taskproc 0.25.0

A lightweight task processing library written in pure Python

taskproc

A Python library for building/executing/managing task pipelines that is

• Minimal yet sufficient for general use.
• Focused on the composability of tasks for building large/complex pipelines effortlessly.

Features

• User defines a long and complex computation by composing shorter and simpler units of work, i.e., tasks.
• taskproc automatically executes them in a distributed way, supporting multithreading/multiprocessing and third-party job-controlling commands such as jbsub and docker. It also creates/reuses/discards a cache per task automatically.
• Full type-hinting support.

Nonfeatures

• Periodic scheduling
• Automatic retry
• External service integration (GCP, AWS, ...)
• Graphical user interface

Installation

pip install taskproc

Documentation

Basics

We define a task by class.

from taskproc import Task, Const, Graph

class Choose(Task):
    """ Compute the binomial coefficient. """

    def __init__(self, n: int, k: int):
        # Declaration of the upstream tasks.
        # Any instance of `Task` registered as an attirbute is considered as an upstream task.
        if 0 < k < n:
            self.left = Choose(n - 1, k - 1)
            self.right = Choose(n - 1, k)
        elif k == 0 or k == n:
            # We can also set dummy tasks with their value already computed.
            self.left = Const(0)
            self.right = Const(1)
        else:
            raise ValueError(f'{(n, k)}')

    def run_task(self) -> int:
        # The main computation of the task, which is delayed until necessary.
        # The return values of the prerequisite tasks are accessible via `.get_result()`.
        return self.left.get_result() + self.right.get_result()

# A task pipeline is constructed with instantiation, which should be done inside the `Graph` context.
with Graph('./cache'):  # Specifies the directory to store the cache.
    task = Choose(6, 3)  # Builds a pipeline to compute 6 chooses 3.

# Use the `run_graph()` method to run the pipeline.
ans, stats = task.run_graph()  # `ans` should be 6 chooses 3, which is 20. `stats` is the execution statistics.

Commandline Interface

taskproc has a utility classmethod to run with commandline arguments, which is useful if all you need is to run a single task. For example, if you have

# taskfile.py
from taskproc import Task, DefaultCliArguments
# ...

class Main(Task):
    def __init__(self):
        self.result = Choose(100, 50)
    
    def run_task(self):
        print(self.result.get_result())

# Optionally you can configure default CLI arguments.
DefaultCliArguments(
    # ...
).populate()

Then Main task can be run with CLI:

taskproc /path/to/taskfile.py -o /path/to/cache/directory

or

python -m taskproc /path/to/taskfile.py -o /path/to/cache/directory

Besides, if you have the entrypoint inside some module, you can run it with

# taskfile.py
...

class Main(Task):
    ...

# Must call the entrypoint explicitly.
if __name__ == '__main__':
    Main.cli()

and

python -m module.path.to.taskfile -o /path/to/cache/directory

See also taskproc /path/to/taskfile.py --help or python -m module.path.to.taskfile --help for more details.

Futures and Task Composition

To be more precise, any attributes of a task implementing the Future protocol are considered as upstream tasks. For example, Tasks and Consts are Futures. One can pass a future into the initialization of another task to compose the computation.

from taskproc import Future

class DownstreamTask(Task):
    def __init__(self, upstream: Future[int], other_args: Any):
        self.upstream = upstream  # Register upstream task
        ...

class Main(Task):
    def __init__(self):
        self.composed = DownstreamTask(
            upstream=UpstreamTaskProducingInt(),
            ...
        )

FutureList and FutureDict can be used to aggregate multiple futures into one, allowing us to register a batch of upstream futures.

from taskproc import FutureList, FutureDict

class SummarizeScores(Task):
    def __init__(self, task_dict: dict[str, Future[float]]):
        self.score_list = FutureList([ScoringTask(i) for i in range(10)])
        self.score_dict = FutureDict(task_dict)

    def run_task(self) -> float:
        # `.get_result()` evaluates `FutureList[T]` and `FutureDict[K, T]` into
        # `list[T]` and `dict[K, T]`, respectively.
        return sum(self.score_dict.get_result().values()) / len(self.score_dict.get_result())

If a future is wrapping a sequence or a mapping, one can directly access its element with the standard indexing operation. The result is also a Future.

class MultiOutputTask(Task):
    def run_task(self) -> dict[str, int]:
        return {'foo': 42, ...}

class DownstreamTask(Task):
    def __init__(self):
        self.dep = MultiOutputTask()['foo']  # type of Future[int]

Input and Output Specifications

In general, tasks can be initialized with any JSON serializable objects which may or may not contain futures. Any non-jsonable objects can be also passed, as the output of a task.

SomeTask(1, 'foo', bar={'baz': TaskProducingNonJsonableObj(), 'other': [1, 2, 3]})

On the other hand, the output of a task, i.e., the return value of the .run_task() method, should be serializable with cloudpickle.

Data Directories

Use task.task_directory to get a fresh path dedicated to each task. The directory is automatically created and managed along with the cache.

class TrainModel(Task):
    def run_task(self) -> str:
        ...
        model_path = self.task_directory / 'model.bin'
        model.save(model_path)
        return model_path

Task Label for Computational Resource Control

Each task class can be tagged with multiple labels. The task labels are useful to configure prefix commands and concurrency limits for controlling of computational resources.

class TaskUsingGPU(Task):
    task_label = 'gpu'
    ...

class AnotherTaskUsingGPU(Task):
    task_label = ['gpu', 'memory']
    ...

with Graph('./cache'):
    # Label-wise prefix/concurrency control
    SomeDownstreamTask().run_graph(
        # The number of tasks labeled with "gpu" running simultaneously is at most 2 (resp. "memory" is at most 1).
        rate_limits={
            'gpu': 2,
            'memory': 1,
            TaskUsingGPU.task_name: 5,  # Each task is also labeld with `cls.task_name` by default.
            },
        prefixes={
            # Task labeled with "gpu" is run with the job-dispatching command "jbsub ...".
            # Left labels prevail in the prefix collision.
            'gpu': 'jbsub -wait -queue x86_1h -cores 16+1 -mem 64g'  
            }
    ) 

Advanced Topics

Execution Policy Configuration

One can control the policy of parallelism with concurrent.futures.Executor classes.

from concurrent.futures import ProcessPoolExecutor, ThreadPoolExecutor

class MyTask(Task):
    ...

with Graph('./cache'):
    # Limit the number of parallel workers
    MyTask().run_graph(executor=ProcessPoolExecutor(max_workers=2))
    
    # Thread-based parallelism
    MyTask().run_graph(executor=ThreadPoolExecutor())

Selective Cache Deletion

It is possible to selectively discard cache:

with Graph('./cache'):
    # Selectively discard the cache of a specific task.
    Choose(3, 3).clear_task()

    # `ans` is recomputed tracing back to the computation of `Choose(3, 3)`.
    ans, _ = Choose(6, 3).run_graph()
    
    # Delete all the cache associated with `Choose`.
    Choose.clear_all_tasks()            

One can also manage caches directly from the disk location, i.e., ./cache in the above.

Cache Compression

The task output is compressed with gzip. The compression level can be changed as follows (defaults to 9), trading the space efficiency with the time efficiency.

class NoCompressionTask(Task):
    task_compress_level = 0
    ...

Built-in properties/methods

Below is the list of the built-in attributes/properties/methods of Task. Do not override these attributes in the subclass.

Name	Owner	Type	Description
run_task	instance	method	Run the task
task_name	class	property	String id of the task class
task_directory	instance	property	Path to the data directory of the task
run_graph	instance	method	Run the task after necessary upstream tasks and save the results in the cache
cli	class	method	run_graph with command line arguments
clear_task	instance	method	Clear the cache of the task instance
clear_all_tasks	class	method	Clear the cache of the task class
get_task_config	class	method	Get task config from the current graph
task_worker	instance	attribute	Task worker of instance
task_config	instance	attribute	Task config of instance
task_compress_level	instance	attribute	Compression level of instance
task_label	instance	attribute	Label of instance
get_result	instance	method	Get the result of the task (fails if the result is missing)
to_json	instance	method	Serialize itself as a JSON dictionary
get_workers	instance	method	Get the dictionary of the workers

Browsing caches

Show the whole task dependency tree:

tree -l /<path_to_cache_directory>/<task_name>/results/<root_task_id>

Show finished tasks only:

tree -l -P result.pkl.gz --prune /<path_to_cache_directory>/<task_name>/results/<root_task_id>

Show finished tasks + running tasks:

tree -l -P *.txt --prune /<path_to_cache_directory>/<task_name>/results/<root_task_id>

TODO

• Known issue

  • Current task argument serialization is not ideal since JSON is mapping two different values into the same text representation (e.g., tuple and list). Consider using consistency check x == json.loads(json.dumps(x)), or redesign the format.

• Feature enhancement

  • Error handling policy (eager/lazy) as an argument.
  • Task-state tracker.
  • Simple task graph visualizer.
  • Pydantic/dataclass support in task arguments (as an incompatible, but better-UX object with TypedDict).
  • Dynamic prefix generation with prefix template (e.g., for specifying the log locations).