pipe:

re.split(r'[,;|]', 'a,b;c|d')

# ['a', 'b', 'c', 'd']

# Split on whitespace (runs included):

re.split(r'\s+', 'a b c')

# A capture group keeps the separator:

re.split(r'(,)', 'a,b,c')

# ['a', ',', 'b', ',', 'c']

# Limit:

re.split(r',', 'a,b,c', maxsplit=1)

# ['a', 'b,c']

# Split into lines:

re.split(r'\r?\n', text)

# By paragraph:

re.split(r'\n\s*\n', text)

# Use re.split when str.split cannot handle several separators

Groups

() captures; (?P<name>) names a group; (?:) groups without capturing; | alternation.

import re

# Groups:

m = re.search(r'(\w+)@(\w+)\.(\w+)', '[email protected]')

m.group(0) # '[email protected]', the whole match

m.group(1) # 'a'

m.group(2) # 'b'

m.groups() # ('a','b','com')

# Named groups:

m = re.search(r'(?P<user>\w+)@(?P<domain>\w+)', '[email protected]')

m.group('user') # 'a'

m.groupdict() # {'user': 'a', 'domain': 'b'}

# Non-capturing:

re.search(r'(?:ab)+', 'abab')

# Alternation:

re.search(r'cat|dog', 'a dog')

# Group reference in a replacement: \1

# Nested groups are numbered by opening parenthesis

Flags

re.I case-insensitive, re.M multiline anchors, re.S dot matches newlines, re.X verbose.

import re

# re.I ignores case:

re.search(r'hello', 'HELLO', re.I)

# re.M makes the anchors multiline:

re.search(r'^line', 'a\nline b', re.M)

# re.S lets the dot match newlines:

re.search(r'a.b', 'a\nb', re.S)

# re.X verbose (comments and whitespace):

pattern = re.compile(r'''

(\d{4}) # year

[-/]

(\d{2}) # month

''', re.X)

# Combining flags:

re.search(r'x', 'X', re.I | re.M)

# Compile once and reuse:

pat = re.compile(r'\d+', re.IGNORECASE)

pat.search('a1')

# Precompiling is faster and pins the options

Common Patterns

Common regex snippets for email, phone, URL, IP, Chinese, etc. Validate business formats.

import re

# Email:

EMAIL = r'^[^\s@]+@[^\s@]+\.[^\s@]+$'

# Chinese mainland mobile number:

MOBILE = r'^1[3-9]\d{9}$'

# URL:

URL = r'^https?://[^\s]+$'

# IPv4:

IP = r'^(\d{1,3}\.){3}\d{1,3}$'

# Chinese characters:

CHINESE = re.compile(r'[\u4e00-\u9fff]+')

# ID card number (18 digits):

ID = r'^\d{17}[0-9Xx]$'

# Date:

DATE = r'^\d{4}-(0[1-9]|1[0-2])-(0[1-9]|[12]\d|3[01])$'

# Validation helper:

def is_email(s): return bool(re.match(EMAIL, s))

# Extract Chinese text:

CHINESE.findall('hello 世界你好')

# Colour:

COLOR = r'^#[0-9a-fA-F]{6}$'

Regex Performance

Precompile patterns, avoid catastrophic backtracking (nested quantifiers), and bound widths. Process big text in chunks.

import re

# Precompile:

pattern = re.compile(r'\d+') # at module level

pattern.findall(text)

# Avoid catastrophic backtracking:

# BAD: nested quantifiers

# re.match(r'(a+)+$', 'a' * 30 + '!') extremely slow

# GOOD: simplify it

re.match(r'a+$', 'aaa')

# Bound the width:

re.match(r'[a-z]{1,20}', 'x' * 100)

# Use a character class instead of alternation:

# BAD: [a|b|c]

# GOOD: [abc]

# Large text:

# process it in sections rather than all at once

# Prefer str methods for simple parsing:

# s.startswith / s.split

# Let regex handle structural matching

19.Build & Engineering

Dependency management, virtual environments, testing, linting, and CI.

requirements.txt

Pin versions with pip freeze -> requirements.txt. pip install -r installs. Lock for reproducibility.

# Export the current environment:

pip freeze > requirements.txt

# Example file content:

# requests==2.31.0

# httpx>=0.24.0

# # a comment explaining the purpose

# Install:

pip install -r requirements.txt

# Keep dev dependencies separate:

# requirements-dev.txt holds the test tooling

# Version pinning strategy:

# == pins exactly

# >= allows upgrades

# Pin exact versions for production

# Modern alternative:

# declare in pyproject.toml + a lock file

# uv installs extremely fast

venv in Practice

Best practices: one venv per project, never commit it, recreate on broken dependency trees.

# Create:

python3 -m venv .venv

# Activate (POSIX):

source .venv/bin/activate

# Activate (Windows):

# .venv\Scripts\activate

# Check the current interpreter:

which python

# Leave it: deactivate

# Delete and rebuild (when dependencies break):

# rm -rf .venv && python3 -m venv .venv

# Add .venv/ to .gitignore

# Install dependencies into the venv:

# never pip install into the system environment

# Point your editor at the venv interpreter

Modern Package Managers

uv for blazing-fast installs; poetry for dependency + packaging. Lock files make builds reproducible.

# uv (recommended, extremely fast):

# pip install uv

uv venv # create a virtual environment

uv add requests # add a dependency

uv sync # install everything in sync

# As a pip replacement:

uv pip install -r requirements.txt

# poetry:

# poetry new mypkg

# poetry add requests

# poetry install

# pyproject.toml + poetry.lock

# The consistent workflow:

# declare dependencies -> lock file -> CI installs from the lock

# The lock file gives everyone the same versions

# Upgrade: uv lock --upgrade

Testing with pytest

Name tests test_* and use assert. fixtures share setup, parametrize runs over inputs.

import pytest

from app import add

def test_add():

assert add(1, 2) == 3

def test_add_negative():

assert add(-1, 1) == 0

# Parametrised:

@pytest.mark.parametrize('a,b,expected', [

(1, 2, 3),

(0, 0, 0),

])

def test_add_param(a, b, expected):

assert add(a, b) == expected

# fixture:

def setup_db():

db = create_db()

yield db

db.close()

def test_using_fixture(setup_db):

assert setup_db.query()

# Asserting an exception:

with pytest.raises(ValueError):

int('x')

# Run: pytest or python3 -m pytest

# Coverage: pytest --cov

unittest Standard Library

The stdlib testing framework. Subclass TestCase, use assertXxx methods, override setUp/tearDown.

import unittest

from app import add

class TestAdd(unittest.TestCase):

def setUp(self):

self.data = [1, 2]

def test_add(self):

self.assertEqual(add(1, 2), 3)

def test_types(self):

self.assertIsInstance(add(1, 2), int)

def test_raises(self):

with self.assertRaises(TypeError):

add('a', 1)

if __name__ == '__main__':

unittest.main()

# Run:

# python3 -m unittest test_app

# Discovery mode:

# python3 -m unittest discover

# Assertion methods:

# assertEqual/assertTrue/assertIn

# assertAlmostEqual for floats

# mock: unittest.mock fakes dependencies

Lint & Formatting

ruff for fast lint + format, black for style, mypy for types. Pin the rules in pyproject.toml.

# ruff (recommended, extremely fast):

# pip install ruff

ruff check . # check

ruff format . # format

ruff check --fix . # fix automatically

# black:

# pip install black

black --line-length 88 .

# mypy type checking:

# pip install mypy

mypy src/

# Configure it in pyproject.toml:

# [tool.ruff]

# line-length = 88

# [tool.mypy]

# strict = true

# Editor integration:

# format on save

# Before committing: a pre-commit hook

Packaging & Publishing

Configure packaging in pyproject.toml, build sdist + wheel, publish to PyPI with twine.

# pyproject.toml:

# [project]

# name = "my-pkg"

# version = "0.1.0"

# description = "..."

# requires-python = ">=3.10"

# dependencies = ["requests"]

# [build-system]

# requires = ["setuptools>=68"]

# build-backend = "setuptools.build_meta"

# [tool.setuptools.packages.find]

# include = ["mypkg*"]

# Build:

# pip install build

# python3 -m build

# produces dist/*.whl and tar.gz

# Publish:

# pip install twine

# twine upload dist/*

# Test index: test.pypi.org

# Semantic versioning: 1.2.0

# Try the install before publishing: pip install .

CI & Deployment

Pipeline stages: lint, type-check, test, build. Configure in GitHub Actions.

# GitHub Actions:

# .github/workflows/ci.yml

# name: CI

# on: [push, pull_request]

# jobs:

# test:

# runs-on: ubuntu-latest

# steps:

# - uses: actions/checkout@v4

# - uses: actions/setup-python@v5

# with: python-version: '3.11'

# - run: pip install -e .[dev]

# - run: ruff check .

# - run: mypy src/

# - run: pytest --cov

# Cache dependencies: actions/cache

# Multi-version matrix: python-version: [3.10, 3.11, 3.12]

# A failure blocks the merge

# Deployment: upload to PyPI on release

# Environment variables: keep secrets in GitHub Secrets

Official Links

Direct links to the official docs and resources.

Trademark & Use Notice

These programming language cheatsheets are open community references. The names, logos, and trademarks of the programming languages and runtimes referenced here (such as Java™, C#, Swift™, Kotlin™, Dart™, TypeScript™, Go™, Rust™, Python™, JavaScript™, Perl™, PHP, R, Scratch™, C, C++) are the property of their respective owners and are used here solely for identification and reference purposes.

All content in these cheatsheets is independently authored as a simplified, self-contained reference. It is not affiliated with, endorsed by, or sourced from any official documentation. Code examples, syntax references, and descriptions are original simplifications; for authoritative information please consult the official documentation of each language or runtime.

Where applicable, language version tags and badges (e.g. "C++17 / C++20", "C# 12", "Swift 5.9") reflect specific stable releases; behavior may differ in other versions. The choice of which version to document does not imply any preference or recommendation by the trademark holders.

If a trademark or copyright owner believes that any content on this site infringes their rights, please send an email to [email protected]. Upon receiving valid proof of rights, we will modify or remove the relevant content within a reasonable period.

This website assumes no legal liability for any direct or indirect losses arising from the use of information presented on these pages.

Version 2.1.1

Python Cheat Sheet

Recommended Learning Path

1.Hello World & Build Environment

Minimal Program

Ways to Run

Virtual Environments

pip Package Management

pyproject.toml

Shebang Scripts

REPL Interaction

__main__ & Modules

2.Variables & Type Annotations

Assignment & Inference

Type Annotations

Multiple Assignment

Constant Conventions

LEGB Scope

Walrus Operator

None & Booleans

del & Garbage Collection

3.Data Types

Numeric Types

Booleans & None

Sequence Types

Mapping dict

Set Types

Bytes Types

Advanced Type Annotations

Duck Typing

4.References & Mutability

Reference Semantics

Mutable vs Immutable

Shallow Copy

Deep Copy

Shared Reference Pitfalls

is vs ==

Default Argument Pitfall

Interning & Caching

5.Control Flow

if / elif / else

Pattern Matching

for Loop

while Loop

break & continue

Comprehensions

Truthiness

Iteration Idioms

6.Functions

Function Definition

Parameter Types

*args & **kwargs

Lambda

Closures

Decorators

Generators

Functional Tools

Recursion

7.Strings

String Basics

Common Methods

f-string Formatting

format & % Formatting

Split & Join

Strip & Pad

Unicode & Encoding

Search & Replace

Text Wrapping

8.Collections & Containers

List Operations

Dict Operations

Set Operations

Tuple Usage

collections Utilities

Sorting

Slicing

Heaps & Queues

9.Memory & Performance

Garbage Collection

memoryview & Zero-Copy

Weak References

main & Modules

slots for Memory