Skip to content

Review: the R side of the main series, igraph 0.10.17 to 1.0.1 - #2964

Draft
krlmlr wants to merge 101 commits into
igraph:mainfrom
krlmlr:main-land-review
Draft

krlmlr wants to merge 101 commits into
igraph:mainfrom
krlmlr:main-land-review

Conversation

@krlmlr

@krlmlr krlmlr commented Oct 17, 2026 •

Copy link
Copy Markdown
Contributor

Superseded by #2995, built from main-green 2242c53.

For reading only, never merged. This shows what the vendoring series changes on main outside the vendored engine so far: the glue, R code, tests and generated interface that main will take with the 263 vendor commits replaying igraph 0.10.17 to 1.0.1, up to krlmlr/igraph@3c54c752e6 (engine 0.11.0-372-g3c54c752e). main itself will take the series as a landing, main-land-dev, which is not built yet; this pull request is how a reviewer of main reads the R side before that.

How this branch is built

The series is main-green in krlmlr/rigraph at krlmlr/rigraph@4fa92b502d, mirrored here as main-green: 263 vendor commits on a seed of main, and per-commit CI has a passing verdict for every one of them. It carries the igraph.dev flavor. scripts/series-rc.sh main, run locally without --push, gives the unflavored twin of every series commit, and the branch is written from those twins the way scripts/series-land.sh --review writes one from a landing, onto main at 267cc39:

  • The first commit removes every workflow. Without its engine the R side compiles against nothing, so a check could only fail, and with no workflow files this pull request starts no build or check. Only format-suggest.yaml runs, because as a pull_request_target workflow it is read from main; it formats the files and builds nothing.
  • Then there is one commit for each of the 100 vendor commits (out of 263) that change anything outside the vendored strand. The vendored strand is src/vendor/cigraph/, src/vendor/igraph_version.h, src/vendor/igraph_threading.h, src/vendor/io/, patch/, the engine's source lists src/sources-{igraph,glpk,mini-gmp}.mk, and DESCRIPTION's version. Each commit is titled by its upstream commit, carries the series commit's message with its R-side fix notes, and names that commit in a Series-commit: trailer.
  • Unlike --review, the generated interface is shown: R/aaa-*.R, src/rinterface.c, src/cpp11.cpp, R/cpp11.R and src/sources-glue-*.mk. Stimulus generates it from the vendored functions.yaml as well as from tools/stimulus/, so an interface change the engine makes has no tools/stimulus/ change to stand for it.
  • Outside the vendored strand this branch's tree is identical to the series tip's, without the flavor. The remaining difference to main-green is the igraph.dev rename, line for line the seed's flavor commit, then DESCRIPTION's version, and NEWS.md, where main has the two fledge bumps the series leaves out. Each commit's diff equals its series commit's diff with the flavor's names spelled back.
  • The series' own first commit, which cherry-picks four of main's commits in front of the vendor commits, is not shown. main has them already, and the series' tree there equals main's outside DESCRIPTION's version and NEWS.md.

Net change: 74 files, +2,937 −3,009, not counting the workflow files the first commit removes. The generated interface is 19 of those files, +1,981 −2,171. The rest is R code (11 files, +184 −31), glue and build files (16 files, +185 −85), tools/ (5 files, +450 −68), tests (7 files, +121 −638) and 16 Rd files with one line each.

Worth a look

  • R-side fixes in R/:
    • R/layout.R (+62): layout_with_sugiyama() rebuilds the extended graph, its map to the original edges and the dummy vertices' coordinates from the control points upstream now returns per edge, in the order igraph used to number them (sugiyama_extended_graph()). This is the largest piece of new R logic, and it relies on vendored patch 0106 for graphs with more than one component.
    • R/centrality.R: upstream always scales the largest hub and authority score to one. hits_scores(scale = FALSE) rescales to unit length in R, signals the superseded stage, and leaves all scores at one when there are no nonzero edges. eigen_centrality(), centr_eigen() and centr_eigen_tmax() stop passing the scale argument the engine deprecates.
    • R/topology.R: upstream reads and returns permutations the other way round. canonical_permutation() inverts the labeling with order(), and permute() inverts its argument by assignment, so an invalid permutation still fails upstream's check. graph.isoclass.subgraph() leaves converting vids to the impl, as upstream takes a vertex selector.
    • R/games.R: sample_pa() and sample_pa_age() pass NULL for an empty out-degree sequence, and sample_correlated_gnp() and sample_correlated_gnp_pair() invert the permutation as permute() does.
    • R/rewire.R: keeping_degseq() and each_edge() map loops and multiple to upstream's allowed_edge_types.
    • R/paths.R and R/structural-properties.R: all_simple_paths() passes maxlen and drops its post-processing, the weighted _dijkstra functions are merged into their base functions upstream, and six *_dijkstra_impl() names stay as aliases.
    • R/minimum.spanning.tree.R: mst() builds the forest from the edges minimum_spanning_tree_impl() returns, with all vertices.
    • R/assortativity.R, R/conversion.R and R/structural-properties.R pass weights = NA to impls that gained a weights argument.
  • Glue:
    • src/rrandom.c, src/rinterface.h and src/uuid.c: upstream removed RNG_BEGIN()/RNG_END(). R's RNG state is now loaded on the first draw of a call, saved after every call in IGRAPH_R_CHECK() and by the finalizer, and left alone by a call that draws nothing.
    • src/rinterface_extra.c (+112 −73) and src/rcallback.c follow upstream's API changes: attribute handlers, matrix lists, callback signatures, Johnson distances, maximal cliques and LAD.
    • IGRAPH_I_DESTROY() in src/rinterface.h is redefined because upstream dropped IGRAPH_I_ATTRIBUTE_DESTROY().
    • Build: src/Makevars.in, .ucrt and .win add the Infomap include path, src/vendor/config.h defines HAVE_INFOMAP, and src/vendor/igraph_config.h uses IGRAPH_BEGIN_C_DECLS.
  • Generated interface: src/rinterface.c (+1,037 −1,189), src/cpp11.cpp and R/aaa-*.R, driven by tools/stimulus/functions-R.yaml (+223 −43), types-RC.yaml and types-RR.yaml. 19 _impl() functions are removed and 12 are added. All 12 new ones land in a new R/aaa-uncategorized.R (+293) because tools/aaa-categories.yaml has no category for them: the spatial graphs (Delaunay, Gabriel, beta skeletons, relative and nearest neighbor graphs, edge lengths), the IEA games, community_leiden_simple and weighted_biadjacency.
  • Tests removed for removed C functions: 15 test_that() blocks in tests/testthat/test-aaa-auto.R and 14 snapshots in _snaps/aaa-auto.md go with the nine C functions the engine removes. They are hub and authority scores, the bipartite game, the Moran process and three imitation functions, transitive_closure_dag and weighted_sparsemat. No R code outside the generated files calls them. scripts/series-tests.sh main main-land-review --strict lists all 29 as lost, so the landing's test gate has to accept them before main-land-dev can be built.
  • Snapshot drift: in _snaps/aaa-auto.md, the Sugiyama layout and hub and authority scores change, and eigenvector_centrality_impl(scale = FALSE) warns that scale is deprecated and returns scores with a maximum of one. The GEM and Davidson-Harel layouts and the islands game draw a different random stream with the new RNG handling, and error sources move by a line. In _snaps/centrality.md, the ARPACK error is worded differently. In _snaps/incidence.md, graph_from_biadjacency_matrix() orders its edges differently.
  • New tests for R-side fixes are in test-assortativity.R, test-community.R and test-structural-properties.R.
  • Vendored patches 0105 to 0107 are in patch/, which belongs to the vendored strand, so they are not in this diff:
    • 0105 normalises a boolean in igraph_i_edge_type_to_loops_multiple(). Without it, sample_smallworld(loops = TRUE, multiple = TRUE) fails where R builds igraph_bool_t as int. It comes with the review commit 3c93667.
    • 0106 maps the Sugiyama control points of every component to their edges. It comes with the review commit 85d1c14.
    • 0107 keeps the vendored Qhull from referencing stderr under USING_R, for R CMD check. Its vendor commit changes nothing outside the strand, so it has no commit here.
    • main's patches 0103 and 0104 leave the stack at the first vendor commit, where upstream carries them.
  • NEWS.md has no entry from the series. The user-visible changes above are not mentioned there: hits_scores(scale = FALSE), the random streams of the islands game and the GEM and Davidson-Harel layouts, the edge order of graph_from_biadjacency_matrix() and the ARPACK error message.

Acting on a finding

Fix it on the series, not here. The loop folds each fix into the vendor commit that introduced the defect on main-dev, CI judges it, and main-green advances over it. The next review is rebuilt from the series and gets a pull request of its own, main-land-review-v2. Close this pull request and delete the branch once main has taken the landing.

🤖 Generated with Claude Code

https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk

claude added 30 commits October 17, 2026 00:15
The branch carries the R side of the main series without its engine,
so it compiles against nothing, and it is never merged.
Without workflow files a pull request from it starts nothing that
builds or tests it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@240ae9b

Date: 2026-04-12 16:28:02 +0200

Align with code in R package
Prune changelog
Align workflows with bleeding edge

Series-commit: 95c4647
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources (tag 0.11.0) to krlmlr/igraph@48766a9

Date: 2026-04-12 16:29:59 +0200

Remove clashing file

Snapshot: version_impl_clean() reports 0.11.0, the tag this commit vendors (from snapshot-cdb4f8abf954c2e4bb621b805e293e4267e51e04-rcc-smoke-null).

Series-commit: 5f67ae8
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@0f194d1

Date: 2026-04-18 13:02:06 +0000

nfc: remove RNG_BEGIN()/RNG_END() macros, seed via igraph_setup()

R-side fix: upstream removes RNG_BEGIN()/RNG_END(), which under R were GetRNGstate()/PutRNGstate() around every random C function, so nothing loaded or saved R's RNG state any more. src/rrandom.c now loads it on the first draw of a call into igraph, IGRAPH_R_CHECK() saves it after every call (igraph_rng_R_end()), and R_igraph_finalizer() saves what a draw outside IGRAPH_R_CHECK() left loaded; a call that draws nothing leaves .Random.seed alone, as before. A graph's id (Rx_igraph_add_env(), UUID_gen()) still draws without advancing .Random.seed. Rx_igraph_ac_random_numeric() drops the removed macros and saves the state itself. The src/*.dd lists name the new rrandom.h include. Snapshots in tests/testthat/_snaps/aaa-auto.md: two error sources move up a line with the macro gone, and the islands game and the GEM and Davidson-Harel layouts draw a different stream, because the macros nested inside one call (igraph_random_sample_real() in the islands game, the random start in the layouts) reloaded .Random.seed and replayed numbers already drawn.

Series-commit: ae1f640
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@868f59f

Date: 2026-04-18 11:58:25 +0000

nfc: remove igraph_multiple_t enum, use igraph_bool_t

R-side fix:

Upstream removes the igraph_multiple_t enum in favour of igraph_bool_t, so Rx_igraph_get_adjlist() in src/rinterface_extra.c now casts its multiple argument to igraph_bool_t for igraph_i_neighbors().

Series-commit: aa3e057
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@b4bd270

Date: 2026-04-18 12:13:06 +0000

nfc: collapse ARPACK error codes into IGRAPH_EARPACK
docs: document single-header inclusion policy for public API

Snapshot: ARPACK errors now give the condition before "ARPACK error" (tests/testthat/_snaps/centrality.md), as upstream rewords them here.

Series-commit: 0fd6dee
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@0ea7f0a

Date: 2026-04-18 12:40:27 +0000

nfc: remove deprecated IGRAPH_ENEGLOOP alias

R-side fix:

Upstream removes the deprecated IGRAPH_ENEGLOOP alias, so src/rinterface_extra.c raises IGRAPH_ENEGCYCLE, the value the alias stood for.

Series-commit: 0fc05e4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@cfce837

Date: 2026-04-18 12:55:19 +0000

nfc: drop void* argument from interruption handlers
nfc: collapse ARPACK error codes into IGRAPH_EARPACK, requires IGRAPH_EINVEID

R-side fix: upstream drops the void* argument of interruption handlers and has them return whether to interrupt. Rx_igraph_interrupt_handler() takes no argument and returns true when R has an interrupt pending, false otherwise.

Series-commit: 488aa4b
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@426e8ee

Date: 2026-04-12 18:41:49 +0000

nfc: igraph_vector_view() and related functions now return by value
nfc: rename igraph_integer_t to igraph_int_t, keep former as alias
chore: Reenable const warnings

R-side fix:

Upstream makes igraph_vector_view() and igraph_matrix_view() return the view by value, so the three calls in src/rinterface_extra.c assign the result instead of passing the target.

Vendoring fix: src/cliques/cliques.c is now as upstream has it at 426e8ee. The vendor run kept the previous content because upstream's change kept the file size, the replay's commits share one committer timestamp, and git's stat check (core.checkStat=minimal, core.trustctime=false) took the regenerated file as unchanged.

Snapshot: the dot-product warning is reported from games/dotproduct.c:87, two lines up, as this commit's edit to that file moves it (tests/testthat/_snaps/aaa-auto.md).

Series-commit: e077b35
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@8590948

Date: 2026-04-15 11:35:31 +0000

nfc: igraph_attribute_handler_t members now take igraph_attribute_record_list_t

R-side fix:

Upstream passes attribute records to the attribute handler as a const igraph_attribute_record_list_t instead of an igraph_vector_ptr_t, and stores the value in a union. The R attribute handler in src/rinterface_extra.c (init, add_vertices, add_edges and their helpers) now takes the record list, reads it with igraph_attribute_record_list_size() and igraph_attribute_record_list_get_ptr(), and reads the value through value.as_vector, value.as_vector_bool and value.as_strvector.

Series-commit: fe8850e
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@7a768e6

Date: 2026-04-15 18:42:14 +0000

nfc: invert semantics of igraph_permute_vertices() permutation argument

R-side fix: upstream inverts the permutation igraph_permute_vertices() takes and igraph_canonical_permutation() returns, now the old ID of each new vertex. permute() and canonical_permutation() in R/topology.R keep the new ID of each old vertex by inverting on the way in and out; permute() inverts by assignment, so an invalid permutation still fails upstream's check. The internal callers in R/operators.R go through permute() and keep working. The canonical_permutation_impl() snapshot shows upstream's direction.

Series-commit: 425b839
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@610de99

Date: 2026-04-15 19:25:38 +0000

nfc: igraph_delete_vertices_map() and igraph_induced_subgraph_map() use -1 for unmapped vertices

R-side fix: upstream renames igraph_delete_vertices_idx() to igraph_delete_vertices_map() and marks deleted or unmapped vertices with -1 instead of 0, here and in igraph_induced_subgraph_map(). tools/stimulus/functions-R.yaml keeps the R name delete_vertices_idx_impl() and declares idx as INDEX_VECTOR, so its values stay 1-based with 0 for a deleted vertex, as before; the categories follow the rename. induced_subgraph_map_impl()'s map now reads as 1-based indices with 0 for an unmapped vertex, where it used to be off by one, and its snapshot shows that.

Series-commit: 41a811c
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@e6da4d7

Date: 2026-04-15 23:35:33 +0000

nfc: change loops parameter from igraph_bool_t to igraph_loops_t in 7 functions

R-side fix: upstream turns the loops argument of igraph_degree(), igraph_maxdegree(), igraph_strength(), igraph_centralization_degree() and igraph_centralization_degree_tmax() from a boolean into igraph_loops_t, so degree() and the other R callers, which pass a logical, failed in the generated argument check. The LOOPS type in tools/stimulus/types-RR.yaml now takes a logical as before (TRUE counts loops, as "twice" does, FALSE ignores them) and knows the TWICE default, which generated a bare TWICE symbol; the R interface is unchanged.

Series-commit: 69c6d27
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@782d268

Date: 2026-04-16 00:36:16 +0000

nfc: igraph_bipartite_game_gnm() gains allowed_edge_types and edge_labeled parameters, add igraph_bipartite_iea_game()

R-side fix:

Upstream gives igraph_bipartite_game_gnm() allowed_edge_types and edge_labeled parameters, so its override in tools/stimulus/functions-R.yaml declares them with the upstream defaults (SIMPLE, False), and the generated interface is regenerated. sample_bipartite_gnm() keeps its interface and behaviour.

R-side fix: upstream's new igraph_i_edge_type_to_loops_multiple() stores `allowed_edge_types & IGRAPH_I_MULTI_LOOPS_SW` in an igraph_bool_t, which R builds as int (src/vendor/igraph_config.h), so the value is 4, never equals true, and allowing self-loops with multi-edges fails with "Either both multi-edges and multi-loops should be allowed or neither"; sample_smallworld(loops = TRUE, multiple = TRUE) reaches it from krlmlr/igraph@eb3a92b on. patch/0105-Read-multi-loops-as-a-boolean-in-igraph_i_edge_type_.patch reads the bit as a boolean, as the two lines above it do. Snapshot: bipartite_game_gnm_impl()'s error takes upstream's wording, "for bipartite G(n,m) model" (tests/testthat/_snaps/aaa-auto.md).

Series-commit: ff3c18c
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@3cbb8ba

Date: 2026-04-16 01:20:48 +0000

nfc: igraph_bipartite_game_gnp() gains allowed_edge_types and edge_labeled parameters

R-side fix:

Upstream gives igraph_bipartite_game_gnp() allowed_edge_types and edge_labeled parameters, so its override in tools/stimulus/functions-R.yaml declares them with the upstream defaults (SIMPLE, False), and the generated interface is regenerated. sample_bipartite_gnp() keeps its interface and behaviour.

Snapshot: bipartite_game_gnp_impl()'s error for an invalid vertex count takes upstream's wording, "for bipartite G(n,p) model" (tests/testthat/_snaps/aaa-auto.md).

Series-commit: 095c646
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@6bc87ad

Date: 2026-04-16 03:32:45 +0000

nfc: igraph_erdos_renyi_game_gnm() uses igraph_edge_type_sw_t, add igraph_iea_game()

R-side fix: upstream replaces the loops argument of igraph_erdos_renyi_game_gnm() with allowed_edge_types and adds edge_labeled, so sample_gnm() failed with an unused loops argument. A new Stimulus type, EDGE_TYPE_SW_LOOPS in tools/stimulus/types-RR.yaml and types-RC.yaml, takes a logical and passes IGRAPH_LOOPS_SW or IGRAPH_SIMPLE_SW, and the override in tools/stimulus/functions-R.yaml keeps loops in erdos_renyi_game_gnm_impl() with it, beside upstream's edge_labeled = FALSE. The generated interface is regenerated, sample_gnm() is unchanged, and the impl's error snapshot takes upstream's wording, "for G(n,m) model".

Series-commit: 7bbb485
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@9373961

Date: 2026-04-16 05:21:52 +0000

nfc: igraph_erdos_renyi_game_gnp() uses allowed_edge_types and edge_labeled parameters

R-side fix: upstream makes the same change to igraph_erdos_renyi_game_gnp(); the override in tools/stimulus/functions-R.yaml keeps loops in erdos_renyi_game_gnp_impl() as EDGE_TYPE_SW_LOOPS beside edge_labeled = FALSE, and the generated interface is regenerated, so sample_gnp() is unchanged. The error snapshots of erdos_renyi_game_gnp_impl() and correlated_pair_game_impl(), which samples through G(n,p), take upstream's wording, "for G(n,p) model".

Series-commit: 2aa2109
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@eb3a92b

Date: 2026-04-16 06:18:12 +0000

nfc: igraph_rewire_edges() uses igraph_edge_type_sw_t parameter

R-side fix: upstream replaces the loops and multiple arguments of igraph_rewire_edges() with allowed_edge_types. rewire_each_edge() in R/rewire.R, behind each_edge(), maps the two logicals onto "simple", "loops", "multi" or "all", so each_edge() keeps its interface and behaviour.

Series-commit: 326c6f6
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@17177c7

Date: 2026-04-16 07:14:21 +0000

nfc: igraph_sbm_game() uses igraph_edge_type_sw_t and removes n parameter

R-side fix: upstream drops the n argument of igraph_sbm_game(), now the sum of the block sizes, and replaces loops with allowed_edge_types. The override in tools/stimulus/functions-R.yaml keeps both in sbm_game_impl(): n as BLOCK_SIZES_SUM, a new R-only Stimulus type that checks n against the sum of the block sizes with igraph's former message, and loops as EDGE_TYPE_SW_LOOPS; the generated interface is regenerated and sample_sbm() is unchanged. The error snapshot loses igraph's "Invalid value" suffix and source line, because R raises the error now.

Series-commit: 3696bd1
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@a396457

Date: 2026-04-16 08:12:59 +0000

nfc: barabasi_game and related functions no longer treat empty outseq as NULL

R-side fix: upstream no longer reads an empty outseq of igraph_barabasi_game() and the aging and recent-degree games as a missing one, so sample_pa() and sample_pa_age() in R/games.R, which turned a missing out.seq into an empty vector, now pass NULL for an empty or missing one, keeping their interface and behaviour.

Series-commit: d0a288e
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@2bd0802

Date: 2026-04-16 09:32:04 +0000

nfc: port igraph_rewire() API change from next branch

R-side fix:

Upstream gives igraph_rewire() an optional igraph_rewiring_stats_t output. tools/stimulus/types-RC.yaml maps the new REWIRING_STATS type to a null pointer, so R_igraph_rewire() still returns the rewired graph alone, and the generated interface is regenerated.

Vendoring fix: src/games/degree_sequence.c is now as upstream has it at 2bd0802. The vendor run kept the previous content because upstream's change kept the file size, the replay's commits share one committer timestamp, and git's stat check took the regenerated file as unchanged.

R-side fix: igraph_rewire() takes allowed_edge_types in place of the rewiring mode, so rewire_impl() lost mode and keeping_degseq() failed with an unused argument. rewire_keeping_degseq() in R/rewire.R passes "loops" or "simple", what "simple_loops" and "simple" were, so keeping_degseq() is unchanged.

Series-commit: 2aef6e5
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@3ced2ff

Date: 2026-04-16 10:16:03 +0000

nfc: Reorder igraph_correlated_game() arguments

Vendoring fix: include/igraph_games.h is now as upstream has it at 3ced2ff. The vendor run kept the previous content because upstream's change kept the file size, the replay's commits share one committer timestamp, and git's stat check took the regenerated file as unchanged.

R-side fix: upstream now reads the permutation of igraph_correlated_game() and igraph_correlated_pair_game() as the old ID of each new vertex, as igraph_permute_vertices() does since krlmlr/igraph@7a768e6. sample_correlated_gnp() and sample_correlated_gnp_pair() in R/games.R invert it by assignment on the way in, as permute() does, so permutation stays the new ID of each old vertex.

Series-commit: 9633ae1
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@311f5af

Date: 2026-04-16 12:31:14 +0000

nfc: Rename igraph_lcf() to igraph_lcf_small() and igraph_lcf_vector() to igraph_lcf()
nfc: Document permutation semantics change in correlated game functions

R-side fix: upstream renames igraph_lcf_vector() to igraph_lcf() (and the variadic igraph_lcf() to igraph_lcf_small()). tools/stimulus/functions-R.yaml keys the definition by the new name and keeps the R name with NAME-R: lcf_vector, so lcf_vector_impl() and its callers are unchanged; tools/aaa-categories.yaml and tools/rebuild-cats.R follow the rename.

Series-commit: 526ba46
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@0a4030a

Date: 2026-04-16 13:26:20 +0000

nfc: igraph_static_fitness_game() uses igraph_edge_type_sw_t parameter

R-side fix: upstream replaces the loops and multiple arguments of igraph_static_fitness_game() and igraph_static_power_law_game() with allowed_edge_types. A new Stimulus type, EDGE_TYPE_SW_MULTIPLE, takes a logical multiple argument and adds IGRAPH_MULTI_SW to the EDGE_TYPE_SW_LOOPS loops argument it depends on. The overrides of both functions in tools/stimulus/functions-R.yaml use the pair, so static_fitness_game_impl(), static_power_law_game_impl(), sample_fitness() and sample_fitness_pl() keep their logicals.

Series-commit: d30bdd2
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@6319e3d

Date: 2026-04-16 15:28:46 +0000

nfc: igraph_watts_strogatz_game() uses igraph_edge_type_sw_t parameter
nfc: igraph_static_power_law_game() uses igraph_edge_type_sw_t

R-side fix: upstream replaces the loops and multiple arguments of igraph_watts_strogatz_game() with allowed_edge_types. Its override takes loops as EDGE_TYPE_SW_LOOPS and multiple as EDGE_TYPE_SW_MULTIPLE, so watts_strogatz_game_impl() and sample_smallworld() are unchanged.

Series-commit: 8249eda
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@d8febb1

Date: 2026-04-16 16:28:21 +0000

nfc: merge weighted variants of diameter, eccentricity, radius, graph_center, pseudo_diameter, average_path_length

R-side fix:

Upstream merges the weighted _dijkstra variants of diameter, pseudo_diameter, radius, graph_center, eccentricity and average_path_length into their base functions, which gain a weights argument in second position. The overrides in tools/stimulus/functions-R.yaml move to the base functions, Rx_igraph_diameter() and friends in src/rinterface_extra.c call igraph_diameter(), and eccentricity(), radius(), graph_center() and mean_distance() call the base _impl functions; the generated interface is regenerated.

R-side fix: merging the weighted _dijkstra variants into the base functions removed average_path_length_dijkstra_impl(), diameter_dijkstra_impl(), eccentricity_dijkstra_impl(), graph_center_dijkstra_impl(), pseudo_diameter_dijkstra_impl() and radius_dijkstra_impl(), which test-aaa-auto.R still calls. R/paths.R keeps the six names as aliases of the merged _impl() functions, which take the same arguments.

R-side fix: roxygen2 regenerates the documentation of average.path.length, distances, eccentricity, graph_center and radius, whose related C documentation now links the merged igraph_average_path_length(), igraph_eccentricity(), igraph_graph_center() and igraph_radius() instead of their _dijkstra variants.

Series-commit: 41e9f00
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@14deb59

Date: 2026-04-16 18:25:30 +0000

nfc: move weights parameter to 2nd position for efficiency functions

Series-commit: 068653e
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@f5ccd96

Date: 2026-04-16 19:25:02 +0000

nfc: igraph_get_all_simple_paths() returns results in vector list

R-side fix:

Upstream returns the paths of igraph_get_all_simple_paths() in an igraph_vector_int_list_t, declared as the VERTEX_INDICES_LIST type, which has no R conversion yet. tools/stimulus/types-RC.yaml and types-RR.yaml now give VERTEX_INDICES_LIST the conversions of VERTEXSET_LIST, and the generated interface is regenerated.

R-side fix: upstream returns the paths of igraph_get_all_simple_paths() as a list and replaces cutoff with minlen, maxlen and max_results, which default to UNLIMITED. tools/stimulus/types-RR.yaml maps the UNLIMITED default of an INTEGER argument to -1, and the generated interface is regenerated.

R-side fix: all_simple_paths() passes its cutoff as maxlen and takes the list as is. The snapshot of get_all_simple_paths_impl() shows the list, and its class check in test-aaa-auto.R follows it.

Series-commit: a26b455
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@87f2479

Date: 2026-04-16 21:28:32 +0000

Port: igraph_distances() and 4 others gain weights argument

R-side fix:

patch/0004-chore-Fix-remaining-Stimulus-types.patch rebased: upstream adds an OPTIONAL EDGEWEIGHTS weights argument to igraph_get_shortest_paths() and igraph_get_all_shortest_paths() and rewrites their DEPS lines, so the patch now adds "vertices ON graph" after "weights ON graph" there; the bipartite-types hunks are unchanged. The header-only upstream commit fb033592 before it is folded into this one, as the vendor run skips it.

R-side fix (buffer):

Upstream gives igraph_distances() and igraph_get_shortest_paths() a weights argument in second position, and igraph_distances_johnson() a mode argument. The unweighted branches in src/rinterface_extra.c pass null weights, and the Johnson calls pass IGRAPH_OUT, as the R side still simulates mode = "in" by swapping from and to.

Series-commit: ec06971
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@00e3d00

Date: 2026-04-16 22:12:05 +0000

nfc: Rename igraph_community_infomap() parameters e_weights/v_weights

R-side fix: upstream renames the e_weights and v_weights arguments of igraph_community_infomap() to edge_weights and vertex_weights. Its override keeps the R names with PARAM_NAMES, so community_infomap_impl() and cluster_infomap() are unchanged.

Series-commit: 91e9ba8
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
claude added 27 commits October 17, 2026 00:15
vendor: Update vendored sources to krlmlr/igraph@8bb67ce

Date: 2026-04-18 07:25:04 +0000

deprecated: Remove igraph_minimum_spanning_tree_unweighted()

Series-commit: 5e93c78
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@fe684fe

Date: 2026-04-18 07:39:09 +0000

deprecated: Remove igraph_sparsemat() and igraph_weighted_sparsemat()

R-side fix: upstream removes igraph_weighted_sparsemat(), and with it weighted_sparsemat_impl(), which only its generated tests called. The tests and snapshots go, with a comment naming the upstream commit.

Series-commit: 633c618
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@66ba309

Date: 2026-04-18 08:36:37 +0000

deprecated: Remove igraph_sample_dirichlet(), igraph_sample_sphere_surface(), igraph_sample_sphere_volume()

R-side fix: upstream removes igraph_sample_sphere_surface(), igraph_sample_sphere_volume() and igraph_sample_dirichlet() in favour of igraph_rng_sample_*(), which take an RNG; the exported sample_sphere_surface(), sample_sphere_volume() and sample_dirichlet() lost their impls. Overrides in functions-R.yaml generate the impls under their R names from the new functions. A new Stimulus type RNG_DEFAULT (types-RC/RR.yaml) passes igraph_rng_default(), R's RNG, as the removed functions used. aaa-categories.yaml lists the new names. Snapshot: the dot_product_game_impl() warning moves to games/dotproduct.c:86.

Series-commit: 4a814f0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@0bfd80a

Date: 2026-04-18 08:47:57 +0000

deprecated: Remove igraph_Calloc(), igraph_Realloc(), igraph_Free() macros

R-side fix:

Upstream removes the igraph_Calloc() macro. src/rinterface_extra.c uses IGRAPH_CALLOC() instead, its replacement.

Series-commit: 2d6d7f0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@49f029a

Date: 2026-04-18 21:22:35 +0200

chore: EDGE_WEIGHTS

R-side fix:

patch/0004-chore-Fix-remaining-Stimulus-types.patch rebased: upstream renames the EDGEWEIGHTS type to EDGE_WEIGHTS, which the patch carries as context; its changes are unchanged. The header-only upstream commits 8f6c9024 and 2d634933 before it are folded into this one, as the vendor run skips them.

R-side fix:

Upstream renames the EDGEWEIGHTS type to EDGE_WEIGHTS. tools/stimulus/types-RC.yaml, types-RR.yaml and functions-R.yaml follow the rename, so the generated interface again passes weights by pointer and looks up the "weight" edge attribute.

Series-commit: 28854f2
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@8a8c86e

Date: 2026-04-18 21:24:01 +0200

chore: VERTEX_WEIGHTS

R-side fix:

Upstream renames the VERTEXWEIGHTS type to VERTEX_WEIGHTS. tools/stimulus/types-RC.yaml and types-RR.yaml follow the rename, so the generated interface again passes vertex weights by pointer and looks up the "weight" vertex attribute.

Series-commit: b98a57e
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@dbcec08

Date: 2026-04-18 21:29:43 +0200

chore: EDGE_INDICES_LIST

R-side fix:

patch/0004-chore-Fix-remaining-Stimulus-types.patch rebased: upstream renames the EDGESET_LIST type to EDGE_INDICES_LIST, which the patch carries as context in igraph_get_all_shortest_paths(); its changes are unchanged.

R-side fix:

Upstream renames the EDGESET_LIST type to EDGE_INDICES_LIST. tools/stimulus/types-RC.yaml, types-RR.yaml and functions-R.yaml follow the rename, so the generated interface again converts the edge paths of the shortest path functions.

Series-commit: e375370
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@3b3709c

Date: 2026-04-18 21:31:21 +0200

chore: VERTEX_COLORS and EDGE_COLORS

R-side fix:

Upstream renames the VERTEX_COLOR and EDGE_COLOR types to VERTEX_COLORS and EDGE_COLORS. tools/stimulus/types-RC.yaml, types-RR.yaml and functions-R.yaml follow the rename, so the generated interface again converts the vertex and edge colors of the isomorphism functions.

Series-commit: 0e62ef9
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@8043c0f

Date: 2026-04-18 19:51:41 +0000

nfc: rename ARPACKFUNC to ARPACK_FUNC in Stimulus

R-side fix:

Upstream renames the ARPACKFUNC type to ARPACK_FUNC. tools/stimulus/types-RC.yaml and types-RR.yaml follow the rename; without it, Stimulus generates eigen_matrix_impl() and eigen_matrix_symmetric_impl() without the conversion of their fun argument.

Series-commit: 10c072e
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@4bf87c0

Date: 2026-04-18 19:52:35 +0000

nfc: rename ARPACKOPT to ARPACK_OPTIONS in Stimulus

R-side fix:

Upstream renames the ARPACKOPT type to ARPACK_OPTIONS. tools/stimulus/types-RC.yaml, types-RR.yaml and functions-R.yaml follow the rename, in the overrides of the eigenvector centralities, the HITS scores and the leading eigenvector communities; without it, Stimulus fails on those overrides, which name a type it no longer knows.

Series-commit: 47475e9
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@f21bac8

Date: 2026-04-18 19:52:53 +0000

nfc: rename ARPACKSTORAGE to ARPACK_STORAGE in Stimulus

R-side fix:

Upstream renames the ARPACKSTORAGE type to ARPACK_STORAGE. tools/stimulus/types-RC.yaml and types-RR.yaml follow the rename; without it, Stimulus generates eigen_adjacency_impl(), eigen_matrix_impl() and eigen_matrix_symmetric_impl() without the conversion of their storage argument, and the glue does not compile.

Series-commit: 856c3b0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@8edd11b

Date: 2026-04-18 19:53:19 +0000

nfc: rename EDGE_CAPACITY to EDGE_CAPACITIES in Stimulus

R-side fix:

Upstream renames the EDGE_CAPACITY type to EDGE_CAPACITIES. tools/stimulus/types-RC.yaml and types-RR.yaml follow the rename; without it, Stimulus generates the flow and cut functions, from maxflow_impl() to gomory_hu_tree_impl(), without the conversion of their capacities, and the glue does not compile.

Series-commit: b95d1c6
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@afeb097

Date: 2026-04-18 19:55:39 +0000

nfc: rename MST_ALGORITHM to MSTALGORITHM and relocate LPA_VARIANT

R-side fix:

Upstream renames the MST_ALGORITHM type to MSTALGORITHM. tools/stimulus/types-RR.yaml follows the rename; without it, Stimulus generates minimum_spanning_tree_impl() without the mapping of its method argument and with the undefined default AUTOMATIC.

Series-commit: 0c05bf4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@57535cf

Date: 2026-04-18 20:06:21 +0000

feat: expose igraph_community_leiden_simple via Stimulus
nfc: port Stimulus type renames and additions from next

R-side fix: upstream adds igraph_community_leiden_simple(), whose objective is an igraph_leiden_objective_t that tools/stimulus/types-RR.yaml did not map, so community_leiden_simple_impl(objective = "modularity") coerced the string to NA. A new LEIDEN_OBJECTIVE mapping takes "modularity", "cpm" and "er"; a test in test-community.R checks it.

Series-commit: fe01943
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@610a8ab

Date: 2026-04-18 20:11:31 +0000

feat: add weight argument to igraph_assortativity[_nominal]

R-side fix: upstream igraph_assortativity() and igraph_assortativity_nominal() take optional edge weights, which their impls take from the weight attribute, so assortativity() of a weighted graph became weighted and assortativity_nominal() failed on any weighted graph with "Weighted nominal assortativity is not yet implemented". assortativity(), assortativity_legacy() and assortativity_nominal() pass weights = NA; a test checks this. The roxygenized "Related documentation in the C library" of assortativity() and assortativity.nominal() follows the impls, which now read the weight attribute.

Series-commit: e65f137
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@a73d82c

Date: 2026-04-18 20:15:29 +0000

nfc: rename igraph_girth output parameter circle -> cycle

R-side fix: upstream renames the circle result of igraph_girth() to cycle; PARAM_NAMES in tools/stimulus/functions-R.yaml keeps girth_impl()'s $circle.

Series-commit: 2910604
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@82e0439

Date: 2026-04-18 20:21:17 +0000

nfc: rename igraph_biadjacency Stimulus parameter incidence -> biadjmatrix

R-side fix: upstream renames the incidence argument of igraph_biadjacency() to biadjmatrix, so graph_from_biadjacency_matrix() failed on every dense unweighted matrix with an unused argument; PARAM_NAMES in tools/stimulus/functions-R.yaml keeps biadjacency_impl()'s incidence.

Series-commit: bffe6b9
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@112bba9

Date: 2026-04-18 20:21:52 +0000

nfc: reorder igraph_betweenness_subset Stimulus parameters

Series-commit: f8d4a88
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@6065302

Date: 2026-04-18 20:23:59 +0000

nfc: align igraph_community_infomap Stimulus defaults with C API

Series-commit: 9061d21
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@01f2049

Date: 2026-04-18 21:21:53 +0000

nfc: rename __BEGIN_DECLS/__END_DECLS to IGRAPH_BEGIN_C_DECLS/IGRAPH_END_C_DECLS

R-side fix:

Upstream renames the __BEGIN_DECLS and __END_DECLS macros of igraph_decls.h to IGRAPH_BEGIN_C_DECLS and IGRAPH_END_C_DECLS. src/vendor/igraph_config.h, rigraph's own copy of the configured header, follows as upstream's igraph_config.h.in does; without it, the C++ glue that includes it does not compile.

Series-commit: 3c086b1
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@d69d60d

Date: 2026-04-18 21:48:19 +0000

refactor: replace IGRAPH_I_ATTRIBUTE_DESTROY/COPY macros with igraph_i_attribute_copy()

R-side fix: upstream drops the IGRAPH_I_ATTRIBUTE_DESTROY() macro of src/graph/attributes.h, which src/rinterface.h maps the generated interface's IGRAPH_I_DESTROY() to. IGRAPH_I_DESTROY() now releases a graph's attributes with igraph_i_attribute_destroy() itself, as the macro did; without it, the glue still passes the syntax-only gate, with an implicit declaration, but the package does not link.

Series-commit: 1200d08
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@54ebcd6

Date: 2026-04-18 21:49:25 +0000

nfc: drop remaining vim modelines (triangles, plfit.h)

Snapshot: upstream drops the vim modeline at the top of properties/triangles.c, which moves the warning that Barrat's transitivity of an unweighted graph falls back to the unweighted version from line 890 to line 889; the transitivity_barrat_impl() snapshots follow.

Series-commit: 95cf99b
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@c6f0863

Date: 2026-04-18 22:28:13 +0000

nfc: branch rename, infomap summary, and a batch of file cleanups

Snapshot: upstream drops the vim modeline at the top of centrality/hub_authority.c, which moves the warning that the hub and authority scores of an undirected graph are its eigenvector centralities from line 295 to line 294; the hub_and_authority_scores_impl() snapshots follow.

Series-commit: 3d43f88
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@3a73fdd

Date: 2026-04-18 23:29:26 +0000

nfc: style + typo cleanups in community_leading_eigenvector

Snapshot: upstream corrects "memberhsip" to "membership" in the error igraph_community_leading_eigenvector() raises for a start membership vector of the wrong length; the community_leading_eigenvector_callback_closure_impl() error snapshot follows.

Series-commit: 373842d
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@28547aa

Date: 2026-04-19 09:23:53 +0000

refactor: biadjacency accepts negative matrix elements unless multiple=true
nfc: sync leading_eigenvector example and tutorial2/3 with next
nfc: switch igraph_sparsemat3 example locals to igraph_int_t
nfc: add tests/regression/bug_2608.c (companion for existing .out)
nfc: rename comment above igraph_rng_sample_* block to "sampling"
nfc: add spanning_tree benchmark
nfc: drop empty "# setup" section header from tests/CMakeLists.txt
nfc: reorder igraph_*biadjacency tests alphabetically
nfc: add max_results stress test to the cliques legacy-test list
nfc: reorder minimum_spanning_tree legacy test entry
nfc: document why eigen_stress is disabled by default
nfc: drop the igraph_arpack_error legacy unit test

Snapshot: upstream igraph_biadjacency() now walks the matrix column by column, so graph_from_biadjacency_matrix() of a dense unweighted matrix, with or without multiple = TRUE, creates the same edges in column-major order, the order the weighted and sparse paths already used; the two dense snapshots follow.

Series-commit: 4a513e0
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@cea0af7

Date: 2026-04-19 09:39:04 +0000

nfc: reformat long motifs calls and add graph/internal.h include

Snapshot: upstream adds an include and wraps the long calls in misc/motifs.c over more lines, which moves the triad census warning on undirected graphs from line 1155 to line 1170; the triad_census_impl() snapshot follows.

Series-commit: b66af9d
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk
vendor: Update vendored sources to krlmlr/igraph@fb5d4a6

Date: 2026-04-19 09:49:48 +0000

nfc: store strvector elements as const char *

R-side fix: upstream stores the elements of igraph_strvector_t as const char * rather than char *. Rx_igraph_SEXP_to_strvector() in src/rinterface_extra.c, which points the vector at R's own strings, follows; without it, the assignment is an incompatible-pointer warning in the syntax-only glue gate, and an error under GCC 14 and later.

Series-commit: a1aa5b2
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01JXnA76zWTu4ZD8BQCSYRpk

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants