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 (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)
  29 import Gargantext.Core.Viz.Graph.Index (createIndices, toIndex, map2mat, mat2map, Index, MatrixShape(..))
  30 import Gargantext.Core.Viz.Graph.Tools.IGraph (mkGraphUfromEdges, 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                 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
  52 data PartitionMethod = Spinglass | Confluence | Infomap
  53     deriving (Generic, Eq, Ord, Enum, Bounded, Show)
  54 instance FromJSON  PartitionMethod
  55 instance ToJSON    PartitionMethod
  56 instance ToSchema  PartitionMethod
  57 instance Arbitrary PartitionMethod where
  58   arbitrary = elements [ minBound .. maxBound ]
  59
  60 data BridgenessMethod = BridgenessMethod_Basic | BridgenessMethod_Advanced
  61     deriving (Generic, Eq, Ord, Enum, Bounded, Show)
  62 instance FromJSON  BridgenessMethod
  63 instance ToJSON    BridgenessMethod
  64 instance ToSchema  BridgenessMethod
  65 instance Arbitrary BridgenessMethod where
  66   arbitrary = elements [ minBound .. maxBound ]
  67
  68
  69 -------------------------------------------------------------
  70 defaultClustering :: Map (Int, Int) Double -> IO [ClusterNode]
  71 -- defaultClustering x = pure $ BAC.defaultClustering x
  72 defaultClustering x = spinglass 1 x
  73
  74 -------------------------------------------------------------
  75 type Threshold = Double
  76
  77
  78 cooc2graph' :: Ord t => Similarity
  79                      -> Double
  80                      -> Map (t, t) Int
  81                      -> Map (Index, Index) Double
  82 cooc2graph' distance threshold myCooc
  83     = Map.filter (> threshold)
  84     $ mat2map
  85     $ measure distance
  86     $ case distance of
  87         Conditional    -> map2mat Triangle 0 tiSize
  88         Distributional -> map2mat Square   0 tiSize
  89     $ Map.filter (> 1) myCooc'
  90
  91      where
  92         (ti, _) = createIndices myCooc
  93         tiSize  = Map.size ti
  94         myCooc' = toIndex ti myCooc
  95
  96
  97
  98 -- coocurrences graph computation
  99 cooc2graphWith :: PartitionMethod
 100                -> BridgenessMethod
 101                -> MultiPartite
 102                -> Similarity
 103                -> Threshold
 104                -> Strength
 105                -> HashMap (NgramsTerm, NgramsTerm) Int
 106                -> IO Graph
 107 cooc2graphWith Spinglass = cooc2graphWith' (spinglass 1)
 108 cooc2graphWith Confluence= cooc2graphWith' (\x -> pure $ BAC.defaultClustering x)
 109 cooc2graphWith Infomap   = cooc2graphWith' (infomap "-v -N2")
 110 --cooc2graphWith Infomap   = cooc2graphWith' (infomap "--silent --two-level -N2")
 111                         -- TODO: change these options, or make them configurable in UI?
 112
 113
 114 cooc2graphWith' :: Partitions
 115                 -> BridgenessMethod
 116                 -> MultiPartite
 117                 -> Similarity
 118                 -> Threshold
 119                 -> Strength
 120                 -> HashMap (NgramsTerm, NgramsTerm) Int
 121                 -> IO Graph
 122 cooc2graphWith' doPartitions bridgenessMethod multi similarity threshold strength myCooc = do
 123   let (distanceMap, diag, ti) = doSimilarityMap similarity threshold strength myCooc
 124   distanceMap `seq` diag `seq` ti `seq` return ()
 125
 126 --{- -- Debug
 127   -- saveAsFileDebug "/tmp/distanceMap" distanceMap
 128   -- saveAsFileDebug "/tmp/distanceMap.keys" (List.length $ Map.keys distanceMap)
 129   -- printDebug "similarities" similarities
 130 --}
 131
 132   partitions <- if (Map.size distanceMap > 0)
 133       then doPartitions distanceMap
 134       else panic $ Text.unlines [ "[Gargantext.C.V.Graph.Tools] Similarity Matrix is empty"
 135                                 , "Maybe you should add more Map Terms in your list"
 136                                 , "Tutorial: link todo"
 137                                 ]
 138   length partitions `seq` return ()
 139
 140   let
 141     !confluence' = BAC.computeConfluences 3 (Map.keys distanceMap) True
 142     !bridgeness' = if bridgenessMethod == BridgenessMethod_Basic
 143                       then bridgeness (Bridgeness_Basic partitions 1.0) distanceMap
 144                       else bridgeness (Bridgeness_Advanced similarity confluence') distanceMap
 145
 146   pure $ data2graph multi ti diag bridgeness' confluence' partitions
 147
 148 type Reverse = Bool
 149
 150 doSimilarityMap :: Similarity
 151               -> Threshold
 152               -> Strength
 153               -> HashMap (NgramsTerm, NgramsTerm) Int
 154               -> ( Map (Int,Int) Double
 155                  , Map (Index, Index) Int
 156                  , Map NgramsTerm Index
 157                  )
 158 doSimilarityMap Distributional threshold strength myCooc = (distanceMap, toIndex ti diag, ti)
 159   where
 160     -- TODO remove below
 161     (diag, theMatrix) = Map.partitionWithKey (\(x,y) _ -> x == y)
 162                       $ Map.fromList
 163                       $ HashMap.toList myCooc
 164
 165     (ti, _it) = createIndices theMatrix
 166     tiSize  = Map.size ti
 167
 168     similarities = (\m -> m `seq` m)
 169                  $ (\m -> m `seq` measure Distributional m)
 170                  $ (\m -> m `seq` map2mat Square 0 tiSize m)
 171                  $ theMatrix `seq` toIndex ti theMatrix
 172
 173     links = round (let n :: Double = fromIntegral tiSize in n * (log n)^(2::Int))
 174
 175     distanceMap = Map.fromList
 176                 $ List.take links
 177                 $ (if strength == Weak then List.reverse else identity)
 178                 $ List.sortOn snd
 179                 $ Map.toList
 180                 $ edgesFilter
 181                 $ (\m -> m `seq` Map.filter (> threshold) m)
 182                 $ similarities `seq` mat2map similarities
 183
 184 doSimilarityMap Conditional threshold strength myCooc = (distanceMap, toIndex ti myCooc', ti)
 185   where
 186     myCooc' = Map.fromList $ HashMap.toList myCooc
 187     (ti, _it) = createIndices myCooc'
 188     links = round (let n :: Double = fromIntegral (Map.size ti) in n * (log n)^(2::Int))
 189     distanceMap = toIndex ti
 190                 $ Map.fromList
 191                 $ List.take links
 192                 $ (if strength == Weak then List.reverse else identity)
 193                 $ List.sortOn snd
 194                 $ HashMap.toList
 195                 $ HashMap.filter (> threshold)
 196                 $ conditional myCooc
 197
 198 ----------------------------------------------------------
 199 -- | From data to Graph
 200 type Occurrences      = Int
 201
 202 nodeTypeWith :: MultiPartite -> NgramsTerm -> NgramsType
 203 nodeTypeWith (MultiPartite (Partite s1 t1) (Partite _s2 t2)) t =
 204   if HashSet.member t s1
 205      then t1
 206      else t2
 207
 208
 209 data2graph :: MultiPartite
 210            -> Map NgramsTerm Int
 211            -> Map (Int, Int) Occurrences
 212            -> Map (Int, Int) Double
 213            -> Map (Int, Int) Double
 214            -> [ClusterNode]
 215            -> Graph
 216 data2graph multi labels' occurences bridge conf partitions =
 217   Graph { _graph_nodes = nodes
 218         , _graph_edges = edges
 219         , _graph_metadata = Nothing
 220         }
 221
 222    where
 223
 224     nodes = map (setCoord ForceAtlas labels bridge)
 225           [ (n, Node { node_size    = maybe 0 identity (Map.lookup (n,n) occurences)
 226                      , node_type    = nodeTypeWith multi label
 227                      , node_id      = (cs . show) n
 228                      , node_label   = unNgramsTerm label
 229                      , node_x_coord = 0
 230                      , node_y_coord = 0
 231                      , node_attributes =
 232                               Attributes { clust_default = fromMaybe 0
 233                                                            (Map.lookup n community_id_by_node_id)
 234                                          }
 235                      , node_children = []
 236                      }
 237                )
 238             | (label, n) <- labels
 239             , Set.member n toKeep
 240             ]
 241
 242     (bridge', toKeep) = nodesFilter (\v -> v > 1) bridge
 243
 244     edges = [ Edge { edge_source = cs (show s)
 245                    , edge_hidden = Nothing
 246                    , edge_target = cs (show t)
 247                    , edge_weight = weight
 248                    , edge_confluence = maybe 0 identity $ Map.lookup (s,t) conf
 249                    , edge_id     = cs (show i)
 250                    }
 251             | (i, ((s,t), weight)) <- zip ([0..]::[Integer] ) $ Map.toList bridge'
 252             , s /= t
 253             , weight > 0
 254             ]
 255
 256     community_id_by_node_id = Map.fromList
 257                             $ map nodeId2comId partitions
 258
 259     labels = Map.toList labels'
 260
 261
 262 ------------------------------------------------------------------------
 263
 264 data Layout = KamadaKawai | ACP | ForceAtlas
 265
 266
 267 setCoord' :: (Int -> (Double, Double)) -> (Int, Node) -> Node
 268 setCoord' f (i,n) = n { node_x_coord = x, node_y_coord = y }
 269   where
 270     (x,y) = f i
 271
 272
 273 -- | ACP
 274 setCoord :: Ord a => Layout -> [(a, Int)] -> Map (Int, Int) Double -> (Int, Node) -> Node
 275 setCoord l labels m (n,node) = node { node_x_coord = x
 276                                     , node_y_coord = y
 277                                     }
 278   where
 279     (x,y) = getCoord l labels m n
 280
 281
 282 getCoord :: Ord a
 283          => Layout
 284          -> [(a, Int)]
 285          -> Map (Int, Int) Double
 286          -> Int
 287          -> (Double, Double)
 288 getCoord KamadaKawai _ _m _n = undefined -- layout m n
 289
 290 getCoord ForceAtlas _ _ n = (sin d, cos d)
 291   where
 292     d = fromIntegral n
 293
 294 getCoord ACP labels m n = to2d $ maybe (panic "Graph.Tools no coordinate") identity
 295              $ Map.lookup n
 296              $ pcaReduceTo (Dimension 2)
 297              $ mapArray labels m
 298   where
 299     to2d :: Vec.Vector Double -> (Double, Double)
 300     to2d v  = (x',y')
 301       where
 302         ds = take 2 $ Vec.toList v
 303         x'  = head' "to2d" ds
 304         y'  = last' "to2d" ds
 305
 306     mapArray :: Ord a => [(a, Int)] -> Map (Int, Int) Double -> Map Int (Vec.Vector Double)
 307     mapArray items m' = Map.fromList [ toVec n' ns m' | n' <- ns ]
 308       where
 309         ns = map snd items
 310
 311     toVec :: Int -> [Int] -> Map (Int,Int) Double -> (Int, Vec.Vector Double)
 312     toVec n' ns' m'' = (n', Vec.fromList $ map (\n'' -> maybe 0 identity $ Map.lookup (n',n'') m'') ns')
 313 ------------------------------------------------------------------------
 314
 315 -- | KamadaKawai Layout
 316 -- TODO TEST: check labels, nodeId and coordinates
 317 layout :: Map (Int, Int) Double -> Int -> Gen -> (Double, Double)
 318 layout m n gen = maybe (panic "") identity $ Map.lookup n $ coord
 319   where
 320     coord :: (Map Int (Double,Double))
 321     coord = Map.fromList $ List.zip (Igraph.nodes g) $ (Layout.layout g p gen)
 322     --p = Layout.defaultLGL
 323     p = Layout.kamadaKawai
 324     g = mkGraphUfromEdges $ map fst $ List.filter (\e -> snd e > 0) $ Map.toList m
 325
 326 -----------------------------------------------------------------------------
 327 -- MISC Tools
 328 cooc2graph'' :: Ord t => Similarity
 329                       -> Double
 330                       -> Map (t, t) Int
 331                       -> Map (Index, Index) Double
 332 cooc2graph'' distance threshold myCooc = neighbourMap
 333   where
 334     (ti, _) = createIndices myCooc
 335     myCooc' = toIndex ti myCooc
 336     matCooc = map2mat Triangle 0 (Map.size ti) $ Map.filter (> 1) myCooc'
 337     distanceMat  = measure distance matCooc
 338     neighbourMap = filterByNeighbours threshold
 339                  $ mat2map distanceMat
 340
 341 -- Quentin
 342 filterByNeighbours :: Double -> Map (Index, Index) Double -> Map (Index, Index) Double
 343 filterByNeighbours threshold distanceMap = filteredMap
 344   where
 345     indexes :: [Index]
 346     indexes = List.nub $ List.concat $ map (\(idx,idx') -> [idx,idx'] ) $ Map.keys distanceMap
 347     filteredMap :: Map (Index, Index) Double
 348     filteredMap = Map.fromList
 349                 $ List.concat
 350                 $ map (\idx ->
 351                           let selected = List.reverse
 352                                        $ List.sortOn snd
 353                                        $ Map.toList
 354                                        $ Map.filter (> 0)
 355                                        $ Map.filterWithKey (\(from,_) _ -> idx == from) distanceMap
 356                            in List.take (round threshold) selected
 357                       ) indexes