README.md

dagapp

Making apps from DAGs by just snapping your fingers

To install: pip install dagapp

What's a DAG?

What's an app?

This is a beautiful example of a DAG app woven into to an article: rent-or-buy article+app The app serves as an illustration of the article and the article as the documentation of the app. (Note: It was a lot more "beautiful" in 2014, but they changed the graphic now! Boo!)

Here's another example of interactive calculators.

What dagapp wants to become is a framework to generate such effective communication, effortlessly.

Enough theory. Here's how it works...

A simple example

First make a DAG

from meshed.dag import DAG


def b(a):
    return 2**a


def d(c):
    return 10 - (5**c)


def result(b, d):
    return b * d


dag = DAG((b, d, result))

Then make an app

from dagapp.base import dag_app
from functools import partial

dags = [dag]

if __name__ == "__main__":
    app = partial(dag_app, dags=dags)
    app()

Then run the app

>>> streamlit run example.py

... and this is what you get

A more complicated example

Let's say we want to create two different DAGs and view them on the same streamlit page, keep all the non-root nodes static and represent the number inputs for one of the DAGs as sliders. We will use functions defined in configs_example.py for this example.

Define the DAGs

To start we can create our DAGs just like in the previous example.

profit_dag = DAG((user_clicks, rev, cost, profit))
revenue_dag = DAG((partners, clicks, revenue))

dags = [profit_dag, revnue_dag]

Define the configs

Next we can define some configs for these DAGs. These configs should be a list of dictionaries, with each dictionary representing the configs for each DAG. Each config must contain a dictionary arg_types that matches each of the root nodes in the DAG to an input type (currently the options are: num, slider, text, list, dict). If arg_types is not explicitly defined, then it will try to infer from type annotations in the function definitions, and then default to num. If slider is defined as the arg_type for any of the root nodes, then another dictionary ranges must be defined that matches each of the nodes designated to be slider with a min and max value for that slider. The configs for this example can be seen below.

configs = [
    dict(
        arg_types=dict(
            a="num",
            b="num",
            cost_per_click="num",
            revenue_per_click="num",
        ),
    ),
    dict(
        arg_types=dict(
            max_partners="slider",
            cost_per_click="slider",
            price_elasticity="slider",
            partners="slider",
            clicks_per_partner="slider",
        ),
        ranges=dict(
            max_partners=[0, 2000],
            cost_per_click=[0.0, 1.0],
            price_elasticity=[0, 200],
            partners=[0, 1500],
            clicks_per_partner=[0.0, 10.0],
        ),
    ),
]

Make the app

We can now make our app in a similar manner as the previous example, this time defining our configs, as well as defining StaticPageFunc as our page factory to keep all non-root nodes static.

from dagapp.page_funcs import StaticPageFunc

if __name__ == "__main__":
    app = partial(dag_app, dags=dags, page_factory=StaticPageFunc, configs=configs)
    app()

Run the app

>>> streamlit run configs_example.py

... and this is what you get