Meta-ticket: Unify the use of weights in graphs

[dkrenn]

Consider the following example:

G = Graph([(1,2,0.1), (2,3,0.5), (1,3,0.3), (1,4,0.1)])  # whatever
G.min_spanning_tree(weight_function=lambda e:e[2])
w, L = T.longest_path(use_edge_labels=True)

In the last two lines, the functions are called differently, which is annoying, since one always has to look up the doc, to find out, which one is the correct.

Since both ways of calling are good (use_edge_labels simple and easy, weight_function more flexible), either should be possible for each function.

A first step is to unify the use of parameters by_weight, weight_function and check_weight

• Helper function to avoid duplicated code when using weights #32723 Helper function to avoid duplicated code when using weights
• Clean the use of weights in generic_graph.py - part 1 #32798 Clean the use of weights in generic_graph.py - part 1
• Clean the use of weights in generic_graph.py - part 2 #32800 Clean the use of weights in generic_graph.py - part 2
• Clean the use of weights in generic_graph.py - part 3 #32801 Clean the use of weights in generic_graph.py - part 3
• Clean the use of weights in graph.py #32802 Clean the use of weights in graph.py
• Clean the use of weights in digraph.py #32803 Clean the use of weights in digraph.py
• Clean the use of weights in path_enumeration.pyx #32804 Clean the use of weights in path_enumeration.pyx
• Clean the use of weights in spanning_tree.pyx #32805 Clean the use of weights in spanning_tree.pyx. Method random_spanning_tree in random_spanning_tree ignores weights #30566.
• c_graph.pyx - depends on other tickets
• boost_graph.pyx - depends on other tickets

We also have to deprecate use_edge_labels and use instead by_weight, etc.

and check methods silently using if self.weighted()

• Bad error message for weighted adjacency matrix #33562 Bad error message for weighted adjacency matrix
• distance_graph - requires to add parameters by_weights, ...

CC: @sagetrac-brunellus @fchapoton @tscrim

Component: graph theory

Keywords: weight graph

Issue created by migration from https://trac.sagemath.org/ticket/13112

[sagetrac-brunellus]

comment:1

Moreover, there are arguments named "by_weight" in distance, shortest_path, shortest_path_length…

[sagetrac-brunellus]

comment:3

I would go for single argument "by_weight" with following semantics of its values:

• False would be default with meaning "all the same".
• True would use labels.
• Custom function would be used translate edges to weight.

I think this would be easy to remember. On the other side, it is quite a radical change that would require long transition period.

What do you think?

[dcoudert]

comment:8

I'm turning this ticket to a task since it requires a long transition period.

[dcoudert]

comment:16

Help is more than welcome for reviewing the open tickets listed in this meta-ticket.