Python 2.7 即將停止支援，請收下這份 3.x 遷移指南

Python 2 總是試圖使用字串級聯（準確，但不好），現在有了 pathlib，程式碼安全、準確、可讀性強。

此外，pathlib.Path 具備大量方法，這樣 Python 新使用者就不用每個方法都去搜索了：

p.exists()
p.is_dir()
p.parts()
p.with_name('sibling.png') # only change the name, but keep the folder
p.with_suffix('.jpg') # only change the extension, but keep the folder and the name
p.chmod(mode)
p.rmdir()

pathlib 會節約大量時間，詳見：

• 檔案：https://docs.python.org/3/library/pathlib.html；

• 參考資訊：https://pymotw.com/3/pathlib/。

型別提示（Type hinting）成為語言的一部分

PyCharm 中的型別提示示例：

Python 不只是適合指令碼的語言，現在的資料流程還包括大量步驟，每一步都包括不同的框架（有時也包括不同的邏輯）。

型別提示被引入 Python，以幫助處理越來越複雜的專案，使機器可以更好地進行程式碼驗證。而之前需要不同的模組使用自定義方式在檔案字串中指定型別（註意：PyCharm 可以將舊的檔案字串轉換成新的型別提示）。

下列程式碼是一個簡單示例，可以處理不同型別的資料（這就是我們喜歡 Python 資料棧之處）。

def repeat_each_entry(data):
    """ Each entry in the data is doubled 
    
    """
    index = numpy.repeat(numpy.arange(len(data)), 2)
    return data[index]

上述程式碼適用於 numpy.array（包括多維）、astropy.Table 和 astropy.Column、bcolz、cupy、mxnet.ndarray 等。

該程式碼同樣可用於 pandas.Series，但是方式是錯誤的：

repeat_each_entry(pandas.Series(data=[0, 1, 2], index=[3, 4, 5])) # returns Series with Nones inside

這是一個兩行程式碼。想象一下複雜系統的行為多麼難預測，有時一個函式就可能導致錯誤的行為。明確瞭解哪些型別方法適合大型系統很有幫助，它會在函式未得到此類引數時給出提醒。

def repeat_each_entry(data: Union[numpy.ndarray, bcolz.carray]):

如果你有一個很棒的程式碼庫，型別提示工具如 MyPy 可能成為整合流程中的一部分。不幸的是，提示沒有強大到足以為 ndarrays/tensors 提供細粒度型別，但是或許我們很快就可以擁有這樣的提示工具了，這將是 DS 的偉大功能。

型別提示 → 執行時的型別檢查

預設情況下，函式註釋不會影響程式碼的執行，不過它也只能幫你指出程式碼的意圖。

但是，你可以在執行時中使用 enforce 等工具強制進行型別檢查，這可以幫助你除錯程式碼（很多情況下型別提示不起作用）。

@enforce.runtime_validation
def foo(text: str) -> None:
    print(text)
foo('Hi') # ok
foo(5)    # fails
@enforce.runtime_validation
def any2(x: List[bool]) -> bool:
    return any(x)
any ([False, False, True, False]) # True
any2([False, False, True, False]) # True
any (['False']) # True
any2(['False']) # fails
any ([False, None, "", 0]) # False
any2([False, None, "", 0]) # fails

# Python 3
from astropy import units as u
@u.quantity_input()
def frequency(speed: u.meter / u.s, wavelength: u.m) -> u.terahertz:
    return speed / wavelength
frequency(speed=300_000 * u.km / u.s, wavelength=555 * u.nm)
# output: 540.5405405405404 THz, frequency of green visible light

# l2-regularized linear regression: || AX - b ||^2 + alpha * ||x||^2 -> min
# Python 2
X = np.linalg.inv(np.dot(A.T, A) + alpha * np.eye(A.shape[1])).dot(A.T.dot(b))
# Python 3
X = np.linalg.inv(A.T @ A + alpha * np.eye(A.shape[1])) @ (A.T @ b)

import glob
# Python 2
found_images = 
    glob.glob('/path/*.jpg') 
  + glob.glob('/path/*/*.jpg') 
  + glob.glob('/path/*/*/*.jpg') 
  + glob.glob('/path/*/*/*/*.jpg') 
  + glob.glob('/path/*/*/*/*/*.jpg') 
# Python 3
found_images = glob.glob('/path/**/*.jpg', recursive=True)

# Python 3
found_images = pathlib.Path('/path/').glob('**/*.jpg')

print >>sys.stderr, "critical error"      # Python 2
print("critical error", file=sys.stderr)  # Python 3

# Python 3
print(*array, sep='	')
print(batch, epoch, loss, accuracy, time, sep='	')

# Python 3
_print = print # store the original print function
def print(*args, **kargs):
    pass  # do something useful, e.g. store output to some file

@contextlib.contextmanager
def replace_print():
    import builtins
    _print = print # saving old print function
    # or use some other function here
    builtins.print = lambda *args, **kwargs: _print('new printing', *args, **kwargs)
    yield
    builtins.print = _print
with replace_print():
    <code here will invoke other print function>

# Python 3
result = process(x) if is_valid(x) else print('invalid item: ', x)

# Python 2
print('{batch:3} {epoch:3} / {total_epochs:3}  accuracy: {acc_mean:0.4f}±{acc_std:0.4f} time: {avg_time:3.2f}'.format(
    batch=batch, epoch=epoch, total_epochs=total_epochs, 
    acc_mean=numpy.mean(accuracies), acc_std=numpy.std(accuracies),
    avg_time=time / len(data_batch)
))
# Python 2 (too error-prone during fast modifications, please avoid):
print('{:3} {:3} / {:3}  accuracy: {:0.4f}±{:0.4f} time: {:3.2f}'.format(
    batch, epoch, total_epochs, numpy.mean(accuracies), numpy.std(accuracies),
    time / len(data_batch)
))

120  12 / 300  accuracy: 0.8180±0.4649 time: 56.60

# Python 3.6+
print(f'{batch:3} {epoch:3} / {total_epochs:3}  accuracy: {numpy.mean(accuracies):0.4f}±{numpy.std(accuracies):0.4f} time: {time / len(data_batch):3.2f}')

query = f"INSERT INTO STATION VALUES (13, '{city}', '{state}', {latitude}, {longitude})"

data = pandas.read_csv('timing.csv')
velocity = data['distance'] / data['time']

n_gifts = money // gift_price  # correct for int and float arguments

# All these comparisons are illegal in Python 3
3 < '3'
2 < None
(3, 4) < (3, None)
(4, 5) < [4, 5]
# False in both Python 2 and Python 3
(4, 5) == [4, 5]

sorted([2, '1', 3])  # invalid for Python 3, in Python 2 returns [2, 3, '1']

if a is not None:
  pass
if a: # WRONG check for None
  pass

s = '您好'
print(len(s))
print(s[:2])

x = u'со'
x += 'co' # ok
x += 'со' # fail

'a' < type < u'a'  # Python 2: True
'a' < u'a'         # Python 2: False

from collections import Counter
Counter('Möbelstück')

import json
x = {str(i):i for i in range(5)}
json.loads(json.dumps(x))
# Python 2
{u'1': 1, u'0': 0, u'3': 3, u'2': 2, u'4': 4}
# Python 3
{'0': 0, '1': 1, '2': 2, '3': 3, '4': 4}

from torch import nn
# Python 2
model = nn.Sequential(OrderedDict([
          ('conv1', nn.Conv2d(1,20,5)),
          ('relu1', nn.ReLU()),
          ('conv2', nn.Conv2d(20,64,5)),
          ('relu2', nn.ReLU())
        ]))
# Python 3.6+, how it *can* be done, not supported right now in pytorch
model = nn.Sequential(
    conv1=nn.Conv2d(1,20,5),
    relu1=nn.ReLU(),
    conv2=nn.Conv2d(20,64,5),
    relu2=nn.ReLU())
)        

# handy when amount of additional stored info may vary between experiments, but the same code can be used in all cases
model_paramteres, optimizer_parameters, *other_params = load(checkpoint_name)
# picking two last values from a sequence
*prev, next_to_last, last = values_history
# This also works with any iterables, so if you have a function that yields e.g. qualities,
# below is a simple way to take only last two values from a list 
*prev, next_to_last, last = iter_train(args)

# Python 2
import cPickle as pickle
import numpy
print len(pickle.dumps(numpy.random.normal(size=[1000, 1000])))
# result: 23691675
# Python 3
import pickle
import numpy
len(pickle.dumps(numpy.random.normal(size=[1000, 1000])))
# result: 8000162

labels = <initial_value>
predictions = [model.predict(data) for data, labels in dataset]
# labels are overwritten in Python 2
# labels are not affected by comprehension in Python 3

# Python 2
class MySubClass(MySuperClass):
    def __init__(self, name, **options):
        super(MySubClass, self).__init__(name='subclass', **options)
# Python 3
class MySubClass(MySuperClass):
    def __init__(self, name, **options):
        super().__init__(name='subclass', **options)

x = dict(a=1, b=2)
y = dict(b=3, d=4)
# Python 3.5+
z = {**x, **y} 
# z = {'a': 1, 'b': 3, 'd': 4}, note that value for `b` is taken from the latter dict.

[*a, *b, *c] # list, concatenating 
(*a, *b, *c) # tuple, concatenating 
{*a, *b, *c} # set, union 

Python 3.5+
do_something(**{**default_settings, **custom_settings})
# Also possible, this code also checks there is no intersection between keys of dictionaries
do_something(**first_args, **second_args)

model = sklearn.svm.SVC(2, 'poly', 2, 4, 0.5)

class SVC(BaseSVC):
    def __init__(self, *, C=1.0, kernel='rbf', degree=3, gamma='auto', coef0=0.0, ... )

# Python 3
math.inf # 'largest' number
math.nan # not a number
max_quality = -math.inf  # no more magic initial values!
for model in trained_models:
    max_quality = max(max_quality, compute_quality(model, data))

isinstance(x, numbers.Integral) # Python 2, the canonical way
isinstance(x, (long, int))      # Python 2
isinstance(x, int)              # Python 3, easier to remember

map(lambda x, (y, z): x, z, dict.items())

{x:z for x, (y, z) in d.items()}