Skip to content

Latest commit

 

History

History
374 lines (279 loc) · 7.11 KB

Python.jupyter.md

File metadata and controls

374 lines (279 loc) · 7.11 KB
Raw

python

常用

Dict and Easydict

get 和 [] 获取 key 的区别

result = {'comment1': 'OK'}

# 此时都返回 'OK'
result.get('comment1') == result['comment1']
# 返回 None
print(result.get('comment2'))
# 程序出错
try:
    result['comment2']
except Exception as E:
    print("Exception: {}".format(type(E).__name__))
    print("Exception message: {}".format(E))

Easydict

from easydict import EasyDict as edict

__C = edict()
cfg = __C

__C.DATASET                             = edict()
__C.DATASET.NAME                        = 'GoPro'

索引

b[start:end:step]

b = (1, 3, 2, 4)

print('0:', b[0:4:1])   # 从0开始，到4之前结束，递进1
print('2:', b[0:-1:2])   # 从0开始，到最后一个结束，递进2

Random

import random

random.seed(0)
# 随机生成[-10, 10] 之间的浮点数
random.uniform(-10, 10)
# Return random integer in range [a, b], including both end points.
random.randint(-10, 10)

namedtuple

from collections import namedtuple
gaussian = namedtuple('Gaussian', ['mean', 'var'])
gaussian.__repr__ = lambda s: 'Ν(μ={:.3f}, σ={:.3f})'.format(s[0], s[1])

g = gaussian(0, 1)
g, g.mean, g.var

CSV 和 namedtuple 结合

import csv

EmployeeRecord = namedtuple('EmployeeRecord', 'name, age')

csv_data = open("_files/employee.csv", "r")

for emp in map(EmployeeRecord._make, csv.reader(csv_data)):
    print(emp.name, emp.age)

实用技巧

eval()

eval()可以将输入的字符串转化为Python代码

eval('print("hello")')

有时可以配合input()使用，如eval(intput())

import numpy as np

eval('1+np.pi')

f直接打印变量

a = 'hello'
b = 2
print(f'{a}, {b}')

Deep Copy

import copy

A = [[1], [2]]

B = copy.copy(A)
C = copy.deepcopy(A)

A[0][0] = ['love']

print('Shallow copy:', B, "\nDeep copy:", C)

Print out nicely - pprint

尤其对字典和json文件有用

import pprint

students = {"Dilip": ["English", "Maths", "Science"],
            "Raju": {"English": 50, "Maths": 60, "Science": 70},
            "Kalpana": (50, 60, 70)}

pprint.pprint(students)

pp = pprint.PrettyPrinter(width=20)
pp.pprint(students)

获取 for 循环的当前步数

Q: Accessing the index in 'for' loops? Q: Using a for loop, how do I access the loop index?

for count, item in enumerate(range(4), start=10):
    print(count, item)

自定义排序sorted

students = [['john', 'A', 15],
            ['jane', 'B', 12],
            ['dave', 'B', 10]]

sorted(students, key=lambda student: student[2])

循环移动一维 list/np.ndarray

We can see that we correctly shifted all the values one position to the right, wrapping up from the end of the array back to the begining.

def shift_1d(src, move):
    """ move the position by `move` spaces, where positive is 
    to the right, and negative is to the left
    """
    n = len(src)
    if isinstance(src, list):
        dst = [0] * n
    elif isinstance(src, np.ndarray):
        dst = np.zeros(n)
    else:
        raise Exception

    # 不要用for循环，太耗时
    if move > 0:
        dst = src[n-move%n:] + src[:move%n+1]
    else:
        dst = src[abs(move)%n:] + src[:abs(move)%n]
    return dst

shift_1d([1,2,3,4, 5], 2), shift_1d([1,2,3,4, 5], -4)

with Statement

contextlib

自动执行 __enter__ 和 __exit___，可以用 contextmanager decorator 来创造。

例如自动管理文件打开和关闭

from contextlib import contextmanager

@contextmanager
def open_file(path, mode):
    """
        with open_file() as f:
            ...
    """
    the_file = open(path, mode)
    yield the_fil
    the_file.close()

自动加上html的tag

@contextmanager
def tag(name):
    print("<%s>" % name)
    yield
    print("</%s>" % name)

with tag("h1"):
    print("foo")

with 和 time Benchmark

构造一个Benchmark类，配合with测量代码运行时间

import time

class Benchmark():
    def __init__(self, prefix=None):
        self.prefix = prefix + ' ' if prefix else ''
    
    def __enter__(self):
        self.start = time.time()
        
    def __exit__(self, *args):
        print('%stime: %.4f sec' % (self.prefix, time.time() - self.start))
with Benchmark('Workloads are queued.'):
    print('+++')

with Benchmark('Workloads are finished.'):
    print('---')

保存变量

如果变量非常大，想重复运算，可以保存加快运行速度

import pickle

obj1 = np.eye(2)
obj2 = [1]
fn = 'test.pkl'

# Saving the objects
with open(fn, 'wb') as f:
    pickle.dump([obj1, obj2], f)
    
# Getting back the objects:
with open(fn, 'rb') as f:
    obj1, obj2 = pickle.load(f)

System

import os
import subprocess

os.path

os.path.basename("/home/austin/name.md.txt")
# Getting the name of the file without the extension
os.path.splitext("/home/austin/name.md.txt")[0]
# Get the extension
os.path.splitext("/home/austin/name.md.txt")[1]
os.path.splitext(os.path.basename("/home/austin/name.md"))[0]
os.path.dirname("/home/austin/name.md")

subprocess

得到系统内存 get_mem

# Only for Linux or Mac
def get_mem():
    res = subprocess.check_output(['ps', 'u', '-p', str(os.getpid())])
    return int(str(res).split()[15]) / 1e3

get_mem()

import相对路径文件

DirA
|--main.py
|--DirB
|----fileB.py
|----fileB1.py

在"main.py"中可以直接import DirB.fileB，但是同级/上级调用如在"fileB.py"里import fileB1要分几种情况：

  • python main.py中调用了import DirB.fileB然后间接调用了import fileB1
  • python DirB/fileB.py中调用了import fileB1
  • python fileB.py中调用了import fileB1

鉴于情况比较复杂，尽量避免同级/上级调用，如需要运行python fileB.py来执行if __name__ == "__main__"的入口测试可以先设置以下python绝对或者相对环境：

export PYTHONPATH="/home/austin/.../DirA/DirB"

argparse

import argparse

parser = argparse.ArgumentParser(description='Process some integers.')
parser.add_argument('integers', metavar='N', type=int, nargs='+',
                    help='an integer for the accumulator')
parser.add_argument('--sum', dest='accumulate', action='store_const',
                    const=sum, default=max,
                    help='sum the integers (default: find the max)')

args = parser.parse_args()
print(args.accumulate(args.integers))

ipython

display

如果需要显示np.array，请查看sympy.jupyter.md

import numpy as np
from  IPython.display import display, Math, Latex
raw_latex = "\\text{%s} \quad W*X^T" % ("e.g.")

raw_latex
display(Math(raw_latex))
display(Latex("$"+raw_latex+"$"))