[DB][Optim] Cooc without groups (wip).

bin/gargantext-init/Main.hs

@@ -50,7 +50,7 @@ main = do
     initMaster = do
       (masterUserId, masterRootId, masterCorpusId) <- getOrMk_RootWithCorpus userMaster (Left corpusMasterName) (Nothing :: Maybe HyperdataCorpus)
       masterListId <- getOrMkList masterCorpusId masterUserId
-      _ <- initTriggers masterListId
+      _triggers <- initTriggers masterListId
       pure (masterUserId, masterRootId, masterCorpusId, masterListId)
 
   withDevEnv iniPath $ \env -> do

devops/postgres/schema.sql

@@ -110,9 +110,9 @@ ALTER TABLE public.node_node_ngrams OWNER TO gargantua;
 
 CREATE TABLE public.node_node_ngrams2 (
 node_id   INTEGER NOT NULL REFERENCES public.nodes  (id) ON DELETE CASCADE,
-node_ngrams_id   INTEGER NOT NULL REFERENCES public.node_ngrams  (id) ON DELETE CASCADE,
+nodengrams_id   INTEGER NOT NULL REFERENCES public.node_ngrams  (id) ON DELETE CASCADE,
 weight double precision,
-PRIMARY KEY (node_id, node_ngrams_id, ngrams_field)
+PRIMARY KEY (node_id, nodengrams_id)
 );
 ALTER TABLE public.node_node_ngrams2 OWNER TO gargantua;
 
@@ -164,7 +164,8 @@ CREATE UNIQUE INDEX ON public.node_node_ngrams USING btree (node1_id, node2_id,
 CREATE        INDEX ON public.node_node_ngrams USING btree (node1_id,  node2_id);
 CREATE        INDEX ON public.node_node_ngrams USING btree (ngrams_id, node2_id);
 
-
+CREATE INDEX ON public.node_nodengrams_nodengrams USING btree (node_id, node_ngrams1_id, node_ngrams2_id);
+CREATE INDEX ON public.node_node_ngrams2 USING btree (node_id, nodengrams_id);
 ------------------------------------------------------------------------
 -- Ngrams Full DB Extraction Optim
 -- TODO remove hard parameter and move elsewhere

src/Gargantext/Database/Flow.hs

@@ -37,8 +37,10 @@ module Gargantext.Database.Flow -- (flowDatabase, ngrams2list)
   , flowAnnuaire
   )
     where
+
 import Prelude (String)
 import Data.Either
+import Data.Tuple.Extra (first, second)
 import Data.Traversable (traverse)
 import Debug.Trace (trace)
 import Control.Lens ((^.), view, _Just)
@@ -63,7 +65,8 @@ import Gargantext.Database.Root (getRoot)
 
 import Gargantext.Database.Schema.Ngrams -- (insertNgrams, Ngrams(..), NgramsIndexed(..), indexNgrams,  NgramsType(..), text2ngrams, ngramsTypeId)
 import Gargantext.Database.Schema.Node -- (mkRoot, mkCorpus, getOrMkList, mkGraph, {-mkPhylo,-} mkDashboard, mkAnnuaire, getCorporaWithParentId, HasNodeError, NodeError(..), nodeError)
-
+import Gargantext.Database.Schema.NodeNgrams (listInsertDb, getCgramsId)
+import Gargantext.Database.Schema.NodeNodeNgrams2 -- (NodeNodeNgrams2, insertNodeNodeNgrams2)
 import Gargantext.Database.Schema.User (getUser, UserLight(..))
 import Gargantext.Database.TextSearch (searchInDatabase)
 import Gargantext.Database.Types.Node -- (HyperdataDocument(..), NodeType(..), NodeId, UserId, ListId, CorpusId, RootId, MasterCorpusId, MasterUserId)
@@ -266,11 +269,27 @@ insertMasterDocs c lang hs  =  do
   maps <- mapNodeIdNgrams
        <$> documentIdWithNgrams (extractNgramsT $ withLang lang documentsWithId) documentsWithId
 
-  lId      <- getOrMkList masterCorpusId masterUserId
   terms2id <- insertNgrams $ Map.keys maps
+  -- to be removed
   let indexedNgrams = Map.mapKeys (indexNgrams terms2id) maps
 
+  -- new
+  lId      <- getOrMkList masterCorpusId masterUserId
+  mapCgramsId <- listInsertDb lId toNodeNgramsW'
+                $ map (first _ngramsTerms . second Map.keys)
+                $ Map.toList maps
+  -- insertDocNgrams
+  _return <- insertNodeNodeNgrams2
+           $ catMaybes [ NodeNodeNgrams2 <$> Just nId
+                                         <*> getCgramsId mapCgramsId ngrams_type (_ngramsTerms terms)
+                                         <*> Just (fromIntegral w :: Double)
+                       | (terms, mapNgramsTypes) <- Map.toList maps
+                       , (ngrams_type, mapNodeIdWeight) <- Map.toList mapNgramsTypes
+                       , (nId, w) <- Map.toList mapNodeIdWeight
+                       ]
+
   _cooc <- mkNode NodeListCooc lId masterUserId
+  -- to be removed
   _   <- insertDocNgrams lId indexedNgrams
 
   pure ids'

src/Gargantext/Database/Flow/List.hs

@@ -23,11 +23,13 @@ Portability : POSIX
 
 module Gargantext.Database.Flow.List
     where
+import Data.Text (Text)
 import Control.Monad (mapM_)
 import Data.Map (Map, toList)
 import Data.Maybe (Maybe(..), catMaybes)
 import Gargantext.API.Ngrams (NgramsElement(..), putListNgrams)
 import Gargantext.Database.Schema.Ngrams -- (insertNgrams, Ngrams(..), NgramsIndexed(..), indexNgrams,  NgramsType(..), text2ngrams, ngramsTypeId)
+import Gargantext.Core.Types.Main (ListType(CandidateTerm))
 import Gargantext.Database.Schema.NodeNgrams (NodeNgramsPoly(..), NodeNgramsW, listInsertDb, getCgramsId)
 import Gargantext.Database.Schema.Node_NodeNgramsNodeNgrams -- (insert_Node_NodeNgrams_NodeNgrams, Node_NodeNgrams_NodeNgrams(..))
 import Gargantext.Database.Types.Node -- (HyperdataDocument(..), NodeType(..), NodeId, UserId, ListId, CorpusId, RootId, MasterCorpusId, MasterUserId)
@@ -74,16 +76,25 @@ flowList_DbRepo lId ngs = do
 toNodeNgramsW :: ListId
               -> [(NgramsType, [NgramsElement])]
               -> [NodeNgramsW]
-toNodeNgramsW l ngs = List.concat $ map (toNodeNgramsW' l) ngs
+toNodeNgramsW l ngs = List.concat $ map (toNodeNgramsW'' l) ngs
   where
-    toNodeNgramsW' :: ListId
+    toNodeNgramsW'' :: ListId
                   -> (NgramsType, [NgramsElement])
                   -> [NodeNgramsW]
-    toNodeNgramsW' l' (ngrams_type, elms) =
+    toNodeNgramsW'' l' (ngrams_type, elms) =
       [ NodeNgrams Nothing l' list_type ngrams_terms' ngrams_type Nothing Nothing Nothing 0 |
        (NgramsElement ngrams_terms' _size list_type _occ _root _parent _children) <- elms
       ]
 
+toNodeNgramsW' :: ListId
+               -> [(Text, [NgramsType])]
+               -> [NodeNgramsW]
+toNodeNgramsW' l'' ngs = [ NodeNgrams Nothing l'' CandidateTerm terms ngrams_type Nothing Nothing Nothing 0
+                         | (terms, ngrams_types) <- ngs
+                         , ngrams_type <- ngrams_types
+                         ]
+
+
 listInsert :: FlowCmdM env err m
              => ListId
              -> Map NgramsType [NgramsElement]

src/Gargantext/Database/Init.hs

@@ -24,17 +24,19 @@ import Gargantext.Database.Utils (Cmd)
 import Gargantext.Prelude
 import Gargantext.Database.Triggers.Nodes (triggerSearchUpdate)
 import Gargantext.Database.Triggers.NodesNodes (triggerDeleteCount, triggerInsertCount, triggerUpdateAdd, triggerUpdateDel, MasterListId)
-import Gargantext.Database.Triggers.NodeNodeNgrams (triggerCountInsert)
+import Gargantext.Database.Triggers.NodeNodeNgrams (triggerCountInsert, triggerCountInsert2, triggerCoocInsert)
 ------------------------------------------------------------------------
 
 initTriggers :: MasterListId -> Cmd err [Int64]
 initTriggers lId = do
   t0 <- triggerSearchUpdate
   t1 <- triggerCountInsert
+  t1' <- triggerCountInsert2
+  t1'' <- triggerCoocInsert
   t2 <- triggerDeleteCount lId
   t3 <- triggerInsertCount lId
   t4 <- triggerUpdateAdd   lId
   t5 <- triggerUpdateDel   lId
-  pure [t0, t1,t2,t3,t4,t5]
+  pure [t0,t1,t1',t1'',t2,t3,t4,t5]
 
 

src/Gargantext/Database/Schema/NodeNgrams.hs

@@ -43,6 +43,8 @@ import Gargantext.Database.Utils
 import Gargantext.Database.Schema.Ngrams (NgramsType, ngramsTypeId, fromNgramsTypeId)
 import Gargantext.Prelude
 
+type NodeNgramsId = Int
+
 data NodeNgramsPoly id
                     node_id'
                     node_subtype

src/Gargantext/Database/Schema/NodeNodeNgrams2.hs

+{-|
+Module      : Gargantext.Database.Schema.NodeNodeNgrams
+Description : TODO: remove this module and table in database
+Copyright   : (c) CNRS, 2017-Present
+License     : AGPL + CECILL v3
+Maintainer  : team@gargantext.org
+Stability   : experimental
+Portability : POSIX
+
+-}
+
+{-# OPTIONS_GHC -fno-warn-orphans #-}
+
+{-# LANGUAGE Arrows                 #-}
+{-# LANGUAGE FlexibleInstances      #-}
+{-# LANGUAGE FunctionalDependencies #-}
+{-# LANGUAGE MultiParamTypeClasses  #-}
+{-# LANGUAGE NoImplicitPrelude      #-}
+{-# LANGUAGE RankNTypes             #-}
+{-# LANGUAGE TemplateHaskell        #-}
+
+module Gargantext.Database.Schema.NodeNodeNgrams2
+  where
+
+import Prelude
+import Data.Profunctor.Product.TH (makeAdaptorAndInstance)
+import Control.Lens.TH (makeLenses)
+import Gargantext.Database.Utils (Cmd, mkCmd)
+import Gargantext.Database.Schema.NodeNgrams (NodeNgramsId)
+import Gargantext.Database.Schema.Node (pgNodeId)
+import Gargantext.Database.Types.Node
+import Opaleye
+
+data NodeNodeNgrams2Poly node_id nodengrams_id w
+   = NodeNodeNgrams2 { _nnng2_node_id      :: node_id
+                     , _nnng2_nodengrams_id :: nodengrams_id
+                     , _nnng2_weight        :: w
+                     } deriving (Show)
+
+type NodeNodeNgrams2Write =
+     NodeNodeNgrams2Poly (Column PGInt4  )
+                        (Column PGInt4  )
+                        (Column PGFloat8)
+
+type NodeNodeNgrams2Read  =
+     NodeNodeNgrams2Poly (Column PGInt4  )
+                         (Column PGInt4  )
+                         (Column PGFloat8)
+
+type NodeNodeNgrams2ReadNull =
+     NodeNodeNgrams2Poly (Column (Nullable PGInt4  ))
+                        (Column (Nullable PGInt4  ))
+                        (Column (Nullable PGFloat8))
+
+type NodeNodeNgrams2 =
+  NodeNodeNgrams2Poly DocId NodeNgramsId Double
+
+$(makeAdaptorAndInstance "pNodeNodeNgrams2" ''NodeNodeNgrams2Poly)
+makeLenses ''NodeNodeNgrams2Poly
+
+
+nodeNodeNgrams2Table :: Table NodeNodeNgrams2Write NodeNodeNgrams2Read
+nodeNodeNgrams2Table  = Table "node_node_ngrams2"
+                          ( pNodeNodeNgrams2 NodeNodeNgrams2
+                               { _nnng2_node_id       = required "node_id"
+                               , _nnng2_nodengrams_id = required "nodengrams_id"
+                               , _nnng2_weight         = required "weight"
+                               }
+                          )
+
+queryNodeNodeNgrams2Table :: Query NodeNodeNgrams2Read
+queryNodeNodeNgrams2Table = queryTable nodeNodeNgrams2Table
+
+-- | Insert utils
+insertNodeNodeNgrams2 :: [NodeNodeNgrams2] -> Cmd err Int
+insertNodeNodeNgrams2 = insertNodeNodeNgrams2W
+                     . map (\(NodeNodeNgrams2 n1 n2 w) ->
+                              NodeNodeNgrams2 (pgNodeId n1)
+                                             (pgInt4 n2)
+                                             (pgDouble w)
+                                                  )
+
+insertNodeNodeNgrams2W :: [NodeNodeNgrams2Write] -> Cmd err Int
+insertNodeNodeNgrams2W nnnw =
+  mkCmd $ \c -> fromIntegral <$> runInsert_ c insertNothing
+    where
+      insertNothing = (Insert { iTable = nodeNodeNgrams2Table
+                              , iRows  = nnnw
+                              , iReturning = rCount
+                              , iOnConflict = (Just DoNothing)
+      })

src/Gargantext/Database/Triggers/NodeNodeNgrams.hs

@@ -63,4 +63,91 @@ triggerCountInsert = execPGSQuery query (nodeTypeId NodeDocument, nodeTypeId Nod
           EXECUTE PROCEDURE set_ngrams_global_count();
    |]
 
+triggerCountInsert2 :: Cmd err Int64
+triggerCountInsert2 = execPGSQuery query (nodeTypeId NodeCorpus, nodeTypeId NodeDocument, nodeTypeId NodeList)
+  where
+    query :: DPS.Query
+    query = [sql|
+          CREATE OR REPLACE FUNCTION set_ngrams_global_count2() RETURNS trigger AS $$
+          BEGIN
+            IF pg_trigger_depth() <> 1 THEN
+              RETURN NEW;
+            END IF;
+            IF TG_OP = 'INSERT' THEN
+                INSERT INTO node_node_ngrams2 (node_id, nodengrams_id, weight)
+                  SELECT corpus.id, nng.id, count(*) from NEW as new1
+                        INNER JOIN node_ngrams nng    ON nng.id = new1.nodengrams_id
+                        INNER JOIN nodes       list   ON list.id   = nng.node_id
+                        INNER JOIN nodes_nodes nn     ON nn.node2_id = new1.node_id
+                        INNER JOIN nodes       corpus ON corpus.id   = nn.node1_id
+                        INNER JOIN nodes       doc    ON doc.id      = nn.node2_id
+                        WHERE corpus.typename = ? -- 30 -- corpus
+                          AND doc.typename    = ? -- 4  -- maybe not mandatory
+                          AND list.typename   = ? -- 5  -- list
+                        GROUP BY corpus.id, nng.id
+
+                ON CONFLICT (node_id, nodengrams_id)
+                   DO UPDATE set weight = node_node_ngrams2.weight + excluded.weight
+                   ;
+            END IF;
+
+            RETURN NULL;
+          END
+          $$ LANGUAGE plpgsql;
+
+          -- DROP trigger trigger_count_insert2 on node_node_ngrams2;
+
+          CREATE TRIGGER trigger_count_insert2 AFTER INSERT on node_node_ngrams2
+          REFERENCING NEW TABLE AS NEW
+          FOR EACH STATEMENT
+          EXECUTE PROCEDURE set_ngrams_global_count2();
+   |]
+
+-- TODO add the groups
+triggerCoocInsert :: Cmd err Int64
+triggerCoocInsert = execPGSQuery query (nodeTypeId NodeCorpus, nodeTypeId NodeDocument, nodeTypeId NodeList)
+  where
+    query :: DPS.Query
+    query = [sql|
+          CREATE OR REPLACE FUNCTION set_cooc() RETURNS trigger AS $$
+          BEGIN
+            IF pg_trigger_depth() <> 1 THEN
+              RETURN NEW;
+            END IF;
+            IF TG_OP = 'INSERT' THEN
+                INSERT INTO node_nodengrams_nodengrams (node_id, node_ngrams1_id, node_ngrams2_id, weight)
+
+                  SELECT corpus.id, nng1.id, nng2.id, count(*) from NEW as new1
+                        INNER JOIN node_ngrams nng1   ON nng1.id     = new1.nodengrams_id
+                        INNER JOIN nodes       list   ON list.id     = nng1.node_id
+                        INNER JOIN nodes_nodes nn     ON nn.node2_id = new1.node_id
+                        INNER JOIN nodes       corpus ON corpus.id   = nn.node1_id
+                        INNER JOIN nodes       doc    ON doc.id      = nn.node2_id
+
+                        INNER JOIN node_node_ngrams2 nnng2 ON nnng2.node_id  = doc.id
+                        INNER JOIN node_ngrams       nng2  ON nng2.id        = nnng2.nodengrams_id
+
+                        WHERE corpus.typename = ? -- 30 -- corpus
+                          AND doc.typename    = ? -- 4  -- maybe not mandatory
+                          AND list.typename   = ? -- 5  -- list
+                          AND nng2.node_id    = list.id
+                          -- AND nng1.id <> nng2.id
+                        GROUP BY corpus.id, nng1.id, nng2.id
+
+                ON CONFLICT (node_id, node_ngrams1_id, node_ngrams2_id)
+                   DO UPDATE set weight = node_nodengrams_nodengrams.weight + excluded.weight
+                   ;
+            END IF;
+
+            RETURN NULL;
+          END
+          $$ LANGUAGE plpgsql;
+
+          -- DROP trigger trigger_cooc on node_node_ngrams2;
+
+          CREATE TRIGGER trigger_cooc_insert AFTER INSERT on node_node_ngrams2
+          REFERENCING NEW TABLE AS NEW
+          FOR EACH STATEMENT
+          EXECUTE PROCEDURE set_cooc();
+   |]