用Python創(chuàng)建你自己的Shell
介紹
很多人討厭bash腳本。每當(dāng)我要做最簡單的事情時,我都必須查閱文檔。如何將函數(shù)的參數(shù)轉(zhuǎn)發(fā)給子命令?如何將字符串分配給變量,然后作為命令調(diào)用該字符串?如何檢查兩個字符串變量是否相等?如何分割字符串并獲得后半部分?等等。不是我找不到這些答案,而是每次都必須查找它們。
但是,我們不能否認(rèn)將整個程序當(dāng)作純粹的功能發(fā)揮作用的能力,以及將一個程序的輸出傳遞到另一個程序的自然程度。因此,我想知道,我們能否將bash的某些功能與Python結(jié)合起來?
基礎(chǔ)知識
讓我們從一個類開始。這是一個簡單的方法,將其初始化參數(shù)保存到局部變量,然后使用subprocess.run對其自身進(jìn)行延遲求值并保存結(jié)果。
- import subprocess
- class PipePy:
- def __init__(self, *args):
- self._args = args
- self._result = None
- def _evaluate(self):
- if self._result is not None:
- return
- self._result = subprocess.run(self._args,
- capture_output=True,
- text=True)
- @property
- def returncode(self):
- self._evaluate()
- return self._result.returncode
- @property
- def stdout(self):
- self._evaluate()
- return self._result.stdout
- def __str__(self):
- return self.stdout
- @property
- def stderr(self):
- self._evaluate()
- return self._result.stderr
我們讓它旋轉(zhuǎn)一下:
- ls = PipePy('ls')
- ls_l = PipePy('ls', '-l')
- print(ls)
- # <<< files.txt
- # ... main.py
- # ... tags
- print(ls_l)
- # <<< total 16
- # ... -rw-r--r-- 1 kbairak kbairak 125 Jan 22 08:53 files.txt
- # ... -rw-r--r-- 1 kbairak kbairak 5425 Feb 1 21:54 main.py
- # ... -rw-r--r-- 1 kbairak kbairak 1838 Feb 1 21:54 tags
使其看起來更像“命令式”
不用每次我們要自定義命令時都去調(diào)用PipePy。
- ls_l = PipePy('ls', '-l')
- print(ls_l)
相當(dāng)于
- ls = PipePy('ls')
- print(ls('-l'))
換句話說,我們要使:
- PipePy('ls', '-l')
相當(dāng)于
- PipePy('ls')('-l')
值得慶幸的是,我們的類創(chuàng)建了惰性對象這一事實在很大程度上幫助了我們:
- class PipePy:
- # __init__, etc
- def __call__(self, *args):
- args = self._args + args
- return self.__class__(*args)
- ls = PipePy('ls')
- print(ls('-l'))
- # <<< total 16
- # ... -rw-r--r-- 1 kbairak kbairak 125 Jan 22 08:53 files.txt
- # ... -rw-r--r-- 1 kbairak kbairak 5425 Feb 1 21:54 main.py
- # ... -rw-r--r-- 1 kbairak kbairak 1838 Feb 1 21:54 tags
關(guān)鍵字參數(shù)
如果要向ls傳遞更多參數(shù),則可能會遇到--sort = size。我們可以輕松地執(zhí)行l(wèi)s('-l','--sort = size')。我們可以做得更好嗎?
- class PipePy:
- - def __init__(self, *args):
- + def __init__(self, *args, **kwargs):
- self._args = args
- + self._kwargs = kwargs
- self._result = None
- def _evaluate(self):
- if self._result is not None:
- return
- - self._result = subprocess.run(self._args,
- + self._result = subprocess.run(self._convert_args(),
- capture_output=True,
- text=True)
- + def _convert_args(self):
- + args = [str(arg) for arg in self._args]
- + for key, value in self._kwargs.items():
- + keykey = key.replace('_', '-')
- + args.append(f"--{key}={value}")
- + return args
- - def __call__(self, *args):
- + def __call__(self, *args, **kwargs):
- args = self._args + args
- + kwargs = {**self._kwargs, **kwargs}
- - return self.__class__(*args)
- + return self.__class__(*args, **kwargs)
- # returncode, etc
讓我們來旋轉(zhuǎn)一下:
- print(ls('-l'))
- # <<< total 16
- # ... -rw-r--r-- 1 kbairak kbairak 125 Jan 22 08:53 files.txt
- # ... -rw-r--r-- 1 kbairak kbairak 5425 Feb 1 21:54 main.py
- # ... -rw-r--r-- 1 kbairak kbairak 1838 Feb 1 21:54 tags
- print(ls('-l', sort="size"))
- # <<< total 16
- # ... -rw-r--r-- 1 kbairak kbairak 5425 Feb 1 21:54 main.py
- # ... -rw-r--r-- 1 kbairak kbairak 1838 Feb 1 21:54 tags
- # ... -rw-r--r-- 1 kbairak kbairak 125 Jan 22 08:53 files.txt
Piping
事情開始變得有趣起來。我們的最終目標(biāo)是能夠做到:
- ls = PipePy('ls')
- grep = PipePy('grep')
- print(ls | grep('tags'))
- # <<< tags
我們的過程是:
1、讓__init__和__call__方法接受一個僅用于關(guān)鍵字的新_pipe_input關(guān)鍵字參數(shù),該參數(shù)將保存在self上。
2、在評估期間,如果設(shè)置了_pipe_input,它將作為輸入?yún)?shù)傳遞給subprocess.run。
3、重寫__or__方法以將左操作數(shù)的結(jié)果作為pipe輸入傳遞給右操作數(shù)。
- class PipePy:
- - def __init__(self, *args, **kwargs):
- + def __init__(self, *args, _pipe_input=None, **kwargs):
- self._args = args
- self._kwargs = kwargs
- + self._pipe_input = _pipe_input
- self._result = None
- - def __call__(self, *args, **kwargs):
- + def __call__(self, *args, _pipe_input=None, **kwargs):
- args = self._args + args
- kwargs = {**self._kwargs, **kwargs}
- - return self.__class__(*args, **kwargs)
- + return self.__class__(*args, _pipe_input_pipe_input=_pipe_input, **kwargs)
- def _evaluate(self):
- if self._result is not None:
- return
- self._result = subprocess.run(self._convert_args(),
- + input=self._pipe_input,
- capture_output=True,
- text=True)
- + def __or__(left, right):
- + return right(_pipe_input=left.stdout)
讓我們嘗試一下(從之前稍微修改命令以證明它確實有效):
- ls = PipePy('ls')
- grep = PipePy('grep')
- print(ls('-l') | grep('tags'))
- # <<< -rw-r--r-- 1 kbairak kbairak 1838 Feb 1 21:54 tags
讓我們添加一些簡單的東西
1、真實性:
- class PipePy:
- # __init__, etc
- def __bool__(self):
- return self.returncode == 0
現(xiàn)在我們可以作出如下處理:
- git = PipePy('git')
- grep = PipePy('grep')
- if git('branch') | grep('my_feature'):
- print("Branch 'my_feature' found")
2、讀取/寫入文件:
- class PipePy:
- # __init__, etc
- def __gt__(self, filename):
- with open(filename, 'w') as f:
- f.write(self.stdout)
- def __rshift__(self, filename):
- with open(filename, 'a') as f:
- f.write(self.stdout)
- def __lt__(self, filename):
- with open(filename) as f:
- return self(_pipe_input=f.read())
現(xiàn)在可以作出如下操作:
- ls = PipePy('ls')
- grep = PipePy('grep')
- cat = PipePy('cat')
- ls > 'files.txt'
- print(grep('main') < 'files.txt')
- # <<< main.py
- ls >> 'files.txt'
- print(cat('files.txt'))
- # <<< files.txt
- # ... main.py
- # ... tags
- # ... files.txt
- # ... main.py
- # ... tags
3、迭代
- class PipePy:
- # __init__, etc
- def __iter__(self):
- return iter(self.stdout.split())
現(xiàn)在可以作出如下操作:
- ls = PipePy('ls')
- for name in ls:
- print(name.upper())
- # <<< FILES.TXT
- # ... MAIN.PY
- # ... TAGS
4、表格:
- class PipePy:
- # __init__, etc
- def as_table(self):
- lines = self.stdout.splitlines()
- fields = lines[0].split()
- result = []
- for line in lines[1:]:
- item = {}
- for i, value in enumerate(line.split(maxsplit=len(fields) - 1)):
- item[fields[i]] = value
- result.append(item)
- return result
現(xiàn)在可以作出下面操作:
- ps = PipePy('ps')
- print(ps)
- # <<< PID TTY TIME CMD
- # ... 4205 pts/4 00:00:00 zsh
- # ... 13592 pts/4 00:00:22 ptipython
- # ... 16253 pts/4 00:00:00 ps
- ps.as_table()
- # <<< [{'PID': '4205', 'TTY': 'pts/4', 'TIME': '00:00:00', 'CMD': 'zsh'},
- # ... {'PID': '13592', 'TTY': 'pts/4', 'TIME': '00:00:22', 'CMD': 'ptipython'},
- # ... {'PID': '16208', 'TTY': 'pts/4', 'TIME': '00:00:00', 'CMD': 'ps'}]
5、普通bash實用程序:
在子進(jìn)程中更改工作目錄不會影響當(dāng)前的腳本或python shell。與更改環(huán)境變量相同,以下內(nèi)容不是PipePy的補(bǔ)充,但很不錯:
- import os
- cd = os.chdir
- export = os.environ.__setitem__
- pwd = PipePy('pwd')
- pwd
- # <<< /home/kbairak/prog/python/pipepy
- cd('..')
- pwd
- # <<< /home/kbairak/prog/python
使事情看起來更shell-like
如果我在交互式shell中,則希望能夠簡單地鍵入ls并完成它。
- class PipePy:
- # __init__, etc
- def __repr__(self):
- return self.stdout + self.stderr
交互式shell
- >>> ls = PipePy('ls')
- >>> ls
- files.txt
- main.py
- tags
我們的實例是惰性的,這意味著如果我們對它們的結(jié)果感興趣,則將對它們進(jìn)行評估,此后不再進(jìn)行評估。如果我們只是想確保已執(zhí)行該操作怎么辦?例如,假設(shè)我們有以下腳本:
- from pipepy import PipePy
- tar = PipePy('tar')
- tar('-xf', 'some_archive.tar')
- print("File extracted")
該腳本實際上不會執(zhí)行任何操作,因為tar調(diào)用實際上并未得到評估。我認(rèn)為一個不錯的慣例是,如果不帶參數(shù)調(diào)用__call__強(qiáng)制求值:
- class PipePy:
- def __call__(self, *args, _pipe_input=None, **kwargs):
- args = self._args + args
- kwargs = {**self._kwargs, **kwargs}
- - return self.__class__(*args, _pipe_input_pipe_input=_pipe_input, **kwargs)
- + result = self.__class__(*args, _pipe_input_pipe_input=_pipe_input, **kwargs)
- + if not args and not _pipe_input and not kwargs:
- + result._evaluate()
- + return result
因此在編寫腳本時,如果要確保實際上已調(diào)用命令,則必須用一對括號來調(diào)用它:
- from pipepy import PipePy
- tar = PipePy('tar')
- -tar('-xf', 'some_archive.tar')
- +tar('-xf', 'some_archive.tar')()
- print("File extracted")
但是,我們還沒有解決問題??紤]一下:
- date = PipePy('date')
- date
- # <<< Mon Feb 1 10:43:08 PM EET 2021
- # Wait 5 seconds
- date
- # <<< Mon Feb 1 10:43:08 PM EET 2021
不好!date沒有改變。date對象將其_result保留在內(nèi)存中。隨后的評估實際上不會調(diào)用該命令,而只是返回存儲的值。
一種解決方案是通過使用空括號來強(qiáng)制創(chuàng)建副本:
- date = PipePy('date')
- date()
- # <<< Mon Feb 1 10:45:09 PM EET 2021
- # Wait 5 seconds
- date()
- # <<< Mon Feb 1 10:45:14 PM EET 2021
另一個解決方案是:由PipePy構(gòu)造函數(shù)返回的實例不應(yīng)該是惰性的,但由__call__調(diào)用返回的實例將是惰性的。
- class PipePy:
- - def __init__(self, *args, _pipe_input=None, **kwargs):
- + def __init__(self, *args, _pipe_input=None, _lazy=False, **kwargs):
- self._args = args
- self._kwargs = kwargs
- self._pipe_input = _pipe_input
- + self._lazy = _lazy
- self._result = None
- def __call__(self, *args, _pipe_input=None, **kwargs):
- args = self._args + args
- kwargs = {**self._kwargs, **kwargs}
- - result = self.__class__(*args, _pipe_input_pipe_input=_pipe_input, **kwargs)
- + result = self.__class__(*args,
- + _pipe_input_pipe_input=_pipe_input,
- + _lazy=True,
- + **kwargs)
- if not args and not _pipe_input and not kwargs:
- result._evaluate()
- return result
- def _evaluate(self):
- - if self._result is not None:
- + if self._result is not None and self._lazy:
- return
- self._result = subprocess.run(self._convert_args(),
- input=self._pipe_input,
- capture_output=True,
- text=True)
旋轉(zhuǎn)一下:
- date = PipePy('date')
- date
- # <<< Mon Feb 1 10:54:09 PM EET 2021
- # Wait 5 seconds
- date
- # <<< Mon Feb 1 10:54:14 PM EET 2021
并且可以預(yù)見的是,使用空調(diào)用的返回值將具有之前的行為:
- date = PipePy('date')
- d = date()
- d
- # <<< Mon Feb 1 10:56:21 PM EET 2021
- # Wait 5 seconds
- d
- # <<< Mon Feb 1 10:56:21 PM EET 2021
沒關(guān)系 您不會期望d會更新其值。
越來越危險
好吧,ls('-l')不錯,但是如果我們像人類一樣簡單地做ls -l,那就太好了。嗯,我有個主意:
- class PipePy:
- # __init__, etc
- def __sub__(left, right):
- return left(f"-{right}")
現(xiàn)在可以作如下操作:
- ls = PipePy('ls')
- ls - 'l'
- # <<< total 16
- # ... -rw-r--r-- 1 kbairak kbairak 46 Feb 1 23:04 files.txt
- # ... -rw-r--r-- 1 kbairak kbairak 5425 Feb 1 21:54 main.py
- # ... -rw-r--r-- 1 kbairak kbairak 1838 Feb 1 21:54 tags
我們還有一步:
- l = 'l'
- ls -l
現(xiàn)在無濟(jì)于事:
- import string
- for char in string.ascii_letters:
- if char in locals():
- continue
- locals()[char] = char
- class PipePy:
- # __init__, etc
更危險的事情
用locals()給了我一個靈感。為什么我們必須一直實例化PipePy?我們無法在路徑中找到所有可執(zhí)行文件,并根據(jù)它們創(chuàng)建PipePy實例嗎?我們當(dāng)然可以!
- import os
- import stat
- for path in os.get_exec_path():
- try:
- names = os.listdir(path)
- except FileNotFoundError:
- continue
- for name in names:
- if name in locals():
- continue
- if 'x' in stat.filemode(os.lstat(os.path.join(path, name)).st_mode):
- locals()[name] = PipePy(name)
因此,現(xiàn)在,將我們擁有的所有內(nèi)容都放在一個python文件中,并刪除腳本(這是實際bash腳本的轉(zhuǎn)錄):
- from pipepy import mysqladmin, sleep, drush, grep
- for i in range(10):
- if mysqladmin('ping',
- host="mysql_drupal7",
- user="user",
- password="password"):
- break
- sleep(1)() # Remember to actually invoke
- if not drush('status', 'bootstrap') | grep('-q', 'Successful'):
- drush('-y', 'site-install', 'standard',
- db_url="mysql://user:password@mysql_drupal7:3306/drupal",
- acount_pass="kbairak")() # Remember to actually invoke
- drush('en', 'tmgmt_ui', 'tmgmt_entity_ui', 'tmgmt_node_ui')()