'''

Module provides a class allowing to wrap communication over subprocess.Popen

input, output, error streams into a meaningfull, non-blocking, concurrent

stream processor exposing the output data as an iterator fitting to be a

return value passed by a WSGI applicaiton to a WSGI server per PEP 3333.

Copyright (c) 2011  Daniel Dotsenko <dotsa@hotmail.com>

This file is part of git_http_backend.py Project.

git_http_backend.py Project is free software: you can redistribute it and/or

modify it under the terms of the GNU Lesser General Public License as

published by the Free Software Foundation, either version 2.1 of the License,

or (at your option) any later version.

git_http_backend.py Project is distributed in the hope that it will be useful,

but WITHOUT ANY WARRANTY; without even the implied warranty of

MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License

along with git_http_backend.py Project.

If not, see <http://www.gnu.org/licenses/>.

'''

import os

import subprocess

import threading

from collections import deque

class StreamFeeder(threading.Thread):

    """

    Normal writing into pipe-like is blocking once the buffer is filled.

    This thread allows a thread to seep data from a file-like into a pipe

    without blocking the main thread.

    We close inpipe once the end of the source stream is reached.

    """

    def __init__(self, source):

        super(StreamFeeder, self).__init__()

        self.daemon = True

        filelike = False

        self.bytes = b''

        if type(source) in (type(''), bytes, bytearray): # string-like

            self.bytes = bytes(source)

        else: # can be either file pointer or file-like

            if type(source) in (int, long): # file pointer it is

                ## converting file descriptor (int) stdin into file-like

                try:

                    source = os.fdopen(source, 'rb', 16384)

                    pass

            # let's see if source is file-like by now

            try:

                filelike = source.read

                pass

        if not filelike and not self.bytes:

        self.source = source

        self.readiface, self.writeiface = os.pipe()

    def run(self):

        t = self.writeiface

        if self.bytes:

            os.write(t, self.bytes)

        else:

            s = self.source

            b = s.read(4096)

            while b:

                os.write(t, b)

                b = s.read(4096)

        os.close(t)

    @property

    def output(self):

        return self.readiface

class InputStreamChunker(threading.Thread):

    def __init__(self, source, target, buffer_size, chunk_size):

        super(InputStreamChunker, self).__init__()

        self.daemon = True  # die die die.

        self.source = source

        self.target = target

        self.chunk_count_max = int(buffer_size / chunk_size) + 1

        self.chunk_size = chunk_size

        self.data_added = threading.Event()

        self.data_added.clear()

        self.keep_reading = threading.Event()

        self.keep_reading.set()

        self.EOF = threading.Event()

        self.EOF.clear()

        self.go = threading.Event()

        self.go.set()

    def stop(self):

        self.go.clear()

        self.EOF.set()

        try:

    '''

    Processor class wrapping handling of subprocess IO.

    In a way, this is a "communicate()" replacement with a twist.

    - We are multithreaded. Writing in and reading out, err are all sep threads.

    - We support concurrent (in and out) stream processing.

    - The output is not a stream. It's a queue of read string (bytes, not unicode)

      chunks. The object behaves as an iterable. You can "for chunk in obj:" us.

    - We are non-blocking in more respects than communicate()

      (reading from subprocess out pauses when internal buffer is full, but

       does not block the parent calling code. On the flip side, reading from

       slow-yielding subprocess may block the iteration until data shows up. This

       does not block the parallel inpipe reading occurring parallel thread.)

    The purpose of the object is to allow us to wrap subprocess interactions into

    and interable that can be passed to a WSGI server as the application's return

    value. Because of stream-processing-ability, WSGI does not have to read ALL

    of the subprocess's output and buffer it, before handing it to WSGI server for

    HTTP response. Instead, the class initializer reads just a bit of the stream

    to figure out if error ocurred or likely to occur and if not, just hands the

    further iteration over subprocess output to the server for completion of HTTP

    response.

    The real or perceived subprocess error is trapped and raised as one of

    EnvironmentError family of exceptions

    Example usage:

    #    try:

    #        answer = SubprocessIOChunker(

    #            cmd,

    #            input,

    #            buffer_size = 65536,

    #            chunk_size = 4096

    #            )

    #    except (EnvironmentError) as e:

    #        print str(e)

    #        raise e

    #

    #    return answer

    '''

    def __init__(self, cmd, inputstream=None, buffer_size=65536,

                 chunk_size=4096, starting_values=[], **kwargs):

        '''

        Initializes SubprocessIOChunker

        '''

        if inputstream:

            input_streamer = StreamFeeder(inputstream)

            input_streamer.start()

            inputstream = input_streamer.output

        _p = subprocess.Popen(cmd,

            bufsize=-1,

            shell=True,

            stdin=inputstream,

            stdout=subprocess.PIPE,

            stderr=subprocess.PIPE,

            **kwargs

            )

        bg_out = BufferedGenerator(_p.stdout, buffer_size, chunk_size, starting_values)

        bg_err = BufferedGenerator(_p.stderr, 16000, 1, bottomless=True)

        while not bg_out.done_reading and not bg_out.reading_paused and not bg_err.length:

            # doing this until we reach either end of file, or end of buffer.

            bg_out.data_added_event.wait(1)

            bg_out.data_added_event.clear()

        # at this point it's still ambiguous if we are done reading or just full buffer.

        # Either way, if error (returned by ended process, or implied based on

        # presence of stuff in stderr output) we error out.

        # Else, we are happy.

        _returncode = _p.poll()

        if _returncode or (_returncode == None and bg_err.length):

            try:

                _p.terminate()

            except:

                pass

            bg_out.stop()

            bg_err.stop()

            raise EnvironmentError("Subprocess exited due to an error.\n" + "".join(bg_err))

        self.process = _p

        self.output = bg_out

        self.error = bg_err

    def __iter__(self):

        return self

    def next(self):

        if self.process.poll():

            raise EnvironmentError("Subprocess exited due to an error:\n" + ''.join(self.error))