src/Gargantext/Core/Viz/Graph/Tools.hs

   1 {-|
   2 Module      : Gargantext.Core.Viz.Graph.Tools
   3 Description : Tools to build Graph
   4 Copyright   : (c) CNRS, 2017-Present
   5 License     : AGPL + CECILL v3
   6 Maintainer  : team@gargantext.org
   7 Stability   : experimental
   8 Portability : POSIX
   9
  10 -}
  11
  12 {-# LANGUAGE BangPatterns, ScopedTypeVariables #-}
  13
  14 module Gargantext.Core.Viz.Graph.Tools
  15   where
  16
  17 import Data.Aeson
  18 import Data.HashMap.Strict (HashMap)
  19 import Data.Map.Strict (Map)
  20 import Data.Maybe (fromMaybe)
  21 import Data.Swagger hiding (items)
  22 import GHC.Float (sin, cos)
  23 import GHC.Generics (Generic)
  24 import Gargantext.API.Ngrams.Types (NgramsTerm(..))
  25 import Gargantext.Core.Methods.Similarities (Similarity(..), measure)
  26 -- import Gargantext.Core.Methods.Similarities.Conditional (conditional)
  27 import Gargantext.Core.Statistics
  28 import Gargantext.Core.Viz.Graph.Bridgeness (bridgeness, Bridgeness(..), Partitions, nodeId2comId, {-recursiveClustering,-} recursiveClustering', setNodes2clusterNodes)
  29 import Gargantext.Core.Viz.Graph.Index (createIndices, toIndex, map2mat, mat2map, Index, MatrixShape(..))
  30 import Gargantext.Core.Viz.Graph.Tools.IGraph (mkGraphUfromEdges, spinglass, spinglass')
  31 import Gargantext.Core.Viz.Graph.Tools.Infomap (infomap)
  32 import Gargantext.Core.Viz.Graph.Types (Attributes(..), Edge(..), Graph(..), MultiPartite(..), Node(..), Partite(..), Strength(..))
  33 import Gargantext.Database.Schema.Ngrams (NgramsType(..))
  34 import Gargantext.Core.Viz.Graph.Utils (edgesFilter, nodesFilter)
  35 import Gargantext.Prelude
  36 import Graph.Types (ClusterNode)
  37 import IGraph.Random -- (Gen(..))
  38 import Test.QuickCheck (elements)
  39 import Test.QuickCheck.Arbitrary
  40 import qualified Data.HashMap.Strict      as HashMap
  41 import qualified Data.List                as List
  42 import qualified Data.Map.Strict          as Map
  43 import qualified Data.Set                 as Set
  44 import qualified Data.HashSet             as HashSet
  45 import qualified Data.Text                as Text
  46 import qualified Data.Vector.Storable     as Vec
  47 import qualified Graph.BAC.ProxemyOptim   as BAC
  48 import qualified IGraph                   as Igraph
  49 import qualified IGraph.Algorithms.Layout as Layout
  50
  51 data PartitionMethod = Spinglass | Confluence | Infomap
  52     deriving (Generic, Eq, Ord, Enum, Bounded, Show)
  53 instance FromJSON  PartitionMethod
  54 instance ToJSON    PartitionMethod
  55 instance ToSchema  PartitionMethod
  56 instance Arbitrary PartitionMethod where
  57   arbitrary = elements [ minBound .. maxBound ]
  58
  59 data BridgenessMethod = BridgenessMethod_Basic | BridgenessMethod_Advanced
  60     deriving (Generic, Eq, Ord, Enum, Bounded, Show)
  61 instance FromJSON  BridgenessMethod
  62 instance ToJSON    BridgenessMethod
  63 instance ToSchema  BridgenessMethod
  64 instance Arbitrary BridgenessMethod where
  65   arbitrary = elements [ minBound .. maxBound ]
  66
  67
  68 -------------------------------------------------------------
  69 defaultClustering :: Map (Int, Int) Double -> IO [ClusterNode]
  70 -- defaultClustering x = pure $ BAC.defaultClustering x
  71 defaultClustering x = spinglass 1 x
  72
  73 -------------------------------------------------------------
  74 type Threshold = Double
  75
  76
  77 cooc2graph' :: Ord t => Similarity
  78                      -> Double
  79                      -> Map (t, t) Int
  80                      -> Map (Index, Index) Double
  81 cooc2graph' distance threshold myCooc
  82     = Map.filter (> threshold)
  83     $ mat2map
  84     $ measure distance
  85     $ case distance of
  86         Conditional    -> map2mat Triangle 0 tiSize
  87         Distributional -> map2mat Square   0 tiSize
  88     $ Map.filter (> 1) myCooc'
  89
  90      where
  91         (ti, _) = createIndices myCooc
  92         tiSize  = Map.size ti
  93         myCooc' = toIndex ti myCooc
  94
  95
  96
  97 -- coocurrences graph computation
  98 cooc2graphWith :: PartitionMethod
  99                -> BridgenessMethod
 100                -> MultiPartite
 101                -> Similarity
 102                -> Threshold
 103                -> Strength
 104                -> HashMap (NgramsTerm, NgramsTerm) Int
 105                -> IO Graph
 106 cooc2graphWith Spinglass = cooc2graphWith' (spinglass 1)
 107 cooc2graphWith Confluence= cooc2graphWith' (\x -> pure $ BAC.defaultClustering x)
 108 cooc2graphWith Infomap   = cooc2graphWith' (infomap "-v -N2")
 109 --cooc2graphWith Infomap   = cooc2graphWith' (infomap "--silent --two-level -N2")
 110                         -- TODO: change these options, or make them configurable in UI?
 111
 112 cooc2graphWith' :: Partitions
 113                 -> BridgenessMethod
 114                 -> MultiPartite
 115                 -> Similarity
 116                 -> Threshold
 117                 -> Strength
 118                 -> HashMap (NgramsTerm, NgramsTerm) Int
 119                 -> IO Graph
 120 cooc2graphWith' _doPartitions _bridgenessMethod multi similarity threshold strength myCooc = do
 121   let (distanceMap, diag, ti) = doSimilarityMap similarity threshold strength myCooc
 122   distanceMap `seq` diag `seq` ti `seq` return ()
 123
 124   partitions <- if (Map.size distanceMap > 0)
 125       then recursiveClustering' (spinglass' 1) distanceMap
 126       else panic $ Text.unlines [ "I can not compute the graph you request"
 127                                 , "because either the quantity of documents"
 128                                 , "or the quantity of terms"
 129                                 , "are lacking. "
 130                                 , "Solution: add more either Documents or Map Terms to your analysis. "
 131                                 , "Follow the available tutorials on the Training EcoSystems. "
 132                                 , "Ask your co-users of GarganText how to have access to it."
 133                                 ]
 134   length partitions `seq` return ()
 135
 136   let
 137     !confluence' = BAC.computeConfluences 3 (Map.keys distanceMap) True
 138     !bridgeness' = bridgeness (Bridgeness_Recursive partitions 1.0 similarity) distanceMap
 139
 140   pure $ data2graph multi ti diag bridgeness' confluence' (setNodes2clusterNodes $ List.concat partitions)
 141
 142 {-
 143 cooc2graphWith' _doPartitions _bridgenessMethod multi similarity@Distributional threshold strength myCooc = do
 144   let (distanceMap, diag, ti) = doSimilarityMap similarity threshold strength myCooc
 145   distanceMap `seq` diag `seq` ti `seq` return ()
 146
 147   partitions <- if (Map.size distanceMap > 0)
 148       then recursiveClustering (spinglass 1) distanceMap
 149       else panic $ Text.unlines [ "[Gargantext.C.V.Graph.Tools] Similarity Matrix is empty"
 150                                 , "Maybe you should add more Map Terms in your list"
 151                                 , "Tutorial: TODO"
 152                                 ]
 153   length partitions `seq` return ()
 154
 155   let
 156     !confluence' = BAC.computeConfluences 3 (Map.keys distanceMap) True
 157     !bridgeness' = bridgeness (Bridgeness_Basic partitions 1.0) distanceMap
 158
 159   pure $ data2graph multi ti diag bridgeness' confluence' partitions
 160 -}
 161
 162
 163
 164 type Reverse = Bool
 165
 166 doSimilarityMap :: Similarity
 167               -> Threshold
 168               -> Strength
 169               -> HashMap (NgramsTerm, NgramsTerm) Int
 170               -> ( Map (Int,Int) Double
 171                  , Map (Index, Index) Int
 172                  , Map NgramsTerm Index
 173                  )
 174
 175 doSimilarityMap Conditional threshold strength myCooc = (distanceMap, toIndex ti myCooc', ti)
 176   where
 177     myCooc' = Map.fromList $ HashMap.toList myCooc
 178
 179     (_diag, theMatrix) = Map.partitionWithKey (\(x,y) _ -> x == y)
 180                        $ Map.fromList
 181                        $ HashMap.toList myCooc
 182
 183     (ti, _it) = createIndices theMatrix
 184     tiSize  = Map.size ti
 185
 186     similarities = (\m -> m `seq` m)
 187                  $ (\m -> m `seq` measure Conditional m)
 188                  $ (\m -> m `seq` map2mat Square 0 tiSize m)
 189                  $ theMatrix `seq` toIndex ti theMatrix
 190
 191     links = round (let n :: Double = fromIntegral (Map.size ti) in 10 * n * (log n)^(2::Int))
 192     distanceMap = Map.fromList
 193                 $ List.take links
 194                 $ (if strength == Weak then List.reverse else identity)
 195                 $ List.sortOn snd
 196                 $ Map.toList
 197                 $ Map.filter (> threshold)
 198                 $ similarities `seq` mat2map similarities
 199
 200 doSimilarityMap Distributional threshold strength myCooc = (distanceMap, toIndex ti diag, ti)
 201   where
 202     -- TODO remove below
 203     (diag, theMatrix) = Map.partitionWithKey (\(x,y) _ -> x == y)
 204                       $ Map.fromList
 205                       $ HashMap.toList myCooc
 206
 207     (ti, _it) = createIndices theMatrix
 208     tiSize  = Map.size ti
 209
 210     similarities = (\m -> m `seq` m)
 211                  $ (\m -> m `seq` measure Distributional m)
 212                  $ (\m -> m `seq` map2mat Square 0 tiSize m)
 213                  $ theMatrix `seq` toIndex ti theMatrix
 214
 215     links = round (let n :: Double = fromIntegral tiSize in n * (log n)^(2::Int))
 216
 217     distanceMap = Map.fromList
 218                 $ List.take links
 219                 $ (if strength == Weak then List.reverse else identity)
 220                 $ List.sortOn snd
 221                 $ Map.toList
 222                 $ edgesFilter
 223                 $ (\m -> m `seq` Map.filter (> threshold) m)
 224                 $ similarities `seq` mat2map similarities
 225
 226 ----------------------------------------------------------
 227 -- | From data to Graph
 228 type Occurrences      = Int
 229
 230 nodeTypeWith :: MultiPartite -> NgramsTerm -> NgramsType
 231 nodeTypeWith (MultiPartite (Partite s1 t1) (Partite _s2 t2)) t =
 232   if HashSet.member t s1
 233      then t1
 234      else t2
 235
 236 data2graph :: MultiPartite
 237            -> Map NgramsTerm Int
 238            -> Map (Int, Int) Occurrences
 239            -> Map (Int, Int) Double
 240            -> Map (Int, Int) Double
 241            -> [ClusterNode]
 242            -> Graph
 243 data2graph multi labels' occurences bridge conf partitions =
 244   Graph { _graph_nodes = nodes
 245         , _graph_edges = edges
 246         , _graph_metadata = Nothing
 247         }
 248
 249    where
 250
 251     nodes = map (setCoord ForceAtlas labels bridge)
 252           [ (n, Node { node_size    = maybe 0 identity (Map.lookup (n,n) occurences)
 253                      , node_type    = nodeTypeWith multi label
 254                      , node_id      = (cs . show) n
 255                      , node_label   = unNgramsTerm label
 256                      , node_x_coord = 0
 257                      , node_y_coord = 0
 258                      , node_attributes =
 259                               Attributes { clust_default = fromMaybe 0
 260                                                            (Map.lookup n community_id_by_node_id)
 261                                          }
 262                      , node_children = []
 263                      }
 264                )
 265             | (label, n) <- labels
 266             , Set.member n toKeep
 267             ]
 268
 269     (bridge', toKeep) = nodesFilter (\v -> v > 1) bridge
 270
 271     edges = [ Edge { edge_source = cs (show s)
 272                    , edge_hidden = Nothing
 273                    , edge_target = cs (show t)
 274                    , edge_weight = weight
 275                    , edge_confluence = maybe 0 identity $ Map.lookup (s,t) conf
 276                    , edge_id     = cs (show i)
 277                    }
 278             | (i, ((s,t), weight)) <- zip ([0..]::[Integer] ) $ Map.toList bridge'
 279             , s /= t
 280             , weight > 0
 281             ]
 282
 283     community_id_by_node_id = Map.fromList
 284                             $ map nodeId2comId partitions
 285
 286     labels = Map.toList labels'
 287
 288
 289 ------------------------------------------------------------------------
 290
 291 data Layout = KamadaKawai | ACP | ForceAtlas
 292
 293
 294 setCoord' :: (Int -> (Double, Double)) -> (Int, Node) -> Node
 295 setCoord' f (i,n) = n { node_x_coord = x, node_y_coord = y }
 296   where
 297     (x,y) = f i
 298
 299
 300 -- | ACP
 301 setCoord :: Ord a => Layout -> [(a, Int)] -> Map (Int, Int) Double -> (Int, Node) -> Node
 302 setCoord l labels m (n,node) = node { node_x_coord = x
 303                                     , node_y_coord = y
 304                                     }
 305   where
 306     (x,y) = getCoord l labels m n
 307
 308
 309 getCoord :: Ord a
 310          => Layout
 311          -> [(a, Int)]
 312          -> Map (Int, Int) Double
 313          -> Int
 314          -> (Double, Double)
 315 getCoord KamadaKawai _ _m _n = undefined -- layout m n
 316
 317 getCoord ForceAtlas _ _ n = (sin d, cos d)
 318   where
 319     d = fromIntegral n
 320
 321 getCoord ACP labels m n = to2d $ maybe (panic "Graph.Tools no coordinate") identity
 322              $ Map.lookup n
 323              $ pcaReduceTo (Dimension 2)
 324              $ mapArray labels m
 325   where
 326     to2d :: Vec.Vector Double -> (Double, Double)
 327     to2d v  = (x',y')
 328       where
 329         ds = take 2 $ Vec.toList v
 330         x'  = head' "to2d" ds
 331         y'  = last' "to2d" ds
 332
 333     mapArray :: Ord a => [(a, Int)] -> Map (Int, Int) Double -> Map Int (Vec.Vector Double)
 334     mapArray items m' = Map.fromList [ toVec n' ns m' | n' <- ns ]
 335       where
 336         ns = map snd items
 337
 338     toVec :: Int -> [Int] -> Map (Int,Int) Double -> (Int, Vec.Vector Double)
 339     toVec n' ns' m'' = (n', Vec.fromList $ map (\n'' -> maybe 0 identity $ Map.lookup (n',n'') m'') ns')
 340 ------------------------------------------------------------------------
 341
 342 -- | KamadaKawai Layout
 343 -- TODO TEST: check labels, nodeId and coordinates
 344 layout :: Map (Int, Int) Double -> Int -> Gen -> (Double, Double)
 345 layout m n gen = maybe (panic "") identity $ Map.lookup n $ coord
 346   where
 347     coord :: (Map Int (Double,Double))
 348     coord = Map.fromList $ List.zip (Igraph.nodes g) $ (Layout.layout g p gen)
 349     --p = Layout.defaultLGL
 350     p = Layout.kamadaKawai
 351     g = mkGraphUfromEdges $ map fst $ List.filter (\e -> snd e > 0) $ Map.toList m
 352
 353 -----------------------------------------------------------------------------
 354 -- MISC Tools
 355 cooc2graph'' :: Ord t => Similarity
 356                       -> Double
 357                       -> Map (t, t) Int
 358                       -> Map (Index, Index) Double
 359 cooc2graph'' distance threshold myCooc = neighbourMap
 360   where
 361     (ti, _) = createIndices myCooc
 362     myCooc' = toIndex ti myCooc
 363     matCooc = map2mat Triangle 0 (Map.size ti) $ Map.filter (> 1) myCooc'
 364     distanceMat  = measure distance matCooc
 365     neighbourMap = filterByNeighbours threshold
 366                  $ mat2map distanceMat
 367
 368 -- Quentin
 369 filterByNeighbours :: Double -> Map (Index, Index) Double -> Map (Index, Index) Double
 370 filterByNeighbours threshold distanceMap = filteredMap
 371   where
 372     indexes :: [Index]
 373     indexes = List.nub $ List.concat $ map (\(idx,idx') -> [idx,idx'] ) $ Map.keys distanceMap
 374     filteredMap :: Map (Index, Index) Double
 375     filteredMap = Map.fromList
 376                 $ List.concat
 377                 $ map (\idx ->
 378                           let selected = List.reverse
 379                                        $ List.sortOn snd
 380                                        $ Map.toList
 381                                        $ Map.filter (> 0)
 382                                        $ Map.filterWithKey (\(from,_) _ -> idx == from) distanceMap
 383                            in List.take (round threshold) selected
 384                       ) indexes