Merge remote-tracking branch 'origin/main' into tutorials

Author: gdalle

Project.toml

@@ -1,26 +1,27 @@
 name = "SparseMatrixColorings"
 uuid = "0a514795-09f3-496d-8182-132a7b665d35"
 authors = ["Guillaume Dalle", "Alexis Montoison"]
-version = "0.4.13"
+version = "0.4.19"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
-DataStructures = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
 DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 
 [weakdeps]
+CliqueTrees = "60701a23-6482-424a-84db-faee86b9b1f8"
 Colors = "5ae59095-9a9b-59fe-a467-6f913c188581"
 
 [extensions]
+SparseMatrixColoringsCliqueTreesExt = "CliqueTrees"
 SparseMatrixColoringsColorsExt = "Colors"
 
 [compat]
 ADTypes = "1.2.1"
+CliqueTrees = "1"
 Colors = "0.12.11, 0.13"
-DataStructures = "0.18"
 DocStringExtensions = "0.8,0.9"
 LinearAlgebra = "<0.0.1, 1"
 Random = "<0.0.1, 1"

benchmark/benchmarks.jl

@@ -1,6 +1,7 @@
 using BenchmarkTools
 using LinearAlgebra
 using SparseMatrixColorings
+import SparseMatrixColorings as SMC
 using SparseArrays
 using StableRNGs
 
@@ -19,21 +20,23 @@ for structure in [:nonsymmetric, :symmetric],
     p in [2 / n, 5 / n, 10 / n]
 
     problem = ColoringProblem(; structure, partition)
-    algo = GreedyColoringAlgorithm(; decompression, postprocessing=true)
+    algo = GreedyColoringAlgorithm(
+        RandomOrder(StableRNG(0), 0); decompression, postprocessing=true
+    )
 
     # use several random matrices to reduce variance
     nb_samples = 5
     As = [sparse(Symmetric(sprand(StableRNG(i), Bool, n, n, p))) for i in 1:nb_samples]
-    results = [coloring(A, problem, algo) for A in As]
-    Bs = [compress(A, result) for (A, result) in zip(As, results)]
+    results = [coloring(A, problem, algo; decompression_eltype=Float64) for A in As]
+    Bs = [compress(Float64.(A), result) for (A, result) in zip(As, results)]
 
-    SUITE[:coloring][structure][partition][decompression]["n=$n"]["p=$p"] = @benchmarkable begin
+    bench_col = @benchmarkable begin
         for A in $As
             coloring(A, $problem, $algo)
         end
     end
 
-    SUITE[:decompress][structure][partition][decompression]["n=$n"]["p=$p"] = @benchmarkable begin
+    bench_dec = @benchmarkable begin
         for (B, result) in zip($Bs, $results)
             if B isa AbstractMatrix
                 decompress(B, result)
@@ -42,4 +45,34 @@ for structure in [:nonsymmetric, :symmetric],
             end
         end
     end
+
+    SUITE[:coloring][structure][partition][decompression]["n=$n"]["p=$p"] = bench_col
+    SUITE[:decompress][structure][partition][decompression]["n=$n"]["p=$p"] = bench_dec
+end
+
+for structure in [:nonsymmetric, :symmetric],
+    partition in (structure == :nonsymmetric ? [:column, :row] : [:column]),
+    order in [LargestFirst(), SmallestLast(), IncidenceDegree(), DynamicLargestFirst()],
+    n in [10^3, 10^5],
+    p in [2 / n, 5 / n, 10 / n]
+
+    nb_samples = 5
+    As = [sparse(Symmetric(sprand(StableRNG(i), Bool, n, n, p))) for i in 1:nb_samples]
+    if structure == :symmetric
+        gs = [SMC.AdjacencyGraph(A) for A in As]
+        bench_ord = @benchmarkable begin
+            for g in $gs
+                SMC.vertices(g, $order)
+            end
+        end
+    else
+        gs = [SMC.BipartiteGraph(A) for A in As]
+        valside = partition == :row ? Val(1) : Val(2)
+        bench_ord = @benchmarkable begin
+            for g in $gs
+                SMC.vertices(g, $valside, $order)
+            end
+        end
+    end
+    SUITE[:order][structure][partition][string(order)]["n=$n"]["p=$p"] = bench_ord
 end

docs/Project.toml

@@ -7,3 +7,6 @@ Images = "916415d5-f1e6-5110-898d-aaa5f9f070e0"
 MatrixMarket = "4d4711f2-db25-561a-b6b3-d35e7d4047d3"
 SparseMatrixColorings = "0a514795-09f3-496d-8182-132a7b665d35"
 StableRNGs = "860ef19b-820b-49d6-a774-d7a799459cd3"
+
+[sources]
+SparseMatrixColorings = {path=".."}

docs/src/api.md

@@ -15,6 +15,7 @@ SparseMatrixColorings
 
 ```@docs
 coloring
+fast_coloring
 ColoringProblem
 GreedyColoringAlgorithm
 ConstantColoringAlgorithm
@@ -52,4 +53,5 @@ SmallestLast
 IncidenceDegree
 DynamicLargestFirst
 DynamicDegreeBasedOrder
+PerfectEliminationOrder
 ```

docs/src/dev.md

@@ -16,16 +16,17 @@ SparseMatrixColorings.BipartiteGraph
 SparseMatrixColorings.vertices
 SparseMatrixColorings.neighbors
 transpose
+SparseMatrixColorings.bidirectional_pattern
 ```
 
 ## Low-level coloring
 
 ```@docs
 SparseMatrixColorings.partial_distance2_coloring
-SparseMatrixColorings.symmetric_coefficient
 SparseMatrixColorings.star_coloring
 SparseMatrixColorings.acyclic_coloring
 SparseMatrixColorings.group_by_color
+SparseMatrixColorings.Forest
 SparseMatrixColorings.StarSet
 SparseMatrixColorings.TreeSet
 ```

docs/src/vis.md

@@ -74,22 +74,33 @@ problem_bi = ColoringProblem(; structure=:nonsymmetric, partition=:bidirectional
 algo_bi = GreedyColoringAlgorithm(RandomOrder(StableRNG(0)); postprocessing=true, decompression=:direct)
 result_bi = coloring(S, problem_bi, algo_bi)
 
-A_img, Br_img, Bc_img = show_colors(
+Arc_img, Ar_img, Ac_img, Br_img, Bc_img = show_colors(
     result_bi;
     colorscheme=ColorSchemes.progress,
-    background=RGB(1, 1, 1),  # white
+    background_color=RGB(1, 1, 1),  # white
     scale=10,
+    border=1,
     pad=2
 )
 ```
 
 In the bidirectional case, columns and rows can both get colors:
 
 ```@example img
-A_img
+Ar_img
+```
+
+```@example img
+Ac_img
+```
+
+Together, this yields:
+
+```@example img
+Arc_img
 ```
 
-And there are two compression results, one by row and one by column:
+And there are two associated compression results, one by row and one by column:
 
 ```@example img
 Br_img

ext/SparseMatrixColoringsCliqueTreesExt.jl

@@ -0,0 +1,23 @@
+module SparseMatrixColoringsCliqueTreesExt
+
+import CliqueTrees: permutation, EliminationAlgorithm, MCS
+import SparseArrays: SparseMatrixCSC, rowvals, nnz
+import SparseMatrixColorings:
+    AdjacencyGraph, BipartiteGraph, PerfectEliminationOrder, pattern, vertices
+
+PerfectEliminationOrder() = PerfectEliminationOrder(MCS())
+
+function vertices(g::AdjacencyGraph{T}, order::PerfectEliminationOrder) where {T}
+    S = pattern(g)
+
+    # construct matrix with sparsity pattern S
+    M = SparseMatrixCSC{Bool,T}(size(S)..., S.colptr, rowvals(S), ones(Bool, nnz(S)))
+
+    # construct a perfect elimination order
+    # self-loops are ignored
+    order, _ = permutation(M; alg=order.elimination_algorithm)
+
+    return reverse!(order)
+end
+
+end # module