Passing instance of type self to method as argument

Question

I would like to pass an instance of PipelineTask to PipelineTask.add however when I try I get a NameError which mentions PipelineTask is not defined.

I believe this is because PipelineTask is only bound after PipelineTask.__init__() is called.

class Task(BaseModel, abc.ABC):
    id: str

    @abc.abstractmethod
    async def run(self):
        pass


class PipelineTask(Task):
    @abc.abstractmethod
    async def run(self):
        pass

    def add(self, task: PipelineTask):
        ## TODO: how do a pass a instance of self here?
        next = self
        if self.next == None:
            self.next = task
            next = self.next
        else:
            current = self.next
            while current != None:
                current = current.next
            next = current
            current.next = next
        return self

class Pipeline(BaseModel):
    """
    Pipeline execute a sequence of tasks

    ...
    init = PipelineTask(id=0)
    pipeline = Pipeline(init=PrepareDataPipelineTask(id='prepare'))
    pipeline.add(ExecutePipelineTask(id='execute')).add(CollectResultsPipelineTask(id='collect'))
    pipeline.run()
    ...
    """

    # The pipelines innitial task
    init: PipelineTask

    async def run(self):
        await self.init.run()
        has_next = self.init.next != None
        next = self.init
        while has_next:
            next = next.next
            await next.run()
            has_next = next.next != None

    ## Adds a task to the end of the pipeline
    async def add(self, next: PipelineTask):
        """add Task to the end of the pipeline"""
        self.init.add(next)


class StdoutTask(PipelineTask):
    async def run(self):
        print(f"[Worker {self.id}] testing...")


async def test_create_pipeline():
    tasks = (
        StdoutTask(id=1, next=None)
        .add(StdoutTask(id=2, next=None))
        .add(StdoutTask(id=3, next=None))
    )
    pipeline = Pipeline(init=tasks)
    await pipeline.run()

Example usage:

class StdoutTask(PipelineTask):
    async def run(self):
        print(f"[Worker {self.id}] testing...")

@pytest.mark.asyncio
async def test_create_pipeline():
    tasks = StdoutTask(id=1).add(StdoutTask(id=2)).add(StdoutTask(id=3))
    pipeline = Pipeline(init=tasks)
    await pipeline.run()
    pass

I tried 'un'specifying task's type, but then task is missing the next attributes (e.g. AttributeError: 'XXXX' object has no attributed next)

def add(self, task):
    ...

I have also tried modifying task.__class__ = PipelineTask which adds additional methods but not attributes.

Below is a single file reproducible example

from pydantic import BaseModel
import abc
import asyncio


class Task(BaseModel, abc.ABC):
    id: str

    @abc.abstractmethod
    async def run(self):
        pass


class PipelineTask(Task):
    @abc.abstractmethod
    async def run(self):
        pass

    def add(self, task: PipelineTask):
        ## TODO: how do a pass a instance of self here?
        next = self
        if self.next == None:
            self.next = task
            next = self.next
        else:
            current = self.next
            while current != None:
                current = current.next
            next = current
            current.next = next
        return self


class Pipeline(BaseModel):
    """
    Pipeline execute a sequence of tasks

    ...
    init = PipelineTask(id=0)
    pipeline = Pipeline(init=PrepareDataPipelineTask(id='prepare'))
    pipeline.add(ExecutePipelineTask(id='execute')).add(CollectResultsPipelineTask(id='collect'))
    pipeline.run()
    ...
    """

    # The pipelines innitial task
    init: PipelineTask

    async def run(self):
        await self.init.run()
        has_next = self.init.next != None
        next = self.init
        while has_next:
            next = next.next
            await next.run()
            has_next = next.next != None

    ## Adds a task to the end of the pipeline
    async def add(self, next: PipelineTask):
        """add Task to the end of the pipeline"""
        self.init.add(next)


class StdoutTask(PipelineTask):
    async def run(self):
        print(f"[Worker {self.id}] testing...")


async def test_create_pipeline():
    tasks = (
        StdoutTask(id=1, next=None)
        .add(StdoutTask(id=2, next=None))
        .add(StdoutTask(id=3, next=None))
    )
    pipeline = Pipeline(init=tasks)
    await pipeline.run()

Solution

Use getattr

      def add(self, task: "PipelineTask"):
          next = getattr(self, "next", None)
          if self.next == None:
              self.next = task
              next = self.next
          else:
              current = getattr(self, "next", None)
              while current != None:
                  current = getattr(current, "next", None)
              next = getattr(current, "next", None)
              current.next = next
          return self

You can forward-declare types as string literals, e.g. `def add(self, task: 'PipelineTask')`. — Samwise, Nov 27 '21 at 15:48
Thanks for the tip but unfortunately it hasn't added the next attributed and I still get the same error. — Babbleshack, Nov 27 '21 at 15:53
Is that all your code? Because you don't assign a `next` attribute before trying to read it anywhere. — Jasmijn, Nov 27 '21 at 15:57
I am using pydantic so `__init__` is generated see: https://pydantic-docs.helpmanual.io/usage/models/#basic-model-usage — Babbleshack, Nov 27 '21 at 15:58
See [How to create a Minimal, Reproducible Example](https://stackoverflow.com/help/minimal-reproducible-example) — Jasmijn, Nov 27 '21 at 15:59
I make it a habit to always put all attributes used in a class set in `__init__`, even if only set to None. Even better is to add a comment there describing what it is used for. — Keith, Nov 28 '21 at 18:32

ti7 · Accepted Answer · 2021-11-28T18:17:00.290

2

The issue is likely where you have if self.next == None:; just set self.next inside __init__ or instead refer to it as getattr(self, "next", None)

>> class Test():
...     def add(self):
...         print(getattr(self, "missing", None))
... 
>>> t = Test()
>>> t.add()
None
>>> t.missing = "not missing"
>>> t.add()
not missing

However, it's likely better to set set the next attribute during __init__

class PipelineTask(Task):
    def __init__(self, *args, **kwargs):
        self.next = None                   # create .next attribute
        super().__init__(*args, **kwargs)  # call base class init

    ...

edited Nov 28 '21 at 18:17

answered Nov 27 '21 at 15:46

ti7

16,375
6
40
68

1

It would be better to just ensure that `self.next` is set before `add` can be called. – chepner Nov 27 '21 at 17:20
1

@chepner you're right; added a block for it! – ti7 Nov 28 '21 at 18:17

Passing instance of type self to method as argument

Solution

1 Answers1