posix_mkfifo() - posix函数(可移植操作系统接口)
posix_mkfifo()
(PHP 4, PHP 5, PHP 7)
Create a fifo special file(a named pipe)
说明
posix_mkfifo(string $pathname,int $mode): boolposix_mkfifo() creates a specialFIFOfile which exists in the file system and acts as a bidirectional communication endpoint for processes.
参数
$pathnamePath to theFIFOfile.
$modeThe second parameter$modehas to be given in octal notation(e.g. 0644). The permission of the newly createdFIFOalso depends on the setting of the current umask(). The permissions of the created file are(mode &~umask).
返回值
成功时返回TRUE
,或者在失败时返回FALSE
。
注释
Note:当启用安全模式时,PHP 会检查被操作的文件或目录是否与被执行的脚本有相同的 UID(所有者)。
For non-blocking, fopen'd read access to a "half-connected" pipe (created with /usr/bin/mkfifo, posix_mkfifo, etc.), I just go ahead and do: The "r+" allows fopen to return immediately regardless of external writer channel. You then have to use your own conventions to track $fh as a pseudo-read-only resource, since fwrite would technically be permitted as well. I've successfully used this approach on Linux with PHP 4.3.10 and PHP 5.2.4 with both half-connected (no writer yet) and pre-connected (writer already waiting) pipes, polling with stream_select as usual.
Couldn't you just open up a writer yourself right after you open up a reader? Just to be sure that you won't have any blocking issues...
Here is a possible solution to what - tech at kwur dot com- mentioned:
I faced the problem where i had a process (a server) that needed to take care of socket connection, and in the meanwhile get some data from the database. I didn't wanted to make the clients wait for the query execution time, so i decided to make a separate process that executes the query on the DB, and the two would communicate over a pipe. Of course i didn't wanted the server blocking if no data was available. So what i come up with is to use stream_select() , and to overcome the mentioned problem, i would fork the process, open up the pipe for writing in the child, this way the parent won't block when it opens the pipe.
here is some code
A way to have a non-blocking pipe reader is to check first if the pipe exists. If so, then read from the pipe, otherwise do other stuff. This will work assuming that the writer creates the pipe, writes on it, and after that deletes the pipe. This is a blocking writer: And this is the non-blocking reader:
Note (quoted from `man 7 pipe` on debian linux): "On some systems (but not Linux), pipes are bidirectional: data can be transmitted in both directions between the pipe ends. According to POSIX.1-2001, pipes only need to be unidirectional. Portable applications should avoid reliance on bidirectional pipe semantics." Linux pipes are NOT bidirectional. Also, it appears to me that the use of fifo (named) pipes in php is pretty pointless as there appears to be NO way of determining whether opening (let alone reading) from it will block. stream_select SHOULD be able to accomplish this, unfortunatly you cannot get to this point because even trying to OPEN a pipe for read will block until there is a writer. I even tried to use popen("cat $name_of_pipe", 'r'), and even it blocked until it was opened for write by another process.
Object Oriented FIFO Communication process:
This is still not a solution: if I listen to commands on a pipe and output status on a separate pipe, PHP will block on both opens because something else has not already connected to this pipe. Because I can't do a low-level fcntl() to to set O_NONBLOCK or something like it, this always locks up and is really stupid. The only way I can get it to work is to spawn seperate subshells with system() and have them cat, or echo respectively and then the pipes work properly...usually? Its alot of trouble that we can't set the blocking on the open!!
鹏仔微信 15129739599 鹏仔QQ344225443 鹏仔前端 pjxi.com 共享博客 sharedbk.com
图片声明:本站部分配图来自网络。本站只作为美观性配图使用,无任何非法侵犯第三方意图,一切解释权归图片著作权方,本站不承担任何责任。如有恶意碰瓷者,必当奉陪到底严惩不贷!