Forwards & backwards flows

This recipe demonstrates how forwards and backwards flows work.

For demonstration, the CSV data is written directly in the cell below – in practice you would want to load data a file.

import pandas as pd
from io import StringIO

flows = pd.read_csv(StringIO("""
source,target,type,value
a,b,main,2
a,c,main,1
c,d,main,3
b,c,back,2
"""))

flows

	source	target	type	value
0	a	b	main	2
1	a	c	main	1
2	c	d	main	3
3	b	c	back	2

Here is one structure, with nodes b and c both in the same vertical slice:

from floweaver import *

# Set the default size to fit the documentation better.
size = dict(width=570, height=300)

nodes = {
    'a': ProcessGroup(['a']),
    'b': ProcessGroup(['b']),
    'c': ProcessGroup(['c']),
    'd': ProcessGroup(['d']),
    'back': Waypoint(direction='L'),
}

bundles = [
    Bundle('a', 'b'),
    Bundle('a', 'c'),
    Bundle('b', 'c', waypoints=['back']),
    Bundle('c', 'd'),
    Bundle('c', 'b'),
]

ordering = [
    [['a'], []],
    [['b', 'c'], ['back']],
    [['d'], []],
]

sdd = SankeyDefinition(nodes, bundles, ordering)

weave(sdd, flows).to_widget(**size)

Alternatively, if b is moved to the right, extra hidden waypoints are automatically added to get the b--c flow back to the left of c:

bundles = [
    Bundle('a', 'b'),
    Bundle('a', 'c'),
    Bundle('b', 'c'),
    Bundle('c', 'd'),
    Bundle('c', 'b'),
]

ordering = [
    [['a'], []],
    [['c'], ['back']],
    [['b', 'd'], []],
]

sdd = SankeyDefinition(nodes, bundles, ordering)

weave(sdd, flows).to_widget(**size)