Python: Pass extra arguments to callable

Question

I have the following python code:

import networkx as nx 

def cost(i, j, d, value1, value2):
    # some operation involving all these values
    return cost


# graph is a networkx graph
# src, dst are integers
# cost is a callable that is passed 3 values automatically i.e. src, dst and d 
# d is a dictionary of edge weights
path = nx.dijkstra_path(graph, src, dst, weight=cost)

Now I want to pass two values value1 and value2 to the cost function.

The networkx documentation says the weight can be a callable that accepts exactly 3 arguments. But i need value1 and value2 for calculations. How can this be done?

Edit The solution using functools works well. However, my function is in a class as follows:

import networkx as nx 
import functools
class Myclass:
    def cost(self, i, j, d, value2):
        # some operation involving all these values
        # also need self


    # graph is a networkx graph
    # src, dst are integers
    # cost is a callable that is passed 3 values automatically i.e. src, dst and d 
    # d is a dictionary of edge weights
    # path = nx.dijkstra_path(graph, src, dst, cost)
    cost_partial = functools.partial(cost, self=self, value2=5)
    path = nx.dijkstra_path(graph, src, dst, cost_partial)

Using this approach, nx.dijkstra_path insists upon assigning src to self. Thus the interpreter complains that self is assigned multiple values. I need self for calculating the cost.

where would those values come from? if they are known and won't change for all of the function calls why not make them global variables? — AntiMatterDynamite, Aug 29 '18 at 12:33
@AntiMatterDynamite Because global variables are frowned upon since they tend to lead to a messy code, and are (very) rarely the correct solution — DeepSpace, Aug 29 '18 at 12:35
@DeepSpace frowned upon by who? the python elder gods? it all depends on how big of a script you're making ... — AntiMatterDynamite, Aug 29 '18 at 12:37
@AntiMatterDynamite I'm really not going to try and "convince" you why global variables are a bad idea. — DeepSpace, Aug 29 '18 at 12:39
@DeepSpace good because stackoverflow is not a place for opinions — AntiMatterDynamite, Aug 29 '18 at 12:39
@DeepSpace i wasn't giving advice to use global variables i was asking if the values are changing as the function is called again and again, which could invalidate some answers given here... — AntiMatterDynamite, Aug 29 '18 at 12:50
`self` seems to be undefined in the call to `functools.partial`. — Stop harming Monica, Aug 29 '18 at 13:04
Got it thanks. self is implicitly passed. No need to pass it separately. — Niloy Saha, Aug 29 '18 at 13:11
@AntiMatterDynamite by who? By most experienced developers. As one example, a little company called [google](http://google.github.io/styleguide/pyguide.html#25-global-variables), but I'm sure they don't know what they're doing... — strotter, Apr 28 '20 at 17:57
@strotter did you seriously just comment on a 2 year old post for that — AntiMatterDynamite, May 01 '20 at 01:31
Does this answer your question? [Python, how to pass an argument to a function pointer parameter?](https://stackoverflow.com/questions/13783211/python-how-to-pass-an-argument-to-a-function-pointer-parameter) — user202729, Aug 15 '21 at 06:14

FHTMitchell · Accepted Answer · 2018-08-29T12:35:49.397

12

You just need a function that wraps cost. A quick way is to use functools.partial.

import functools

def cost(i, j, d, value1, value2):
    'Example function'
    return i, j, d, value1, value2

# This makes a new function which takes all parameters of `cost` except those 
# you've already passed. Here I'm going to set value1 and value2 to 1 and 2 
# respectively.
cost_partial = functools.partial(cost, value1=1, value2=2)

# A new function that only accepts the 3 args i, j, d
cost_partial('i', 'j', 'd')  # --> ('i', 'j', 'd', 1, 2)

# use as
path = nx.dijkstra_path(graph, src, dst, weight=cost_partial)

edited Aug 29 '18 at 12:35

answered Aug 29 '18 at 12:30

FHTMitchell

11,793
2
35
47

Just to confirm that this solution worked like a charm for me in a similar case, trying to pass a fourth parameter to the filter_edge callable of the networkx.classes.function.subgraph_view(G, filter_node, filter_edge) function in networkx. Many thanks! – whydoesntwork Jul 04 '21 at 14:14

score 3 · Answer 2 · answered Aug 29 '18 at 12:34

This depends largely on the meaning on the meaning of value1 and value2. I would suggest adding a wrapper callable by networkx:

def cost_wrapper(i, j, d):
    value1 = 0  # set values as needed
    value2 = 1
    return cost(i, j, d, value1, value2)

And provide it to networkx:

path = nx.dijkstra_path(graph, src, dst, weight=cost_wrapper)

Or simply make them global variables, not arguments.

Python: Pass extra arguments to callable

2 Answers2

Linked

Related