An async queue with live progress display
Project description
aqueue
An async queue with live progress display. Good for running and visualizing tree-like I/O-bound processing jobs, such as website scrapes.
Example
import random
import trio
from aqueue import EnqueueFn, Display, run_queue, Item
class RootItem(Item):
async def process(self, enqueue: EnqueueFn, display: Display) -> None:
num_children = 3
display.overall.total = num_children
display.worker.description = "Making child items"
for _ in range(num_children):
# simulate doing work and creating more items
await trio.sleep(random.random())
enqueue(ChildItem())
class ChildItem(Item):
async def process(self, enqueue: EnqueueFn, display: Display) -> None:
display.worker.description = "Doing work..."
# Simulate doing work
await trio.sleep(random.random())
display.overall.completed += 1
def main() -> None:
run_queue(
initial_items=[RootItem()],
num_workers=2,
)
if __name__ == "__main__":
main()
Usage Notes
There's two things you need to do to use aqueue:
- Write your Item classes
- Start your queue with one of those items
Items
Items are your units of work. They can represent whatever you'd like, such as parts of a website that you're trying to scrape: an item for the index page, for subpages, for images, etc.
Each item should be an instance of a class that defines an async progress
method. As arguments, it
should accept two positional arguments:
- a
aqueue.EnqueueFn
that caan be called to enqueue more work. That type is simply an alias forCallable[[Item], None]
. - a
aqueue.Display
object that gives you control of the terminal display:
import aqueue
class MyItem(aqueue.Item):
async def process(self, enqueue: aqueue.EnqueueFn, display: aqueue.Display) -> None:
# make an HTTP request, parse it, etc
print('My item is processing!')
# when you discover more items you want to process, enqueue them:
enqueue(AnotherItem())
class AnotherItem(aqueue.Item):
async def process(self, enqueue: aqueue.EnqueueFn, display: aqueue.Display) -> None:
print('Another item is processing!')
As a rule of thumb, you should make a new item class whenever you notice a one-to-many relationship. For example, this one page has many images I want to download.
Note: process
is async, but because this library uses
Trio under the hood, you may only use
Trio-compatible primitives inside process
. For example, use trio.sleep
, not asyncio.sleep
.
TODO: consider AnyIO to avoid this problem?
Disclaimer: aqueue, or any asynchronous framework, is only going to be helpful if you're performing work is I/O-bound.
Starting your Queue
Then, start your queue with an initial item(s) to kick things off.
aqueue.run_queue(
initial_items=[MyItem()],
num_workers=2,
)
Queue type
By default, the queue is actually ...a queue -- that is to say that items are processed
first-in-first-out. Here are all the types you can specify with the queue_type_name
argument.
queue
- first-in-first-out processing, or breadth-first.stack
- last-in-first-out processing, or depth-first. This one is recommended for website scraping because it yields items fast (versusqueue
that tries to look at all the intermediate pages first).priority
- priority queue processing. In this case, your item objects should be orderable (with__lt__
, etc). Lesser objects will be processed first, because this code uses a minheap.
Number of workers
You can specify the number of workers you'd like to be processing your items with the num_workers
argument.
Ctrl-C
If you decide you want to stop your queue processing, press Ctrl-C.
If you've set the graceful_ctrl_c
to False, this will stop the program immediately. If True, the
default, aqueue will wait for the items currently being worked on to complete (without taking any
additional items), and then stop. Put another way, the choice is between responsiveness and
resource consistency.
Setting the look of the panels
Currently, only support for configuring the "Overall Progress" panel is supported. By default, the
panel is very simple. If you want to customize it, provide an iterable of
rich.progress.ProgressColumn
objects to the overall_progress_columns
argument. See
https://rich.readthedocs.io/en/stable/progress.html for more information. (Note that rich provides
all the nice terminal visualizations for aqueue!)
Updating the display
As mentioned, each process
method gets called with an aqueue.Display
object. The display has two
properties:
worker
, which lets you update the description of the worker who's currently processing this item.display.worker.description
is the getter/setter for that.overall
, which lets you access things in "Overall Progress" terminal panel.display.overall.completed
is a getter/setter for the number of completed things,display.overall.total
for the total number of things (or None), anddisplay.overall.total_f
for the total number of things or 0.
These panels are just an informational display for humans. They don't know about the queue churning through items of work. Therefore, you must decide what things you want to keep track of, and often, you won't be able to determine the complete number of things at the beginning. You'll need to do some intermediate processing and increment it slowly as more work is discovered. For example, if you want to keep track of images found and downloaded, you often won't be able to do that until you are searching deeper into the website.
Sharing state
Often, its beneficial to share state between the items. Using the website scrape example again, you may want to keep track of the URLs you've visited so you don't scrape them twice.
If this is needed, simply keep a global set/dict/list and store a key for the item. For example, a URL string may be a good key.
If you don't want to or can't use a global variable, consider a
ContextVar
.
Persisting state
During development, its probably likely that your program will crash after doing some work. For example, maybe your HTTP request timed out or you had a bug in your HTML parsing.
It's a shame to lose that work that's been done. So, if you're looking for a really handy way to
persist state across runs, check out the built-in
shelve
module. It's like a dict that
automatically saves to a file each time you set a key in it.
Other cool things
This library is fully docstringed and type-hinted 🥳
Installation
pip install "git+https://github.com/t-mart/aqueue"
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.