Add CMake build system, CLI, and project infrastructure by schulzchristian · Pull Request #1 · HeiHGM/Bmatching

schulzchristian · 2026-03-09T09:33:20Z

Summary

CMake build system alongside existing Bazel, with all dependencies fetched via FetchContent (Abseil, Protobuf, GoogleTest, Easylogging++, wide-integer)
bmatching_cli -- user-friendly CLI with flag-based interface replacing verbose textproto input
compile.sh -- one-command build script using all CPU cores
GitHub Actions CI -- build (Release/Debug) + smoke tests on every push
Release workflow -- automatically builds and uploads Linux binaries on tag push
Homebrew tap -- brew install --HEAD HeiHGM/bmatching/bmatching
Example graphs bundled in examples/ for instant tryout
README rewrite -- badges, citation, CLI examples for every algorithm, no proto/Bazel references
Regression tests -- 45/45 tests verify CMake CLI matches Bazel output exactly

Test plan

All 45 regression tests pass (Bazel vs CMake output identical)
CI passes on GitHub Actions
Verify brew install --HEAD HeiHGM/bmatching/bmatching works
Tag v1.0.0 and verify release workflow produces binary

FetchContent-based CMake build that mirrors all Bazel targets including optional features (Gurobi, SCIP, bSuitor, Karp-Sipser, hashing, tcmalloc). Generates build-info.h and hashing-config.h, compiles app_io.proto, and builds all core binaries: app, runner, fork_runner, generate_experiment_config, binary_to_textproto.

Replaces verbose textproto input with clean CLI flags: --graph, --algorithms, --capacity, --ordering_method, --timeout, etc. Supports all algorithm pipelines (greedy, reductions, ils, ilp_exact, presolved_ilp, scip, local_improvement, unfold) and output formats (text, json, binary). Uses OBJECT library for algorithm_impl_lib to ensure all REGISTER_IMPL statics are linked.

Includes compare.sh script, 3 test hypergraph instances, and report showing 45/45 tests pass across greedy (5 orderings), reductions+unfold, and reductions+greedy+unfold pipelines at capacities 1/3/5.

…orithms

- Bundle small.hgr and weighted.hgr example graphs for instant tryout - Add GitHub Actions CI: build (Release/Debug) + smoke test + unit tests - Add release workflow: builds and uploads Linux binary on tag push - Update README Quick Start to use bundled examples, add CI badge

Tap repo: HeiHGM/homebrew-bmatching

henrixapp · 2026-03-09T09:42:31Z

+std::vector<std::string> split(const std::string &s, char delim) {
+  std::vector<std::string> tokens;
+  std::istringstream stream(s);
+  std::string token;
+  while (std::getline(stream, token, delim)) {
+    // Trim whitespace
+    size_t start = token.find_first_not_of(" \t");
+    size_t end = token.find_last_not_of(" \t");
+    if (start != std::string::npos) {
+      tokens.push_back(token.substr(start, end - start + 1));
+    }
+  }
+  return tokens;
+}


why not use absl::StrSplit

ok, we can use this.

henrixapp · 2026-03-09T09:45:34Z

+    (*config.mutable_string_params())["ordering_method"] =
+        absl::GetFlag(FLAGS_ordering_method);


nit: Instead of (*config.mutable_string_params())["ordering_method"] = ...
use config.mutable_string_params()->insert({"key","value"});

not sure what you mean with nit

henrixapp · 2026-03-09T09:54:21Z

+  for cap in 1 3; do
+    run_test "$graph" "reductions+unfold" "$cap" \
+      "command:\"run\" hypergraph { file_path:\"$graph\" format:\"hgr\" } config { algorithm_configs { algorithm_name:\"reductions\" string_params{key:\"assume_sorted\" value:\"true\"} } algorithm_configs { algorithm_name:\"unfold\" string_params{key:\"assume_sorted\" value:\"true\"} } capacity:$cap short_name:\"test\" }" \
+      "--graph $graph --algorithms reductions,unfold --capacity $cap"
+  done


this test does not make sense. Why should you run reductions and directly unfold? unfold requires decisions made from other algorithms in between.

feel free to remove; AI slop

henrixapp · 2026-03-09T09:55:02Z

I don't think we need the report file as artifact in the repository.

sure, lets remove it; had in the repo mainly for you.

henrixapp · 2026-03-09T10:02:10Z

+[![DOI](https://img.shields.io/badge/DOI-10.7155%2Fjgaa.v30i1.3166-green.svg)](https://doi.org/10.7155/jgaa.v30i1.3166)

-**Engineering Hypergraph $b$-Matching Algorithms**
+**Tame your hypergraphs -- fast, modular b-matching at any scale.**


modular in this context could be mistaken for submodular.

henrixapp · 2026-03-09T10:17:44Z

+option(BMATCHING_USE_KARP_SIPSER  "Enable Karp-Sipser algorithm"            OFF)
+option(BMATCHING_USE_HASHING      "Enable hashing support (wide-integer)"   OFF)
+option(BMATCHING_USE_TCMALLOC     "Use tcmalloc from gperftools"            OFF)
+option(BMATCHING_ENABLE_LOGGING   "Enable easylogging++ logging output"     ON)


I would not enable this by default.

henrixapp · 2026-03-09T10:20:50Z

I acknowledge that the readme before was a bit unorganized.

However, I believe that the new state is not helpful, as it erases all information on how to reproduce experiments and use the runner framework. Maybe this can be moved into a different doc file.

create a folder with reproducability/reproducability.md; in any case, there is the zenodo link right? we can mention that in the readme too.

- Use absl::StrSplit instead of custom split() function - Use proto map insert() instead of operator[] for params - Fix --inplace flag description (uses newer ILS interface) - Remove nonsensical reductions+unfold test case - Delete report.txt build artifact - Change BMATCHING_ENABLE_LOGGING default to OFF - Fix tagline: "flexible" instead of "modular" - Add Zenodo data link and reproducibility docs

schulzchristian · 2026-03-09T12:56:32Z

Review feedback addressed

Use absl::StrSplit instead of custom split() function
Use proto map insert() instead of operator[] for params
Fix --inplace flag description (uses newer ILS interface)
Remove nonsensical reductions+unfold test case
Delete build_regression/ directory (compare.sh, report.txt, testdata)
Change BMATCHING_ENABLE_LOGGING default to OFF
Fix tagline: "flexible" instead of "modular"
Add Zenodo data link and reproducibility/reproducibility.md

henrixapp · 2026-03-09T13:35:46Z

+## 5. Plot Results
+
+```sh
+python3 tools/plot/plot.py <path_to_experiment> <path_to_experiment>/visualisation.textproto


The python plotting tool is out of date (only the C++ version works correctly).
So this should be:
bazel run -c opt tools/plot:plot_cc <path_to_experiment> <path_to_experiment>/visualisation.textproto

or add pybind/matplotlibcpp17 bindings and use CMake

…n script

schulzchristian added 15 commits March 9, 2026 08:23

Add compile.sh for CMake build

a1c15ce

Rewrite README for clarity and CMake instructions

e114564

Add build regression tests comparing Bazel and CMake output

31c4e11

Includes compare.sh script, 3 test hypergraph instances, and report showing 45/45 tests pass across greedy (5 orderings), reductions+unfold, and reductions+greedy+unfold pipelines at capacities 1/3/5.

Update README with CLI documentation, remove Bazel references

6163aee

Simplify built binaries section to only mention bmatching_cli

8412f23

Rewrite README: remove proto references, add CLI examples for all alg…

748800a

…orithms

Remove running experiments section from README

3cdd631

Remove adding a new algorithm section from README

d1498cc

Add badges, tagline, and full JGAA citation to README

66fa556

Add Homebrew tap and install instructions

b3df462

Tap repo: HeiHGM/homebrew-bmatching

Use --HEAD for brew install in README

347a48e

Fix CI: build only CLI targets to avoid pre-existing io utility errors

03fe647

henrixapp requested changes Mar 9, 2026

View reviewed changes

schulzchristian added 2 commits March 9, 2026 11:41

Remove build_regression directory

12fd770

henrixapp reviewed Mar 9, 2026

View reviewed changes

Fix plot command: use C++ plotter via Bazel instead of outdated Pytho…

c1cef92

…n script

henrixapp approved these changes Mar 9, 2026

View reviewed changes

henrixapp merged commit 7f0ef58 into main Mar 9, 2026
6 checks passed

schulzchristian deleted the cmake-switch branch March 9, 2026 13:54

		(*config.mutable_string_params())["ordering_method"] =
		absl::GetFlag(FLAGS_ordering_method);

Uh oh!

Conversation

schulzchristian commented Mar 9, 2026

Summary

Test plan

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

schulzchristian commented Mar 9, 2026

Review feedback addressed

Uh oh!

henrixapp Mar 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

henrixapp Mar 9, 2026 •

edited

Loading