Reading a named pipe continuosly

Question

What is the best way to read a named pipe continuously in Python?

This is my current code:

def read_commands():
    try:
        print "Creating read pipe..."
        os.mkfifo(pipe_cmd)    # Create pipe
        print "Pipe created!"
    except:
        print "Pipe already exists"

    with open(pipe_cmd, "r") as pipecmd:
        while True:
            try:
                line = pipecmd.readline()
            except:
                print "Could not read cmd pipe"

            if line != "":
                print line

        #time.sleep(1)

However, when I run this code it seems to be taking a lot of resources from my CPU (One of them will go to 100%). It works fine with the sleep of 1 second. However, I need to read the pipe continuously to make sure if there is new data . Is there a better way to accomplish this?

This is what I am sending to the pipe in C++:

void write_pipe(){
    ofstream pipe("/tmp/okccmd");  // Open the pipe
    string data = "Hi";
    pipe << data << endl;
    pipe.flush();
}

Thanks!

Answer 1

select.poll works fine (at least for Linux, not sure if Windows supports this; select.select ist afaik available, however). Just have a look at the documentation, the module is in the standard library and well documented (there is no need to know how the OS select() function actually works).

Documentation: https://docs.python.org/3/library/select.html

Note: poll() returns a list of file descriptors, not file-objects. So, you should have a dict which maps file descriptors to the corresponding object (I would have this also if I just poll one file.

pollobj = select.poll()
polled_files = dict()

# the following two lines are reuired for every file
pollobj.register(my_file_obj, <EVENTMASK>)
polled_files[my_file_obj.fileno()] = my_file_obj

for fd, evt in pollobj.poll():
    fileobj = polled_files[fd]
    ... process event for fileobj