Advanced: Dynamic Tasks

Sometimes you might not know exactly what other tasks to depend on until runtime. There are several cases of dynamic dependencies.

Fixed Dynamic

If you have a fixed set parameters, you can make requires() “dynamic”.

class TaskInput(d6tflow.tasks.TaskPqPandas):
    param = luigi.Parameter()
    ...

class TaskYieldFixed(d6tflow.tasks.TaskPqPandas):

    def requires(self):
        return dict([(s,TaskInput(param=s)) for s in ['a','b','c']])

    def run(self):
        df = self.inputLoad()
        df = pd.concat(df)

You could also use this to load an unknown number of files as a starting point for the workflow.

def requires(self):
    return dict([(s,TaskInput(param=s)) for s in glob.glob('*.csv')])

Collector Task

If you want to spawn multiple tasks without processing any of the outputs, you can use TaskAggregator. This task should do nothing but yield other tasks.

@d6tflow.requires(TrainModel1,TrainModel2) # inherit all params from input tasks
class TrainAllModels(d6tflow.tasks.TaskAggregator):

    def run(self):
        yield self.clone(TrainModel1)
        yield self.clone(TrainModel2)

If you want to run the workflow with multiple parameters at the same time, you can use TaskAggregator to yield multiple tasks.

class TaskAggregator(d6tflow.tasks.TaskAggregator):

    def run(self):
        yield TaskTrain(do_preprocess=False)
        yield TaskTrain(do_preprocess=True)

Fully Dynamic

This doesn’t work yet… It’s actually quite rare though that you need that though. Parameters normally fall in a fixed range which can be solved with the approaches above. Another typical reason you would want to do this is to load an unknown number of input files which you can do manually, see “Load External Files” in tasks.

class TaskA(d6tflow.tasks.TaskCache):
    param = luigi.IntParameter()
    def run(self):
        self.save(self.param)

class TaskB(d6tflow.tasks.TaskCache):
    param = luigi.IntParameter()

    def requires(self):
        return TaskA()

    def run(self):
        value = 1
        df_train = self.input(param=value).load()