Parallel context manager in the DSL. about stories HOT 8 CLOSED

Is it possible to config a parallel()? For example to choose behavior: failure if any fail and vice versa, wait first finished only, wait all etc

from stories.

I think it could be split into several context managers.

• parallel will execute every step specified
• wait_first will wait for the first step which returns Success. A step could return Skip marker to specify if we should ignore its status. #141 could be considered as well to give the marker a proper name.

In my opinion, this feature should not change the semantic of the Result and Failure markers. We also should not ignore uncaught exceptions in any steps.

Would it work for you?

from stories.

I think it could be split into several context managers.

• parallel will execute every step specified
• wait_first will wait for the first step which returns Success. A step could return Skip marker to specify if we should ignore its status. #141 could be considered as well to give the marker a proper name.

In my opinion, this feature should not change the semantic of the Result and Failure markers. We also should not ignore uncaught exceptions in any steps.

Would it work for you?

Yes wait_first sounds as good solution. Also I agree we have to keep current semantic. More over I think Skip is right naming in this context, because it closely to Ignore semantic then Break in #141. I guess we need something like a parallel_ignore_failure for case when we want run few tasks and want success any one. For example send a notification by mail, slack and push.

from stories.

For example send a notification by mail, slack, and push.

Is it really a parallel task? Or we want it to be a sequence that will exit after the first successful execution?

from stories.

Syntax:

from stories import story, parallel

class Order:
    @story
    def purchase(I):
        I.one
        with parallel:
            I.two
            I.three
        I.four

Requirements:

1. This feature should work both for sync and async stories. The semantic of execution and decision making should be the same. Or at least as close as possible.
2. parallel should have at least two steps in it.
3. Context changes made by two should not be visible to the three and wise versa. Context changes made by two and three should be visible by four.
4. two, three or both of them could be a substory. Context restrictions continue to be the same independent from this detail.
5. If some of the step returns Result this will be a result of the whole story. If other parallel steps are still running, their result should be ignored when they return. The story returns only after all the steps complete. If other running steps returns Failure or Result a warning of this clash should be reported.
6. If some of the step returns Failure this will be a failure of the whole story. If other parallel steps are still running, their result should be ignored when they return. The story fails only after all the steps complete. If other running steps returns Failure or Result a warning of this clash should be reported.
7. <parallel> should be present in the representation of the story.
8. History representation should contain the execution of all steps in order of the actual execution even if it is mixed between parallel substories.
9. The warning mentioned by items 5 and 6 of this list should be present in history representation as well.

Why don't we group the failures' reasons in a tuple and return that as a Failure instead of warning about other failures.
If there's a result and a failure, I think I'd rather the story to fail since it didn't complete successfully.

from stories.

Maybe we should not allow returning Result or Failure markers from the steps and substories within a parallel section?

from stories.

That's going to be a problem. Stories and steps can fail.
They can also return results.

from stories.

Superseded by #39

from stories.