queuelink 2.2.3

Quick start with link()

The link() factory function inspects your source and destination and wires everything up automatically — no need to choose between QueueLink, QueueHandleAdapterReader, or QueueHandleAdapterWriter by hand:

import queue
from queuelink import link

src = queue.Queue()
dst = queue.Queue()

# link() returns a result with stop(), close(), and is_alive()
result = link(src, dst)

src.put("hello")
print(dst.get())   # "hello"

result.stop()

link() accepts queues, file handles, file paths, and multiprocessing.connection.Connection as source or destination. Pass a list for fan-out:

result = link(src, [dst1, dst2])   # fan-out to two queues

Reading from a subprocess pipe into a queue:

import queue
from subprocess import Popen, PIPE
from queuelink import link

dest_q = queue.Queue()
proc = Popen(['myprogram'], stdout=PIPE, universal_newlines=True)

result = link(proc.stdout, dest_q)

line = dest_q.get()
result.stop()

Writing from a queue to a file:

import queue
from queuelink import link

src_q = queue.Queue()

with open("output.txt", "w") as f:
    result = link(src_q, f)
    src_q.put("hello\n")
    result.stop()

Use QueueLink directly when you need fine-grained control (registering/unregistering queues at runtime, accessing metrics via get_metrics()).

With QueueLink directly

A QueueLink is a one-way process that connects queues together. When two or more queues are linked, a sub-process is started to read from the “source” queue and write into the “destination” queue.

Circular references are not allowed.

Users create each queue from the Queue or Multiprocessing libraries. Those queues can then be added to a QueueLink instance as either the source or destination.

With standard queues

from queue import Queue
from queuelink import QueueLink

# Source and destination queues
source_q = Queue()
dest_q = Queue()

# Create the QueueLink
queue_link = QueueLink(name="my link")

# Connect queues to the QueueLink
source_id = queue_link.read(q=source_q)
dest_id = queue_link.write(q=dest_q)

# Text to send
text_in = "a😂" * 10

# Add text to the source queue
source_q.put(text_in)

# Retrieve the text from the destination queue!
text_out = dest_q.get()
print(text_out)

With a process manager

from multiprocessing import Manager
from queuelink import QueueLink

# Create the multiprocessing.Manager
manager = Manager()

# Source and destination queues
source_q = manager.JoinableQueue()
dest_q = manager.JoinableQueue()

# Create the QueueLink
queue_link = QueueLink(name="my link")

# Connect queues to the QueueLink
source_id = queue_link.read(q=source_q)
dest_id = queue_link.write(q=dest_q)

# Text to send
text_in = "a😂" * 10

# Add text to the source queue
source_q.put(text_in)

# Retrieve the text from the destination queue!
text_out = dest_q.get()
print(text_out)

Primary methods

These methods are used most common use cases.

• register_queue(q: UNION_SUPPORTED_QUEUES, direction: DIRECTION, start_method: str=None) -> client id: str

• stop

Secondary methods

These methods are less common.

• destructive_audit(direction: str)

• get_metrics() -> dict — returns latency and message-count data; see Metrics for details.

• get_queue(queue_id: [str, int])

• is_alive

• is_drained

• is_empty(queue_id:str =None)

• unregister_queue(queue_id: str, direction: str, start_method: str=None)

Multiprocessing

Start Method: QueueLink is tested against fork, forkserver, and spawn start methods. It defaults to the system preference, but can be overridden by passing the preferred start method name to the class “start_method” parameter.

Inbound Connections

To quickly link a pipe or handle with a queue, use QueueHandleAdapterReader. The Reader Adapter is tested against Multiprocessing Connections and Subprocess pipes. It calls flush and readline to consume from handles, so it should work against any object implementing those methods, with readline returning a string or byte array. For Multiprocessing Connections, the adapter injects a no-op flush method and a custom readline method.

# Text to send
text_in = "a😂" * 10

# Destination queue
dest_q = multiprocessing.Queue()  # Process-based

# Subprocess, simple example sending some text to stdout
# from subprocess import Popen, PIPE
proc = Popen(['echo', '-n', text_in],  # -n prevents echo from adding a newline character
             stdout=PIPE,
             universal_newlines=True,
             close_fds=True)

# Connect the reader
# from queuelink import QueueHandleAdapterReader
read_adapter = QueueHandleAdapterReader(queue=dest_q,
                                        handle=proc.stdout)

# Get the text from the queue
text_out = dest_q.get()
print(text_out)