Programing Python Tools PySnooper

[junxnone]

PySnooper

• Debug Python Code: 使用 装饰器 实现的调试器，可以打印出每行执行的结果,

pip install pysnooper

import pysnooper
@pysnooper.snoop('debug_xxx.log', color=False)

Source path:... /path/to/your/scripts/tap.py
Starting var:.. gt_sp = [0, 0, 0, 0, 1, 1, 1, 1, 1, 1]
Starting var:.. pr_sp = [0.9919, 0.9923, 0.9942, 0.992, 0.9916, 0.9913, 0.9926, 0.9951, 0.9894, 0.9905]
14:52:56.296655 call        14 def test(gt_sp, pr_sp):
14:52:56.296869 line        15     thres = 0.9979
New var:....... thres = 0.9979
14:52:56.296932 line        17     pr_l = [ 1 if num > thres else 0 for num in pr_sp]
New var:....... pr_l = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
14:52:56.297000 line        18     print(pr_l)
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
14:52:56.297071 line        20     aas = accuracy_score(gt_sp, pr_l)
New var:....... aas = 0.4
14:52:56.298280 line        22     print(f'accuracy_score：{aas}')
accuracy_score：0.4
14:52:56.298393 line        25     ap = average_precision_score(gt_sp, pr_sp)
New var:....... ap = 0.6251322751322752
14:52:56.299719 line        26     print(f'ap: {ap}')
ap: 0.6251322751322752
14:52:56.299832 line        28     precision, recall, thresholds = precision_recall_curve(gt_sp, pr_sp)
New var:....... precision = array([0.6       , 0.55555556, 0.5       , 0.428...66667, 0.5       , 1.        ,       1.        ])
New var:....... recall = array([1.        , 0.83333333, 0.66666667, 0.5  ...33333, 0.16666667, 0.16666667,       0.        ])
New var:....... thresholds = array([0.9894, 0.9905, 0.9913, 0.9916, 0.9919, 0.992 , 0.9923, 0.9926,       0.9942, 0.9951])
14:52:56.300593 return      28     precision, recall, thresholds = precision_recall_curve(gt_sp, pr_sp)
Return value:.. None
Elapsed time: 00:00:00.004629

Use Case

调试函数

import pysnooper

@pysnooper.snoop()
def number_to_bits(number):
    if number:
        bits = []
        while number:
        ....

调试某段代码

import pysnooper
import random

def foo():
    lst = []
    for i in range(10):
        lst.append(random.randrange(1, 1000))

    with pysnooper.snoop():
        lower = min(lst)
        upper = max(lst)
        mid = (lower + upper) / 2
        print(lower, mid, upper)

写入 log 文件

@pysnooper.snoop('/my/log/file.log')

默认是终端带颜色的字体，在编辑器中会有乱码，可以使用 color=False 禁止带颜色

展开列表/词典

@pysnooper.snoop(watch_explode=('foo', 'self'))

使用环境变量 Disable PySnooper

是不是可以用来打印 卷积网络图？？

export PYSNOOPER_DISABLED=1

自定义输出方式

def large(l):
    return isinstance(l, list) and len(l) > 5

def print_list_size(l):
    return 'list(size={})'.format(len(l))

def print_ndarray(a):
    return 'ndarray(shape={}, dtype={})'.format(a.shape, a.dtype)

@pysnooper.snoop(custom_repr=((large, print_list_size), (numpy.ndarray, print_ndarray)))
def sum_to_x(x):
    l = list(range(x))
    a = numpy.zeros((10,10))
    return sum(l)

sum_to_x(10000)

默认打印长度

• 默认打印长度为 100，变量超过此值被省略

    @pysnooper.snoop(max_variable_length=200)

max_variable_length=None 不限制输出

Reference

• PySnooper
• Advanced Usage