Bound Diameter #2739

GrishaVar · 2025-03-25T10:44:09Z

Implements #2684

By submitting this pull request, I assign the copyright of my contribution to The igraph development team.

GrishaVar · 2025-03-25T18:56:23Z

@szhorvat This PR now also implements igraph_set_remove; doing that was the path of least resistance. See #2287: I guess these sets will be deprecated/replaced at some point (which I think is a good idea), but you'll need an interface for the remove function either way. The implementation has the same issues with iteration as igraph_set_add, though I think in my specific case this should only affect performance.

I was hoping everything I do would stay in one PR, let me know if you want the sets in a different one (or if you want a different solution entirely)

codecov · 2025-03-25T19:39:19Z

Codecov Report

Attention: Patch coverage is 94.67681% with 14 lines in your changes missing coverage. Please review.

Project coverage is 84.72%. Comparing base (9f727ac) to head (e506aba).

Files with missing lines	Patch %	Lines
src/paths/dijkstra.c	82.92%	7 Missing ⚠️
src/core/set.c	91.54%	6 Missing ⚠️
src/paths/distances.c	99.17%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2739      +/-   ##
===========================================
+ Coverage    84.68%   84.72%   +0.04%     
===========================================
  Files          380      380              
  Lines        62314    62577     +263     
  Branches     12192    12248      +56     
===========================================
+ Hits         52770    53021     +251     
- Misses        9544     9556      +12

Files with missing lines	Coverage Δ
src/paths/unweighted.c	`96.34% <100.00%> (+0.57%)`	⬆️
src/paths/distances.c	`99.78% <99.17%> (-0.22%)`	⬇️
src/core/set.c	`94.08% <91.54%> (-1.84%)`	⬇️
src/paths/dijkstra.c	`95.58% <82.92%> (-1.20%)`	⬇️

... and 2 files with indirect coverage changes

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9f727ac...e506aba. Read the comment docs.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

szhorvat

Just a few small comments now, we can discuss more in person.

include/igraph_paths.h

src/paths/distances.c

tests/unit/igraph_diameter_bound.c

src/paths/distances.c

GrishaVar · 2025-03-26T19:41:58Z

@szhorvat Added more tests, including Barabasi; they pass. Added clean-up for containers as well, without it the address sanitiser kept complaining. Also, I found an OBOE in the set_remove - it was accessing one element too many, but the unit tests didn't detect it because it was the element that got removed at the end. Wouldn't have found it without the sanitiser!

Remaining big TODOs:

Deal with disconnected graphs
Deal with directed graphs
Add vertex-choosing heuristic

tests/unit/igraph_diameter_bound.c

tests/unit/set.c

szhorvat · 2025-03-27T08:50:22Z

I wrote up some tips on efficient development: https://github.com/igraph/igraph/wiki/Tips-for-quicker-development

szhorvat · 2025-03-30T20:16:14Z

@GrishaVar Do you need any input from me right now? I hope I did not miss any questions.

szhorvat · 2025-04-03T10:08:11Z

@GrishaVar Can you please push what you have so far? This is the equivalent of the draft report submission for you.

szhorvat · 2025-04-05T10:50:40Z

@GrishaVar Can you please push what you have so far? This is the equivalent of the draft report submission for you.

Please don't forget to push.

GrishaVar · 2025-04-08T03:04:34Z

@szhorvat update:

I've added a heuristic for choosing the next vertex to explore. As the paper suggests, the first vertex is of max degree, after than we take largest upper and smallest lower eccentricity bound in a loop.
Disconnected graphs are supported now. We explore one component at a time until none are left; infinities from distance calculations are always ignored except to detect the components themselves, once per component. There is one new test for disconnected graphs, the same one igraph_diameter uses. Since the algorithm for doing this is not in the paper, the comments in that section are a bit more verbose.
The adjacency list thing is done, but only unweighted graphs. It looks like doing it for weighted graphs is a bit annoying because they use igraph_lazy_inclist_t instead. I think we discussed this at some point but I don't remember much about it; we should sync on this.
Remaining todos:
- Convert everything to (optionally) weighted diameter
- Move function to more appropriate places [will be done by @szhorvat]
- Add igraph_set_difference, would be an O(nlogn) -> O(n) speedup
- Figure out weird maxdegree_arg issue
- Documentation and more tests (if needed?)
- Get rid of all the prints

These are all pretty minor now, hopefully I can get this wrapped up soon 🙂

szhorvat · 2025-04-08T09:31:39Z

Convert everything to (optionally) weighted diameter

Let's put this to the bottom of the priority list, as it is very straightforward.

Get rid of all the prints

Optionally, you can use a debug macro that makes it possible to turn this on or off, on demand. You can grep the source tree for DEBUG to see a number of variations on this. The style used by MATCHING_DEBUG is probably one of the better ones.

It is up to you whether you want to do this. I don't have a strong opinion. You can judge best if people might feel the need to turn on debug logging in the future for this function.

Something that would be quite nice to have is a basic benchmark. You can see how these work in tests/benchmarks. I suggest looking at the more recent benchmark additions, and copying one.

The purpose of this is to make future tuning of this function easier.

I would put this second-to-last on the priority list, just above the weighted version. Actually, I would prioritize this over igraph_set_difference().

szhorvat · 2025-04-11T17:29:38Z

Decide if set_remove needs to return an error code.
Add example (or remove example from docs).
Extend benchmarks (sparse, dense, Barabasi-Albert model)
Finish error checking, replace weights ? ... with if (weights) ...
Change lazy_inclist to inclist
Change argument order, https://github.com/igraph/igraph/wiki/Guidelines-for-function-argument-ordering
Include \experimental on a separate line at the beginning of the doc comment

szhorvat · 2025-04-12T14:19:25Z

Could you please factor out the cleanup of debug() macros into a separate PR?

GrishaVar · 2025-04-12T16:10:51Z

@szhorvat I rebased to remove the changes in this commit. See #2749 for the new PR.

I can't check your checkboxes, but all the things you listed are done.

szhorvat · 2025-04-25T22:08:41Z

Copilot reviews got support for C a few days ago. Let's see if produces anything useful.

Co-authored-by: Copilot <[email protected]>

ntamas · 2025-05-15T20:25:24Z

@szhorvat Is this ready for merging, do you still want to have another review round, or shall I review it before merging?

szhorvat · 2025-05-15T20:42:28Z

It'd be great if you could do a review, but I want to go through it again in detail before merging. It's best if I do that only after I'm back in Reykjavík. Note: I did not check the set removal function at all yet.

greptile-apps

PR Summary

Implements a faster algorithm for computing graph diameters based on Borassi et al. (2015) paper, offering potential O(|E|) complexity for real-world networks versus the traditional O(|V||E|).

Added igraph_diameter_bound() in src/paths/distances.c with optimized BFS-based diameter computation
Enhanced src/core/set.c with three new operations: push_back, remove, and difference for O(log n) set manipulations
Added specialized single-source variants igraph_distances_1 and igraph_distances_dijkstra_1 in unweighted.c/dijkstra.c for improved performance
Included comprehensive test suite in tests/unit/igraph_diameter_bound.c covering various graph types and edge cases
Currently supports only undirected graphs, with thorough validation against original diameter calculation methods

_{14 files reviewed, 12 comments}
_{Edit PR Review Bot Settings | Greptile}

greptile-apps · 2025-07-02T23:19:48Z

src/core/set.h

 IGRAPH_PRIVATE_EXPORT igraph_error_t igraph_set_add(igraph_set_t *set, igraph_integer_t e);
+IGRAPH_PRIVATE_EXPORT igraph_error_t igraph_set_push_back(igraph_set_t *set, igraph_integer_t e);
+IGRAPH_PRIVATE_EXPORT void igraph_set_remove(igraph_set_t *set, igraph_integer_t e);
+IGRAPH_PRIVATE_EXPORT void igraph_set_difference(igraph_set_t *set, igraph_set_t *to_remove);


style: consider marking to_remove parameter as const since it's not modified

Suggested change

IGRAPH_PRIVATE_EXPORT void igraph_set_difference(igraph_set_t *set, igraph_set_t *to_remove);

IGRAPH_PRIVATE_EXPORT void igraph_set_difference(igraph_set_t *set, const igraph_set_t *to_remove);

greptile-apps · 2025-07-02T23:19:56Z

tests/benchmarks/igraph_diameter_bound.c

+#define DENS 3.0/VCOUNT
+#define REP 1000
+
+    igraph_erdos_renyi_game_gnp(&g, VCOUNT, DENS, IGRAPH_DIRECTED, IGRAPH_NO_LOOPS);


logic: Graph created with IGRAPH_DIRECTED but all diameter computations use IGRAPH_UNDIRECTED mode. Either the graph creation should be undirected or tests should cover directed case too.

greptile-apps · 2025-07-02T23:19:56Z

src/paths/dijkstra.c

+    IGRAPH_CHECK(igraph_2wheap_init(&Q, no_of_nodes));
+    IGRAPH_FINALLY(igraph_2wheap_destroy, &Q);


logic: Missing FINALLY cleanup handler for Q heap if vector_resize fails on line 284

greptile-apps · 2025-07-02T23:19:56Z

tests/benchmarks/igraph_diameter_bound.c

+    igraph_destroy(&g);
+
+#undef VCOUNT
+#undef DENS


logic: DENS macro undefined but used in Barabási graph section - should be ECOUNT instead

Suggested change

#undef DENS

#undef ECOUNT

greptile-apps · 2025-07-02T23:19:57Z

src/core/set.c

+ * \ingroup set
+ * \function igraph_set_remove
+ * \brief Adds an element to the set.


logic: Documentation header incorrectly says 'Adds an element to the set' but this is the remove function

Suggested change

* \ingroup set

* \function igraph_set_remove

* \brief Adds an element to the set.

/**

* \ingroup set

* \function igraph_set_remove

* \brief Removes an element from the set.

*

greptile-apps · 2025-07-02T23:19:59Z

tests/unit/igraph_diameter_bound.c

+    igraph_diameter_bound(&g, NULL, &result, IGRAPH_UNDIRECTED, false);
+    igraph_diameter(&g, NULL, &reference, NULL, NULL, NULL, NULL, IGRAPH_DIRECTED, 0);
+    IGRAPH_ASSERT(result == reference);


logic: IGRAPH_UNDIRECTED used for bound but IGRAPH_DIRECTED used for reference - this inconsistency could mask errors

greptile-apps · 2025-07-02T23:20:00Z

tests/unit/igraph_diameter_bound.c

+    igraph_destroy(&g);
+
+    // weighted ring graph
+    igraph_vector_init_real(&weights, 9, 1.0, 2.675, 3.0, 4.0, 5.5, 1.0, 1.0, 1.0, 1.0);


logic: weights vector initialized with 9 elements but ring graph has 10 nodes (10 edges) - potential buffer overflow

greptile-apps · 2025-07-02T23:20:02Z

src/paths/distances.c

+        if (!weights && igraph_set_size(&current_component) <= *diameter) {
+            break;
+        }


logic: Possible early termination bug - if !weights check should be comparing against current best bound not global diameter

greptile-apps · 2025-07-02T23:20:03Z

src/paths/distances.c

+                update_to_max(&VECTOR(ecc_lower)[w], ecc_v-d);
+                update_to_max(&VECTOR(ecc_lower)[w], d);
+                update_to_min(&VECTOR(ecc_upper)[w], ecc_v+d);


style: Update of ecc_lower twice using different formulas - these should be combined with a max() operation for clarity

greptile-apps · 2025-07-02T23:20:03Z

src/paths/distances.c

+    const igraph_t *graph,  // input graph
+    const igraph_vector_t *weights,  // optional weights
+    igraph_real_t *diameter,  // output diameter value
+    igraph_bool_t directed,  // treating this graph as undirected
+    igraph_bool_t unconn  // false: disconnected returns INF; true: returns largest diameter
+) {


logic: Parameter 'directed' description seems reversed - says 'treating this graph as undirected' but parameter indicates directedness

Suggested change

const igraph_t *graph, // input graph

const igraph_vector_t *weights, // optional weights

igraph_real_t *diameter, // output diameter value

igraph_bool_t directed, // treating this graph as undirected

igraph_bool_t unconn // false: disconnected returns INF; true: returns largest diameter

) {

const igraph_t *graph, // input graph

const igraph_vector_t *weights, // optional weights

igraph_real_t *diameter, // output diameter value

igraph_bool_t directed, // whether to consider directed paths

igraph_bool_t unconn // false: disconnected returns INF; true: returns largest diameter

Copilot

Pull Request Overview

Implements the diameter‐bounding algorithm (#2684) and adds supporting set utilities, shortest‐path helpers, tests, benchmarks, example, and documentation updates.

Added igraph_diameter_bound (and helper functions) to compute graph diameter with bounding/pruning.
Introduced igraph_distances_1 and igraph_distances_dijkstra_1 for single‐source BFS/Dijkstra.
Expanded set interface with igraph_set_push_back/remove/difference and added unit tests and benchmarks.

Reviewed Changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/paths/distances.c	Core implementation of `igraph_diameter_bound`.
src/paths/unweighted.c	Added `igraph_distances_1` single‐source BFS helper.
src/paths/dijkstra.c	Added `igraph_distances_dijkstra_1` single‐source Dijkstra helper.
src/core/set.h / src/core/set.c	Extended set API: push_back, remove, difference.
tests/unit/igraph_diameter_bound.c	Unit tests for `igraph_diameter_bound`.
tests/unit/set.c / tests/unit/set.out	Unit tests and expected output for new set operations.
tests/benchmarks/igraph_diameter_bound.c	Benchmarks for bound vs. original diameter routines.
include/igraph_paths.h / interfaces/functions.yaml	API and header updates for new functions.
examples/simple/igraph_diameter_bound.c	Example for new function.
doc/structural.xxml	Documentation inclusion for `igraph_diameter_bound`.

Comments suppressed due to low confidence (6)

src/core/set.c:355

The doc comment for this block is labeled \function igraph_set_remove but the implementation below is igraph_set_difference. Please update the \function tag to igraph_set_difference.

 * \function igraph_set_remove

src/paths/dijkstra.c:252

The doc comment for igraph_distances_dijkstra_1 is just TODO. Please provide a proper function description, parameters, and return value.

 * TODO

examples/simple/igraph_diameter_bound.c:33

The example calls igraph_diameter_bound but does not display or verify result. Consider adding printf or assertion to show the computed diameter in this example.

    igraph_diameter_bound(&g, NULL, &result, IGRAPH_UNDIRECTED, 0);

interfaces/functions.yaml:439

The parameter in code is named unconn but here it’s labeled unconnected. To avoid confusion for API consumers, align the YAML parameter name with the code (unconn) or vice versa.

igraph_diameter_bound:

src/paths/unweighted.c:608

No unit tests cover igraph_distances_1. Please add tests for this new helper to ensure correct behavior, including unreachable vertices and small graphs.

igraph_error_t igraph_distances_1(const igraph_adjlist_t *adjlist, igraph_vector_t *res, igraph_integer_t from) {

src/paths/dijkstra.c:254

igraph_distances_dijkstra_1 is introduced without any tests. Add unit tests that validate single‐source weighted distances, negative/NaN weight handling, and empty graphs.

igraph_error_t igraph_distances_dijkstra_1(const igraph_t *graph,

Copilot · 2025-07-02T23:20:40Z

src/paths/distances.c

+        // component cannot increase the lower bound. So we can safely skip it
+        // without exploring
+        if (!weights && igraph_set_size(&current_component) <= *diameter) {
+            break;


Using break here exits the outer while (!igraph_set_empty(&to_inspect)) loop, so later components are never inspected. Replace break with continue to skip only this component.

Suggested change

break;

continue;

ntamas · 2025-08-28T13:10:35Z

Retargeting for 1.1 as it is not API-breaking and we need to clear the issue list for the 1.0 milestone now that we are getting closer to the release date.

GrishaVar force-pushed the grigory_bound-diameter branch from af29a01 to 8482c56 Compare March 25, 2025 13:47

szhorvat reviewed Mar 25, 2025

View reviewed changes

GrishaVar force-pushed the grigory_bound-diameter branch from 9fa9bf4 to 63f4199 Compare March 26, 2025 18:21

szhorvat reviewed Mar 26, 2025

View reviewed changes

tests/unit/igraph_diameter_bound.c Outdated Show resolved Hide resolved

szhorvat reviewed Mar 26, 2025

View reviewed changes

tests/unit/set.c Show resolved Hide resolved

szhorvat self-assigned this Mar 31, 2025

GrishaVar force-pushed the grigory_bound-diameter branch 2 times, most recently from 4a87cbe to 24090e8 Compare April 11, 2025 14:31

GrishaVar force-pushed the grigory_bound-diameter branch from 80d39f5 to 9b7e6dd Compare April 12, 2025 16:05

GrishaVar requested a review from szhorvat April 12, 2025 16:06

szhorvat added this to the 1.0 milestone Apr 25, 2025

szhorvat requested a review from Copilot April 25, 2025 22:08

This comment was marked as outdated.

Sign in to view

szhorvat mentioned this pull request Apr 27, 2025

Improve performance by providing a precalculated adjacency list to igraph_distances_cutoff #2715

Open

GrishaVar added 3 commits May 11, 2025 00:52

feat: initial bound diameter implementation

d185e5b

feat: add igraph_set_remove

f9b0f85

fix: use set removal and add tests to diameter_bound

23b2f2b

GrishaVar and others added 15 commits May 11, 2025 00:52

chore: move prints to disableable logging

04d2a16

add error checks

8dc6882

add interface

26cbed6

style

81d0f75

chore: remove return values for set_remove and set_difference

a31c7c3

doc: add example

f12ee3b

bench: add benchmark

1eb4455

fix: clean up error checking

eaa5618

fix: switch lazy_inclist to inclist

bf342cd

fix: change arg order

d144068

chore: add \experimental to doc

b27d9b3

fix: fix benchmark prints

609b6e4

fix: bug in tie-breaking

9e67686

feat: add component skipping for small components in unweighted graphs

0975f4b

feat: use set_push_back instead of set_add to reduce O(logn) to O(1)

75f846e

szhorvat force-pushed the grigory_bound-diameter branch from 9b7e6dd to 75f846e Compare May 11, 2025 00:52

ntamas and others added 3 commits May 15, 2025 22:20

fix: fix missing comma in debug statement

3ff6d52

Co-authored-by: Copilot <[email protected]>

fix: fix ordering of arguments in igraph_diameter_bound() example

4772c59

Co-authored-by: Copilot <[email protected]>

Merge branch 'develop' into grigory_bound-diameter

5c56e28

test: update test for new API of igraph_diameter()

abe51b8

Merge branch 'develop' into grigory_bound-diameter

e506aba

szhorvat requested a review from Copilot July 2, 2025 23:19

greptile-apps bot reviewed Jul 2, 2025

View reviewed changes

Copilot AI reviewed Jul 2, 2025

View reviewed changes

ntamas modified the milestones: 1.0, 1.1 Aug 28, 2025

ntamas changed the base branch from develop to master September 19, 2025 13:15

	IGRAPH_PRIVATE_EXPORT void igraph_set_difference(igraph_set_t set, igraph_set_t to_remove);
	IGRAPH_PRIVATE_EXPORT void igraph_set_difference(igraph_set_t set, const igraph_set_t to_remove);

		IGRAPH_CHECK(igraph_2wheap_init(&Q, no_of_nodes));
		IGRAPH_FINALLY(igraph_2wheap_destroy, &Q);

Uh oh!

Bound Diameter #2739

Are you sure you want to change the base?

Bound Diameter #2739

Uh oh!

Conversation

GrishaVar commented Mar 25, 2025

Uh oh!

GrishaVar commented Mar 25, 2025

Uh oh!

codecov bot commented Mar 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

szhorvat left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

GrishaVar commented Mar 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

szhorvat commented Mar 27, 2025

Uh oh!

szhorvat commented Mar 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szhorvat commented Apr 3, 2025

Uh oh!

szhorvat commented Apr 5, 2025

Uh oh!

GrishaVar commented Apr 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szhorvat commented Apr 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szhorvat commented Apr 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szhorvat commented Apr 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

GrishaVar commented Apr 12, 2025

Uh oh!

szhorvat commented Apr 25, 2025

Uh oh!

This comment was marked as outdated.

Uh oh!

ntamas commented May 15, 2025

Uh oh!

szhorvat commented May 15, 2025

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

PR Summary

Uh oh!

greptile-apps bot Jul 2, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Jul 2, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Jul 2, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Jul 2, 2025

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Jul 2, 2025

codecov bot commented Mar 25, 2025 •

edited

Loading

GrishaVar commented Mar 26, 2025 •

edited

Loading

szhorvat commented Mar 30, 2025 •

edited

Loading

GrishaVar commented Apr 8, 2025 •

edited

Loading

szhorvat commented Apr 8, 2025 •

edited

Loading

szhorvat commented Apr 11, 2025 •

edited

Loading

szhorvat commented Apr 12, 2025 •

edited

Loading