17.1. subprocess — 子进程管理

2.4 新版功能.

subprocess 模块允许你生成新的进程，连接它们的输入、输出、错误管道，并且获取它们的返回码。此模块打算代替一些老旧的模块与功能：

os.system
os.spawn*
os.popen*
popen2.*
commands.*

Information about how this module can be used to replace the older functions can be found in the subprocess-replacements section.

参见

POSIX users (Linux, BSD, etc.) are strongly encouraged to install and use the much more recent subprocess32 module instead of the version included with python 2.7. It is a drop in replacement with better behavior in many situations.

PEP 324 – 提出 subprocess 模块的 PEP

17.1.1. 使用 subprocess 模块

The recommended way to launch subprocesses is to use the following convenience functions. For more advanced use cases when these do not meet your needs, use the underlying Popen interface.

subprocess.call(args, *, stdin=None, stdout=None, stderr=None, shell=False)

Run the command described by args. Wait for command to complete, then return the returncode attribute.

The arguments shown above are merely the most common ones, described below in 常用参数 (hence the slightly odd notation in the abbreviated signature). The full function signature is the same as that of the Popen constructor - this functions passes all supplied arguments directly through to that interface.

例如:

>>> subprocess.call(["ls", "-l"])
0

>>> subprocess.call("exit 1", shell=True)
1

警告

Using shell=True can be a security hazard. See the warning under 常用参数 for details.

注解

Do not use stdout=PIPE or stderr=PIPE with this function as that can deadlock based on the child process output volume. Use Popen with the communicate() method when you need pipes.

subprocess.check_call(args, *, stdin=None, stdout=None, stderr=None, shell=False)

Run command with arguments. Wait for command to complete. If the return code was zero then return, otherwise raise CalledProcessError. The CalledProcessError object will have the return code in the returncode attribute.

The arguments shown above are merely the most common ones, described below in 常用参数 (hence the slightly odd notation in the abbreviated signature). The full function signature is the same as that of the Popen constructor - this functions passes all supplied arguments directly through to that interface.

例如:

>>> subprocess.check_call(["ls", "-l"])
0

>>> subprocess.check_call("exit 1", shell=True)
Traceback (most recent call last):
   ...
subprocess.CalledProcessError: Command 'exit 1' returned non-zero exit status 1

2.5 新版功能.

警告

Using shell=True can be a security hazard. See the warning under 常用参数 for details.

注解

Do not use stdout=PIPE or stderr=PIPE with this function as that can deadlock based on the child process output volume. Use Popen with the communicate() method when you need pipes.

subprocess.check_output(args, *, stdin=None, stderr=None, shell=False, universal_newlines=False)

Run command with arguments and return its output as a byte string.

If the return code was non-zero it raises a CalledProcessError. The CalledProcessError object will have the return code in the returncode attribute and any output in the output attribute.

The arguments shown above are merely the most common ones, described below in 常用参数 (hence the slightly odd notation in the abbreviated signature). The full function signature is largely the same as that of the Popen constructor, except that stdout is not permitted as it is used internally. All other supplied arguments are passed directly through to the Popen constructor.

例如:

>>> subprocess.check_output(["echo", "Hello World!"])
'Hello World!\n'

>>> subprocess.check_output("exit 1", shell=True)
Traceback (most recent call last):
   ...
subprocess.CalledProcessError: Command 'exit 1' returned non-zero exit status 1

To also capture standard error in the result, use stderr=subprocess.STDOUT:

>>> subprocess.check_output(
...     "ls non_existent_file; exit 0",
...     stderr=subprocess.STDOUT,
...     shell=True)
'ls: non_existent_file: No such file or directory\n'

2.7 新版功能.

警告

Using shell=True can be a security hazard. See the warning under 常用参数 for details.

注解

Do not use stderr=PIPE with this function as that can deadlock based on the child process error volume. Use Popen with the communicate() method when you need a stderr pipe.

subprocess.PIPE

Special value that can be used as the stdin, stdout or stderr argument to Popen and indicates that a pipe to the standard stream should be opened.

subprocess.STDOUT

可被 Popen 的 stdin ， stdout 或者 stderr 参数使用的特殊值， 表示标准错误与标准输出使用同一句柄。

exception subprocess.CalledProcessError

Exception raised when a process run by check_call() or check_output() returns a non-zero exit status.

returncode

Exit status of the child process.

cmd

用于创建子进程的指令。

output

Output of the child process if this exception is raised by check_output(). Otherwise, None.

17.1.1.1. 常用参数

为了支持丰富的使用案例， Popen 的构造函数（以及方便的函数）接受大量可选的参数。对于大多数典型的用例，许多参数可以被安全地留以它们的默认值。通常需要的参数有：

args 被所有调用需要，应当为一个字符串，或者一个程序参数序列。提供一个参数序列通常更好，它可以更小心地使用参数中的转义字符以及引用（例如允许文件名中的空格）。如果传递一个简单的字符串，则 shell 参数必须为 True （见下文）或者该字符串中将被运行的程序名必须用简单的命名而不指定任何参数。

stdin, stdout and stderr specify the executed program’s standard input, standard output and standard error file handles, respectively. Valid values are PIPE, an existing file descriptor (a positive integer), an existing file object, and None. PIPE indicates that a new pipe to the child should be created. With the default settings of None, no redirection will occur; the child’s file handles will be inherited from the parent. Additionally, stderr can be STDOUT, which indicates that the stderr data from the child process should be captured into the same file handle as for stdout.

When stdout or stderr are pipes and universal_newlines is True then all line endings will be converted to '\n' as described for the universal newlines 'U' mode argument to open().

如果 shell 设为 True,，则使用 shell 执行指定的指令。如果您主要使用 Python 增强的控制流（它比大多数系统 shell 提供的强大），并且仍然希望方便地使用其他 shell 功能，如 shell 管道、文件通配符、环境变量展开以及 ~ 展开到用户家目录，这将非常有用。但是，注意 Python 自己也实现了许多类似 shell 的特性（例如 glob, fnmatch, os.walk(), os.path.expandvars(), os.path.expanduser() 和 shutil）。

警告

Executing shell commands that incorporate unsanitized input from an untrusted source makes a program vulnerable to shell injection, a serious security flaw which can result in arbitrary command execution. For this reason, the use of shell=True is strongly discouraged in cases where the command string is constructed from external input:

>>> from subprocess import call
>>> filename = input("What file would you like to display?\n")
What file would you like to display?
non_existent; rm -rf / #
>>> call("cat " + filename, shell=True) # Uh-oh. This will end badly...

shell=False disables all shell based features, but does not suffer from this vulnerability; see the Note in the Popen constructor documentation for helpful hints in getting shell=False to work.

When using shell=True, pipes.quote() can be used to properly escape whitespace and shell metacharacters in strings that are going to be used to construct shell commands.

这些选项以及所有其他选项在 Popen 构造函数文档中有更详细的描述。

17.1.1.2. Popen 构造函数

此模块的底层的进程创建与管理由 Popen 类处理。它提供了很大的灵活性，因此开发者能够处理未被便利函数覆盖的不常见用例。

class subprocess.Popen(args, bufsize=0, executable=None, stdin=None, stdout=None, stderr=None, preexec_fn=None, close_fds=False, shell=False, cwd=None, env=None, universal_newlines=False, startupinfo=None, creationflags=0)

Execute a child program in a new process. On Unix, the class uses os.execvp()-like behavior to execute the child program. On Windows, the class uses the Windows CreateProcess() function. The arguments to Popen are as follows.

args 应当是一个程序的参数列表或者一个简单的字符串。默认情况下，如果 args 是一个序列，将运行的程序是此序列的第一项。如果 args 是一个字符串，解释是平台相关的，如下所述。有关默认行为的其他差异，见 shell 和 executable 参数。除非另有说明，推荐将 args 作为序列传递。

On Unix, if args is a string, the string is interpreted as the name or path of the program to execute. However, this can only be done if not passing arguments to the program.

注解

shlex.split() 在确定正确 args 的正确标记化时非常有用，尤其是在复杂情况下:

>>> import shlex, subprocess
>>> command_line = raw_input()
/bin/vikings -input eggs.txt -output "spam spam.txt" -cmd "echo '$MONEY'"
>>> args = shlex.split(command_line)
>>> print args
['/bin/vikings', '-input', 'eggs.txt', '-output', 'spam spam.txt', '-cmd', "echo '$MONEY'"]
>>> p = subprocess.Popen(args) # Success!

特别注意，由 shell 中的空格分隔的选项（例如 -input）和参数（例如 eggs.txt ）位于分开的列表元素中，而在需要时使用引号或反斜杠转义的参数在 shell （例如包含空格的文件名或上面显示的 echo 命令）是单独的列表元素。

在 Windows，如果 args 是一个序列，他将通过一个在 Converting an argument sequence to a string on Windows 描述的方式被转换为一个字符串。这是因为底层的 CreateProcess() 只处理字符串。

参数 shell （默认为 False）指定是否使用 shell 执行程序。如果 shell 为 True，更推荐将 args 作为字符串传递而非序列。

On Unix with shell=True, the shell defaults to /bin/sh. If args is a string, the string specifies the command to execute through the shell. This means that the string must be formatted exactly as it would be when typed at the shell prompt. This includes, for example, quoting or backslash escaping filenames with spaces in them. If args is a sequence, the first item specifies the command string, and any additional items will be treated as additional arguments to the shell itself. That is to say, Popen does the equivalent of:

Popen(['/bin/sh', '-c', args[0], args[1], ...])

在 Windows，使用 shell=True，环境变量 COMSPEC 指定了默认 shell。在 Windows 你唯一需要指定 shell=True 的情况是你想要执行内置在 shell 中的命令（例如 dir 或者 copy）。在运行一个批处理文件或者基于控制台的可执行文件时，不需要 shell=True。

警告

Passing shell=True can be a security hazard if combined with untrusted input. See the warning under 常用参数 for details.

bufsize, if given, has the same meaning as the corresponding argument to the built-in open() function: 0 means unbuffered, 1 means line buffered, any other positive value means use a buffer of (approximately) that size. A negative bufsize means to use the system default, which usually means fully buffered. The default value for bufsize is 0 (unbuffered).

注解

If you experience performance issues, it is recommended that you try to enable buffering by setting bufsize to either -1 or a large enough positive value (such as 4096).

The executable argument specifies a replacement program to execute. It is very seldom needed. When shell=False, executable replaces the program to execute specified by args. However, the original args is still passed to the program. Most programs treat the program specified by args as the command name, which can then be different from the program actually executed. On Unix, the args name becomes the display name for the executable in utilities such as ps. If shell=True, on Unix the executable argument specifies a replacement shell for the default /bin/sh.

stdin, stdout and stderr specify the executed program’s standard input, standard output and standard error file handles, respectively. Valid values are PIPE, an existing file descriptor (a positive integer), an existing file object, and None. PIPE indicates that a new pipe to the child should be created. With the default settings of None, no redirection will occur; the child’s file handles will be inherited from the parent. Additionally, stderr can be STDOUT, which indicates that the stderr data from the child process should be captured into the same file handle as for stdout.

If preexec_fn is set to a callable object, this object will be called in the child process just before the child is executed. (Unix only)

If close_fds is true, all file descriptors except 0, 1 and 2 will be closed before the child process is executed. (Unix only). Or, on Windows, if close_fds is true then no handles will be inherited by the child process. Note that on Windows, you cannot set close_fds to true and also redirect the standard handles by setting stdin, stdout or stderr.

If cwd is not None, the child’s current directory will be changed to cwd before it is executed. Note that this directory is not considered when searching the executable, so you can’t specify the program’s path relative to cwd.

If env is not None, it must be a mapping that defines the environment variables for the new process; these are used instead of inheriting the current process’ environment, which is the default behavior.

注解

如果指定， env 必须提供所有被子进程需求的变量。在 Windows，为了运行一个 side-by-side assembly ，指定的 env 必须 包含一个有效的 SystemRoot。

If universal_newlines is True, the file objects stdout and stderr are opened as text files in universal newlines mode. Lines may be terminated by any of '\n', the Unix end-of-line convention, '\r', the old Macintosh convention or '\r\n', the Windows convention. All of these external representations are seen as '\n' by the Python program.

注解

This feature is only available if Python is built with universal newline support (the default). Also, the newlines attribute of the file objects stdout, stdin and stderr are not updated by the communicate() method.

If given, startupinfo will be a STARTUPINFO object, which is passed to the underlying CreateProcess function. creationflags, if given, can be CREATE_NEW_CONSOLE or CREATE_NEW_PROCESS_GROUP. (Windows only)

17.1.1.3. 异常

Exceptions raised in the child process, before the new program has started to execute, will be re-raised in the parent. Additionally, the exception object will have one extra attribute called child_traceback, which is a string containing traceback information from the child’s point of view.

最常见的被抛出异常是 OSError。例如，当尝试执行一个不存在的文件时就会发生。应用程序需要为 OSError 异常做好准备。

如果 Popen 调用时有无效的参数，则一个 ValueError 将被抛出。

check_all() 与 check_output() 在调用的进程返回非零退出码时将抛出 CalledProcessError。

17.1.1.4. Security

Unlike some other popen functions, this implementation will never call a system shell implicitly. This means that all characters, including shell metacharacters, can safely be passed to child processes. Obviously, if the shell is invoked explicitly, then it is the application’s responsibility to ensure that all whitespace and metacharacters are quoted appropriately.

17.1.2. Popen 对象

Popen 类的实例拥有以下方法：

Popen.poll()

Check if child process has terminated. Set and return returncode attribute.

Popen.wait()

等待子进程被终止。设置并返回 returncode 属性。

警告

This will deadlock when using stdout=PIPE and/or stderr=PIPE and the child process generates enough output to a pipe such that it blocks waiting for the OS pipe buffer to accept more data. Use communicate() to avoid that.

Popen.communicate(input=None)

Interact with process: Send data to stdin. Read data from stdout and stderr, until end-of-file is reached. Wait for process to terminate. The optional input argument should be a string to be sent to the child process, or None, if no data should be sent to the child.

communicate() returns a tuple (stdoutdata, stderrdata).

注意如果你想要向进程的 stdin 传输数据，你需要通过 stdin=PIPE 创建此 Popen 对象。类似的，要从结果元组获取任何非 None 值，你同样需要设置 stdout=PIPE 或者 stderr=PIPE。

注解

内存里数据读取是缓冲的，所以如果数据尺寸过大或无限，不要使用此方法。

Popen.send_signal(signal)

将信号 signal 发送给子进程。

注解

在 Windows， SIGTERM 是一个 terminate() 的别名。 CTRL_C_EVENT 和 CTRL_BREAK_EVENT 可以被发送给以包含 CREATE_NEW_PROCESS 的 creationflags 形参启动的进程。

2.6 新版功能.

Popen.terminate()

停止子进程。在 Posix 操作系统上，此方法发送 SIGTERM。在 Windows，调用 Win32 API 函数 TerminateProcess() 来停止子进程。

2.6 新版功能.

Popen.kill()

杀死子进程。在 Posix 操作系统上，此函数给子进程发送 SIGKILL 信号。在 Windows 上， kill() 是 terminate() 的别名。

2.6 新版功能.

以下属性也是可用的：

警告

使用 communicate() 而非 .stdin.write， .stdout.read 或者 .stderr.read 来避免由于任意其他 OS 管道缓冲区被子进程填满阻塞而导致的死锁。

Popen.stdin

If the stdin argument was PIPE, this attribute is a file object that provides input to the child process. Otherwise, it is None.

Popen.stdout

If the stdout argument was PIPE, this attribute is a file object that provides output from the child process. Otherwise, it is None.

Popen.stderr

If the stderr argument was PIPE, this attribute is a file object that provides error output from the child process. Otherwise, it is None.

Popen.pid

子进程的进程号。

注意如果你设置了 shell 参数为 True，则这是生成的子 shell 的进程号。

Popen.returncode

此进程的退出码，由 poll() 和 wait() 设置（以及直接由 communicate() 设置）。一个 None 值 表示此进程仍未结束。

A negative value -N indicates that the child was terminated by signal N (Unix only).

17.1.3. Windows Popen 助手

STARTUPINFO 类和以下常数仅在 Windows 有效。

class subprocess.STARTUPINFO

Partial support of the Windows STARTUPINFO structure is used for Popen creation.

dwFlags

A bit field that determines whether certain STARTUPINFO attributes are used when the process creates a window.

si = subprocess.STARTUPINFO()
si.dwFlags = subprocess.STARTF_USESTDHANDLES | subprocess.STARTF_USESHOWWINDOW
hStdInput

If dwFlags specifies STARTF_USESTDHANDLES, this attribute is the standard input handle for the process. If STARTF_USESTDHANDLES is not specified, the default for standard input is the keyboard buffer.

hStdOutput

If dwFlags specifies STARTF_USESTDHANDLES, this attribute is the standard output handle for the process. Otherwise, this attribute is ignored and the default for standard output is the console window’s buffer.

hStdError

If dwFlags specifies STARTF_USESTDHANDLES, this attribute is the standard error handle for the process. Otherwise, this attribute is ignored and the default for standard error is the console window’s buffer.

wShowWindow

If dwFlags specifies STARTF_USESHOWWINDOW, this attribute can be any of the values that can be specified in the nCmdShow parameter for the ShowWindow function, except for SW_SHOWDEFAULT. Otherwise, this attribute is ignored.

SW_HIDE is provided for this attribute. It is used when Popen is called with shell=True.

17.1.3.1. Constants

The subprocess module exposes the following constants.

subprocess.STD_INPUT_HANDLE

The standard input device. Initially, this is the console input buffer, CONIN$.

subprocess.STD_OUTPUT_HANDLE

The standard output device. Initially, this is the active console screen buffer, CONOUT$.

subprocess.STD_ERROR_HANDLE

The standard error device. Initially, this is the active console screen buffer, CONOUT$.

subprocess.SW_HIDE

Hides the window. Another window will be activated.

subprocess.STARTF_USESTDHANDLES

Specifies that the STARTUPINFO.hStdInput, STARTUPINFO.hStdOutput, and STARTUPINFO.hStdError attributes contain additional information.

subprocess.STARTF_USESHOWWINDOW

Specifies that the STARTUPINFO.wShowWindow attribute contains additional information.

subprocess.CREATE_NEW_CONSOLE

The new process has a new console, instead of inheriting its parent’s console (the default).

This flag is always set when Popen is created with shell=True.

subprocess.CREATE_NEW_PROCESS_GROUP

A Popen creationflags parameter to specify that a new process group will be created. This flag is necessary for using os.kill() on the subprocess.

This flag is ignored if CREATE_NEW_CONSOLE is specified.

17.1.4. Replacing Older Functions with the subprocess Module

In this section, “a becomes b” means that b can be used as a replacement for a.

注解

All “a” functions in this section fail (more or less) silently if the executed program cannot be found; the “b” replacements raise OSError instead.

In addition, the replacements using check_output() will fail with a CalledProcessError if the requested operation produces a non-zero return code. The output is still available as the output attribute of the raised exception.

In the following examples, we assume that the relevant functions have already been imported from the subprocess module.

17.1.4.1. Replacing /bin/sh shell backquote

output=`mycmd myarg`

becomes:

output = check_output(["mycmd", "myarg"])

17.1.4.2. Replacing shell pipeline

output=`dmesg | grep hda`

becomes:

p1 = Popen(["dmesg"], stdout=PIPE)
p2 = Popen(["grep", "hda"], stdin=p1.stdout, stdout=PIPE)
p1.stdout.close()  # Allow p1 to receive a SIGPIPE if p2 exits.
output = p2.communicate()[0]

The p1.stdout.close() call after starting the p2 is important in order for p1 to receive a SIGPIPE if p2 exits before p1.

Alternatively, for trusted input, the shell’s own pipeline support may still be used directly:

output=`dmesg | grep hda`

becomes:

output=check_output("dmesg | grep hda", shell=True)

17.1.4.3. Replacing os.system()

status = os.system("mycmd" + " myarg")
# becomes
status = subprocess.call("mycmd" + " myarg", shell=True)

注释:

  • Calling the program through the shell is usually not required.

A more realistic example would look like this:

try:
    retcode = call("mycmd" + " myarg", shell=True)
    if retcode < 0:
        print >>sys.stderr, "Child was terminated by signal", -retcode
    else:
        print >>sys.stderr, "Child returned", retcode
except OSError as e:
    print >>sys.stderr, "Execution failed:", e

17.1.4.4. Replacing the os.spawn family

P_NOWAIT example:

pid = os.spawnlp(os.P_NOWAIT, "/bin/mycmd", "mycmd", "myarg")
==>
pid = Popen(["/bin/mycmd", "myarg"]).pid

P_WAIT example:

retcode = os.spawnlp(os.P_WAIT, "/bin/mycmd", "mycmd", "myarg")
==>
retcode = call(["/bin/mycmd", "myarg"])

Vector example:

os.spawnvp(os.P_NOWAIT, path, args)
==>
Popen([path] + args[1:])

Environment example:

os.spawnlpe(os.P_NOWAIT, "/bin/mycmd", "mycmd", "myarg", env)
==>
Popen(["/bin/mycmd", "myarg"], env={"PATH": "/usr/bin"})

17.1.4.5. Replacing os.popen(), os.popen2(), os.popen3()

pipe = os.popen("cmd", 'r', bufsize)
==>
pipe = Popen("cmd", shell=True, bufsize=bufsize, stdout=PIPE).stdout

pipe = os.popen("cmd", 'w', bufsize)
==>
pipe = Popen("cmd", shell=True, bufsize=bufsize, stdin=PIPE).stdin

(child_stdin, child_stdout) = os.popen2("cmd", mode, bufsize)
==>
p = Popen("cmd", shell=True, bufsize=bufsize,
          stdin=PIPE, stdout=PIPE, close_fds=True)
(child_stdin, child_stdout) = (p.stdin, p.stdout)

(child_stdin,
 child_stdout,
 child_stderr) = os.popen3("cmd", mode, bufsize)
==>
p = Popen("cmd", shell=True, bufsize=bufsize,
          stdin=PIPE, stdout=PIPE, stderr=PIPE, close_fds=True)
(child_stdin,
 child_stdout,
 child_stderr) = (p.stdin, p.stdout, p.stderr)

(child_stdin, child_stdout_and_stderr) = os.popen4("cmd", mode,
                                                   bufsize)
==>
p = Popen("cmd", shell=True, bufsize=bufsize,
          stdin=PIPE, stdout=PIPE, stderr=STDOUT, close_fds=True)
(child_stdin, child_stdout_and_stderr) = (p.stdin, p.stdout)

On Unix, os.popen2, os.popen3 and os.popen4 also accept a sequence as the command to execute, in which case arguments will be passed directly to the program without shell intervention. This usage can be replaced as follows:

(child_stdin, child_stdout) = os.popen2(["/bin/ls", "-l"], mode,
                                        bufsize)
==>
p = Popen(["/bin/ls", "-l"], bufsize=bufsize, stdin=PIPE, stdout=PIPE)
(child_stdin, child_stdout) = (p.stdin, p.stdout)

Return code handling translates as follows:

pipe = os.popen("cmd", 'w')
...
rc = pipe.close()
if rc is not None and rc >> 8:
    print "There were some errors"
==>
process = Popen("cmd", shell=True, stdin=PIPE)
...
process.stdin.close()
if process.wait() != 0:
    print "There were some errors"

17.1.4.6. Replacing functions from the popen2 module

(child_stdout, child_stdin) = popen2.popen2("somestring", bufsize, mode)
==>
p = Popen("somestring", shell=True, bufsize=bufsize,
          stdin=PIPE, stdout=PIPE, close_fds=True)
(child_stdout, child_stdin) = (p.stdout, p.stdin)

On Unix, popen2 also accepts a sequence as the command to execute, in which case arguments will be passed directly to the program without shell intervention. This usage can be replaced as follows:

(child_stdout, child_stdin) = popen2.popen2(["mycmd", "myarg"], bufsize,
                                            mode)
==>
p = Popen(["mycmd", "myarg"], bufsize=bufsize,
          stdin=PIPE, stdout=PIPE, close_fds=True)
(child_stdout, child_stdin) = (p.stdout, p.stdin)

popen2.Popen3 and popen2.Popen4 basically work as subprocess.Popen, except that:

  • Popen raises an exception if the execution fails.
  • the capturestderr argument is replaced with the stderr argument.
  • stdin=PIPE and stdout=PIPE must be specified.
  • popen2 closes all file descriptors by default, but you have to specify close_fds=True with Popen.

17.1.5. 注释

17.1.5.1. Converting an argument sequence to a string on Windows

On Windows, an args sequence is converted to a string that can be parsed using the following rules (which correspond to the rules used by the MS C runtime):

  1. Arguments are delimited by white space, which is either a space or a tab.
  2. A string surrounded by double quotation marks is interpreted as a single argument, regardless of white space contained within. A quoted string can be embedded in an argument.
  3. A double quotation mark preceded by a backslash is interpreted as a literal double quotation mark.
  4. Backslashes are interpreted literally, unless they immediately precede a double quotation mark.
  5. If backslashes immediately precede a double quotation mark, every pair of backslashes is interpreted as a literal backslash. If the number of backslashes is odd, the last backslash escapes the next double quotation mark as described in rule 3.