Matchings in Bipartite Graphs

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

The Hopcroft-Karp and Eppstein algorithms as implemented in NetworkX compute matchings for bipartite graphs faster than the algorithms for general graphs. I've included an override of the matching method in the bipartite graph class, with parameters that allow for using the old algorithms if desired.

Component: graph theory

Author: Zach Gershkoff

Branch: 5924a34

Reviewer: Julian Rüth, Travis Scrimshaw, David Coudert

Issue created by migration from https://trac.sagemath.org/ticket/22559

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

Branch: u/zgershkoff/matchings_in_bipartite_graphs

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

Commit: 7f07803

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

New commits:

7f07803

bipartite graph mathchings

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

Description changed:

--- 
+++ 
@@ -1 +1 @@
-
+The Hopcroft-Karp and Eppstein algorithms as implemented in NetworkX compute matchings for bipartite graphs faster than the algorithms for general graphs. I've included an override of the matching method in the bipartite graph class, with parameters that allow for using the old algorithms if desired.

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

22ba202

libraries for compatibility

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 7f07803 to 22ba202

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

d6215f1

removed tab character

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 22ba202 to d6215f1

[saraedum]

Work Issues: author is missing

[saraedum]

comment:5

zgershkoff, I could not find your handle on the wiki home page.

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

Author: Zachary Gershkoff

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

comment:6

I don't see anything in the developer's guide about this. Is this sufficient?

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

0da102a

fixed whitespace issue

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from d6215f1 to 0da102a

[saraedum]

Changed work issues from author is missing to none

[saraedum]

comment:9

Replying to @sagetrac-zgershkoff:

I don't see anything in the developer's guide about this. Is this sufficient?

Yes. It is probably not in the developer's guide. But the merge commit uses that information (and the Reviewer field) afaik. It would be great if you could also add yourself to the list on the wiki's homepage.

[saraedum]

comment:10

I made some minor changes. Let's see what the patchbots think about this.

[saraedum]

Changed branch from u/zgershkoff/matchings_in_bipartite_graphs to u/saraedum/matchings_in_bipartite_graphs

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

7f3cc61	Merge branch 'develop' into t/22559/matchings_in_bipartite_graphs
32eb46d	Revert "Import iteritems directly"

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 0da102a to 32eb46d

[saraedum]

Reviewer: Julian Rüth

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

comment:14

The patchbot says it fails python3 compatibility when it calls six.iteritems, which I find confusing-- I thought the purpose of the six library was python2-3 compatibility.

[90e8124e-743e-4424-88b0-22f5ae41372b]

comment:15

Just a hunch here: maybe you need to write "for (u, v) in six.iteritems..."?

I couldn't find anything online justifying this.

[fchapoton]

comment:16

Patchbot is not smart enough to recognize the six before iteritems.

You should rather use

from six import iteritems

...

for u,v in iteritems

[tscrim]

comment:17

Three other minor things:

• You should start with

  def matching(self, value_only=False, algorithm="Hopcroft-Karp",
               use_edge_labels=False, solver=None, verbose=0):
      r"""

  (and end with """).

• For the wikipedia, use

  :wikipedia:`Matching_(graph_theory)`

  which handles all of the linking and formatting.

• Change the first use_edge_labels to:

        - when set to ``True``, computes a weighted matching where each
          edge is weighted by its label (if an edge has no label, `1` is
          assumed); ignored if ``algorithm`` is not ``"Edmonds"``, ``"LP"``

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

Changed branch from u/saraedum/matchings_in_bipartite_graphs to u/zgershkoff/matchings_in_bipartite_graphs

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

4cabf0b

wrong kind of apostrophe

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 32eb46d to 4cabf0b

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

comment:20

I made all the documentation changes tscrim suggested except the last one, which I believe saraedum already addressed.

Replying to @fchapoton:

Patchbot is not smart enough to recognize the six before iteritems.

I changed it to just iteritems in bipartite_graph.py, but the way I had it before is exactly how it was used in graph.py. I didn't fix it in graph.py since it's not broken.

[tscrim]

comment:21

No, Julian did not address it already; in particular there are still 2 commas separated by a space. I also (slightly) think you should aim to make it not use periods to be consistent and concise.

[tscrim]

Changed reviewer from Julian Rüth to Julian Rüth, Travis Scrimshaw

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

6dd7eaa

typo and style fixes

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 4cabf0b to 6dd7eaa

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

comment:23

Ok, I see. Thank you for the feedback. I fixed the things you pointed out, but I changed the phrasing away from "ignored if algorithm is not "Edmonds", "LP"" because it's now configured to raise an error if use_edge_labels==True but algorithm is not set to "Edmonds" or "LP".

---

New commits:

6dd7eaa

typo and style fixes

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 6dd7eaa to a20b717

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

a20b717

typo fix

[tscrim]

comment:25

You need to revert back the indentation by 2 spaces.

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from a20b717 to e3c353e

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

e3c353e

fixed indentation

[tscrim]

comment:27

You also need to do the other two lines. See comment:17 (also, it would be good if the arguments for the function started at the same level, it's a PEP8 thing).

[c703ac91-1c31-4f3a-abb2-47af3ac2705d]

comment:28

Replying to @tscrim:

You also need to do the other two lines. See comment:17 (also, it would be good if the arguments for the function started at the same level, it's a PEP8 thing).

I believe the lines you're referring to are addressed in this commit https://github.com/sagemath/sagetrac-mirror/commit/5339528b85dd389a3e7633678fe609796fe787b0 - I'm not sure why it didn't show up in the trac ticket.

I'll look into PEP8 and make changes shortly.

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from e3c353e to fadefcb

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

fadefcb

PEP8 compliance

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from fadefcb to 96b1242

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

96b1242

Removed tab character from new lines

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Changed commit from 96b1242 to 0d0bbbc

[7ed8c4ca-6d56-4ae9-953a-41e42b4ed313]

Branch pushed to git repo; I updated commit sha1. New commits:

0d0bbbc

newline at end of file

[90e8124e-743e-4424-88b0-22f5ae41372b]

comment:32

I think this looks good now. Travis, do you agree?

[tscrim]

comment:33

Whoops, sorry this fell of my radar. Yep, looks good.

[dcoudert]

comment:34

Why don't you use G.networkx_graph() to get the networkx version of the graph ?