1 -----------------------------------------------------------------------------
4 -- Copyright : (c) The University of Glasgow 2002
5 -- License : BSD-style (see the file libraries/base/LICENSE)
7 -- Maintainer : libraries@haskell.org
8 -- Stability : experimental
9 -- Portability : non-portable (requires non-portable module ST)
11 -- A version of the graph algorithms described in:
13 -- /Lazy Depth-First Search and Linear Graph Algorithms in Haskell/,
14 -- by David King and John Launchbury.
16 -----------------------------------------------------------------------------
20 -- * External interface
22 -- At present the only one with a "nice" external interface
23 stronglyConnComp, stronglyConnCompR, SCC(..), flattenSCC, flattenSCCs,
27 Graph, Table, Bounds, Edge, Vertex,
31 graphFromEdges, buildG, transposeG,
34 -- ** Graph properties
46 -- tree, back, cross, forward,
54 import Control.Monad.ST
55 import Data.Array.ST (STArray, newArray, readArray, writeArray)
56 import Data.Tree (Tree(Node), Forest)
67 -------------------------------------------------------------------------
71 -------------------------------------------------------------------------
73 -- | Strongly connected component.
74 data SCC vertex = AcyclicSCC vertex -- ^ A single vertex that is not
76 | CyclicSCC [vertex] -- ^ A maximal set of mutually
77 -- reachable vertices.
79 -- | The vertices of a list of strongly connected components.
80 flattenSCCs :: [SCC a] -> [a]
81 flattenSCCs = concatMap flattenSCC
83 -- | The vertices of a strongly connected component.
84 flattenSCC :: SCC vertex -> [vertex]
85 flattenSCC (AcyclicSCC v) = [v]
86 flattenSCC (CyclicSCC vs) = vs
88 -- | The strongly connected components of a directed graph, topologically
92 => [(node, key, [key])]
93 -- ^ The graph: a list of nodes uniquely identified by keys,
94 -- with a list of keys of nodes this node has edges to.
95 -- The out-list may contain keys that don't correspond to
96 -- nodes of the graph; such edges are ignored.
99 stronglyConnComp edges0
100 = map get_node (stronglyConnCompR edges0)
102 get_node (AcyclicSCC (n, _, _)) = AcyclicSCC n
103 get_node (CyclicSCC triples) = CyclicSCC [n | (n,_,_) <- triples]
105 -- | The strongly connected components of a directed graph, topologically
106 -- sorted. The function is the same as 'stronglyConnComp', except that
107 -- all the information about each node retained.
108 -- This interface is used when you expect to apply 'SCC' to
109 -- (some of) the result of 'SCC', so you don't want to lose the
110 -- dependency information.
113 => [(node, key, [key])]
114 -- ^ The graph: a list of nodes uniquely identified by keys,
115 -- with a list of keys of nodes this node has edges to.
116 -- The out-list may contain keys that don't correspond to
117 -- nodes of the graph; such edges are ignored.
118 -> [SCC (node, key, [key])] -- ^ Topologically sorted
120 stronglyConnCompR [] = [] -- added to avoid creating empty array in graphFromEdges -- SOF
121 stronglyConnCompR edges0
124 (graph, vertex_fn) = graphFromEdges edges0
126 decode (Node v []) | mentions_itself v = CyclicSCC [vertex_fn v]
127 | otherwise = AcyclicSCC (vertex_fn v)
128 decode other = CyclicSCC (dec other [])
130 dec (Node v ts) vs = vertex_fn v : foldr dec vs ts
131 mentions_itself v = v `elem` (graph ! v)
133 -------------------------------------------------------------------------
137 -------------------------------------------------------------------------
139 -- | Abstract representation of vertices.
141 -- | Table indexed by a contiguous set of vertices.
142 type Table a = Array Vertex a
143 -- | Adjacency list representation of a graph, mapping each vertex to its
144 -- list of successors.
145 type Graph = Table [Vertex]
146 -- | The bounds of a 'Table'.
147 type Bounds = (Vertex, Vertex)
148 -- | An edge from the first vertex to the second.
149 type Edge = (Vertex, Vertex)
151 -- | All vertices of a graph.
152 vertices :: Graph -> [Vertex]
155 -- | All edges of a graph.
156 edges :: Graph -> [Edge]
157 edges g = [ (v, w) | v <- vertices g, w <- g!v ]
159 mapT :: (Vertex -> a -> b) -> Table a -> Table b
160 mapT f t = array (bounds t) [ (,) v (f v (t!v)) | v <- indices t ]
162 -- | Build a graph from a list of edges.
163 buildG :: Bounds -> [Edge] -> Graph
164 buildG bounds0 edges0 = accumArray (flip (:)) [] bounds0 edges0
166 -- | The graph obtained by reversing all edges.
167 transposeG :: Graph -> Graph
168 transposeG g = buildG (bounds g) (reverseE g)
170 reverseE :: Graph -> [Edge]
171 reverseE g = [ (w, v) | (v, w) <- edges g ]
173 -- | A table of the count of edges from each node.
174 outdegree :: Graph -> Table Int
175 outdegree = mapT numEdges
176 where numEdges _ ws = length ws
178 -- | A table of the count of edges into each node.
179 indegree :: Graph -> Table Int
180 indegree = outdegree . transposeG
182 -- | Build a graph from a list of nodes uniquely identified by keys,
183 -- with a list of keys of nodes this node should have edges to.
184 -- The out-list may contain keys that don't correspond to
185 -- nodes of the graph; they are ignored.
188 => [(node, key, [key])]
189 -> (Graph, Vertex -> (node, key, [key]))
190 graphFromEdges edges0
191 = (graph, \v -> vertex_map ! v)
193 max_v = length edges0 - 1
194 bounds0 = (0,max_v) :: (Vertex, Vertex)
195 sorted_edges = sortBy lt edges0
196 edges1 = zipWith (,) [0..] sorted_edges
198 graph = array bounds0 [(,) v (mapMaybe key_vertex ks) | (,) v (_, _, ks) <- edges1]
199 key_map = array bounds0 [(,) v k | (,) v (_, k, _ ) <- edges1]
200 vertex_map = array bounds0 edges1
202 (_,k1,_) `lt` (_,k2,_) = k1 `compare` k2
204 -- key_vertex :: key -> Maybe Vertex
205 -- returns Nothing for non-interesting vertices
206 key_vertex k = findVertex 0 max_v
208 findVertex a b | a > b
210 findVertex a b = case compare k (key_map ! mid) of
211 LT -> findVertex a (mid-1)
213 GT -> findVertex (mid+1) b
215 mid = (a + b) `div` 2
217 -------------------------------------------------------------------------
219 -- Depth first search
221 -------------------------------------------------------------------------
223 type Set s = STArray s Vertex Bool
225 mkEmpty :: Bounds -> ST s (Set s)
226 mkEmpty bnds = newArray bnds False
228 contains :: Set s -> Vertex -> ST s Bool
229 contains m v = readArray m v
231 include :: Set s -> Vertex -> ST s ()
232 include m v = writeArray m v True
234 -- | A spanning forest of the graph, obtained from a depth-first search of
235 -- the graph starting from each vertex in an unspecified order.
236 dff :: Graph -> Forest Vertex
237 dff g = dfs g (vertices g)
239 -- | A spanning forest of the part of the graph reachable from the listed
240 -- vertices, obtained from a depth-first search of the graph starting at
241 -- each of the listed vertices in order.
242 dfs :: Graph -> [Vertex] -> Forest Vertex
243 dfs g vs = prune (bounds g) (map (generate g) vs)
245 generate :: Graph -> Vertex -> Tree Vertex
246 generate g v = Node v (map (generate g) (g!v))
248 prune :: Bounds -> Forest Vertex -> Forest Vertex
249 prune bnds ts = runST (mkEmpty bnds >>= \m ->
252 chop :: Set s -> Forest Vertex -> ST s (Forest Vertex)
253 chop _ [] = return []
254 chop m (Node v ts : us)
255 = contains m v >>= \visited ->
259 include m v >>= \_ ->
262 return (Node v as : bs)
264 -------------------------------------------------------------------------
268 -------------------------------------------------------------------------
270 ------------------------------------------------------------
271 -- Algorithm 1: depth first search numbering
272 ------------------------------------------------------------
274 preorder :: Tree a -> [a]
275 preorder (Node a ts) = a : preorderF ts
277 preorderF :: Forest a -> [a]
278 preorderF ts = concat (map preorder ts)
280 tabulate :: Bounds -> [Vertex] -> Table Int
281 tabulate bnds vs = array bnds (zipWith (,) vs [1..])
283 preArr :: Bounds -> Forest Vertex -> Table Int
284 preArr bnds = tabulate bnds . preorderF
286 ------------------------------------------------------------
287 -- Algorithm 2: topological sorting
288 ------------------------------------------------------------
290 postorder :: Tree a -> [a]
291 postorder (Node a ts) = postorderF ts ++ [a]
293 postorderF :: Forest a -> [a]
294 postorderF ts = concat (map postorder ts)
296 postOrd :: Graph -> [Vertex]
297 postOrd = postorderF . dff
299 -- | A topological sort of the graph.
300 -- The order is partially specified by the condition that a vertex /i/
301 -- precedes /j/ whenever /j/ is reachable from /i/ but not vice versa.
302 topSort :: Graph -> [Vertex]
303 topSort = reverse . postOrd
305 ------------------------------------------------------------
306 -- Algorithm 3: connected components
307 ------------------------------------------------------------
309 -- | The connected components of a graph.
310 -- Two vertices are connected if there is a path between them, traversing
311 -- edges in either direction.
312 components :: Graph -> Forest Vertex
313 components = dff . undirected
315 undirected :: Graph -> Graph
316 undirected g = buildG (bounds g) (edges g ++ reverseE g)
318 -- Algorithm 4: strongly connected components
320 -- | The strongly connected components of a graph.
321 scc :: Graph -> Forest Vertex
322 scc g = dfs g (reverse (postOrd (transposeG g)))
324 ------------------------------------------------------------
325 -- Algorithm 5: Classifying edges
326 ------------------------------------------------------------
328 tree :: Bounds -> Forest Vertex -> Graph
329 tree bnds ts = buildG bnds (concat (map flat ts))
330 where flat (Node v ts) = [ (v, w) | Node w _us <- ts ] ++ concat (map flat ts)
332 back :: Graph -> Table Int -> Graph
333 back g post = mapT select g
334 where select v ws = [ w | w <- ws, post!v < post!w ]
336 cross :: Graph -> Table Int -> Table Int -> Graph
337 cross g pre post = mapT select g
338 where select v ws = [ w | w <- ws, post!v > post!w, pre!v > pre!w ]
340 forward :: Graph -> Graph -> Table Int -> Graph
341 forward g tree pre = mapT select g
342 where select v ws = [ w | w <- ws, pre!v < pre!w ] \\ tree!v
344 ------------------------------------------------------------
345 -- Algorithm 6: Finding reachable vertices
346 ------------------------------------------------------------
348 -- | A list of vertices reachable from a given vertex.
349 reachable :: Graph -> Vertex -> [Vertex]
350 reachable g v = preorderF (dfs g [v])
352 -- | Is the second vertex reachable from the first?
353 path :: Graph -> Vertex -> Vertex -> Bool
354 path g v w = w `elem` (reachable g v)
356 ------------------------------------------------------------
357 -- Algorithm 7: Biconnected components
358 ------------------------------------------------------------
360 -- | The biconnected components of a graph.
361 -- An undirected graph is biconnected if the deletion of any vertex
362 -- leaves it connected.
363 bcc :: Graph -> Forest [Vertex]
364 bcc g = (concat . map bicomps . map (do_label g dnum)) forest
366 dnum = preArr (bounds g) forest
368 do_label :: Graph -> Table Int -> Tree Vertex -> Tree (Vertex,Int,Int)
369 do_label g dnum (Node v ts) = Node (v,dnum!v,lv) us
370 where us = map (do_label g dnum) ts
371 lv = minimum ([dnum!v] ++ [dnum!w | w <- g!v]
372 ++ [lu | Node (u,du,lu) xs <- us])
374 bicomps :: Tree (Vertex,Int,Int) -> Forest [Vertex]
375 bicomps (Node (v,_,_) ts)
376 = [ Node (v:vs) us | (l,Node vs us) <- map collect ts]
378 collect :: Tree (Vertex,Int,Int) -> (Int, Tree [Vertex])
379 collect (Node (v,dv,lv) ts) = (lv, Node (v:vs) cs)
380 where collected = map collect ts
381 vs = concat [ ws | (lw, Node ws us) <- collected, lw<dv]
382 cs = concat [ if lw<dv then us else [Node (v:ws) us]
383 | (lw, Node ws us) <- collected ]