Store tiger instance in global task context

[thomasst]

This makes it easier for libraries (e.g. FlowFish) to access the TaskTiger instance to queue further tasks.

from tasktiger._internal import g

def my_task(...):
    tiger = g["tiger"]
    tiger.delay(more_tasks)

I don't like that it's importing _internal but wasn't sure if there's a good way to expose it, perhaps just from tasktiger import g?

[jkemp101]

I think we should keep g hidden so we can refactor without breaking things. Expose it with a helper function or class method on TaskTiger? And do we need to ensure the Redis connections are closed before the child process uses it since we would be reusing the attached Redis instance from the parent process? I thought we did that somewhere but could find it with a quick search.

[thomasst]

The Redis connection is closed before forking in _execute, so we should be good. I added a class property, TaskTiger.current_instance.

[jkemp101]

Add a test or two and do we need to set it here also for eager tasks?

tasktiger/tasktiger/task.py

Lines 292 to 297 in f6ed7b5

	def execute(self):
	func = self.func
	is_batch_func = getattr(func, '_task_batch', False)
	g['current_task_is_batch'] = is_batch_func
	g['current_tasks'] = [self]

[thomasst]

Good catch with eager tasks.