Eval: Using nested graphs

To create this graph we need only to install the tierkreis package:

pip install tierkreis

Graph

We can run graphs from within other graphs by using Graph.eval. Recall the fib_step graph that we wrote in the previous tutorial

from typing import NamedTuple

from tierkreis.builder import Graph
from tierkreis.builtins import iadd
from tierkreis.models import TKR


class FibData(NamedTuple):
    a: TKR[int]
    b: TKR[int]


builder = Graph(FibData, FibData)
sum = builder.task(iadd(builder.inputs.a, builder.inputs.b))
fib_step = builder.finish_with_outputs(FibData(builder.inputs.b, sum))

We create a graph fib4 that calls fib_step three times. The graph will have no inputs and gives a single integer as output:

from tierkreis.models import EmptyModel

fib4 = Graph(EmptyModel, TKR[int])

The Graph.eval method takes a Workflow object as its first argument and the appropriately typed input data as the second object.

second = fib4.eval(fib_step, FibData(a=fib4.const(0), b=fib4.const(1)))

We can iterate manually as follows:

third = fib4.eval(fib_step, second)
fourth = fib4.eval(fib_step, third)
workflow = fib4.finish_with_outputs(fourth.b)

In the next tutorial we will see how to iterate programmatically.

Execution

Since we still only use built-in functions, we execute the graph in the same way as before.

from uuid import UUID
from pathlib import Path

from tierkreis import run_graph
from tierkreis.storage import FileStorage, read_outputs
from tierkreis.executor import ShellExecutor

storage = FileStorage(UUID(int=99), name="Nested graphs using Eval")
executor = ShellExecutor(Path("."), workflow_dir=storage.workflow_dir)

storage.clean_graph_files()
run_graph(storage, executor, workflow, {})
print(read_outputs(workflow, storage))

3