{-| Module : Gargantext.Database.Metrics.NgramsByContext Description : Ngrams by Node user and master Copyright : (c) CNRS, 2017-Present License : AGPL + CECILL v3 Maintainer : team@gargantext.org Stability : experimental Portability : POSIX Ngrams by node enable contextual metrics. -} {-# LANGUAGE QuasiQuotes #-} module Gargantext.Database.Action.Metrics.NgramsByContext where -- import Debug.Trace (trace) --import Data.Map.Strict.Patch (PatchMap, Replace, diff) -- import Control.Monad (void) import Data.HashMap.Strict (HashMap) import Data.Map.Strict (Map) import Data.Set (Set) import Data.Text (Text) import Data.Tuple.Extra (first, second, swap) import Database.PostgreSQL.Simple.SqlQQ (sql) import Database.PostgreSQL.Simple.Types (Values(..), QualifiedIdentifier(..)) import Gargantext.API.Ngrams.Types (NgramsTerm(..)) import Gargantext.Core import Gargantext.Data.HashMap.Strict.Utils as HM import Gargantext.Database.Admin.Types.Node (ListId, CorpusId, NodeId(..), ContextId, MasterCorpusId, NodeType(NodeDocument), UserCorpusId, DocId) import Gargantext.Database.Prelude (Cmd, runPGSQuery) -- , execPGSQuery) import Gargantext.Database.Schema.Ngrams (ngramsTypeId, NgramsType(..)) import Gargantext.Prelude import qualified Data.HashMap.Strict as HM import qualified Data.List as List import qualified Data.Map.Strict as Map import qualified Data.Set as Set import qualified Database.PostgreSQL.Simple as DPS import qualified Database.PostgreSQL.Simple.Types as DPST -- | fst is size of Supra Corpus -- snd is Texts and size of Occurrences (different docs) countContextsByNgramsWith :: (NgramsTerm -> NgramsTerm) -> HashMap NgramsTerm (Set ContextId) -> (Double, HashMap NgramsTerm (Double, Set NgramsTerm)) countContextsByNgramsWith f m = (total, m') where total = fromIntegral $ Set.size $ Set.unions $ HM.elems m m' = HM.map ( swap . second (fromIntegral . Set.size)) $ groupContextsByNgramsWith f m groupContextsByNgramsWith :: (NgramsTerm -> NgramsTerm) -> HashMap NgramsTerm (Set NodeId) -> HashMap NgramsTerm (Set NgramsTerm, Set ContextId) groupContextsByNgramsWith f' m'' = HM.fromListWith (<>) $ map (\(t,ns) -> (f' t, (Set.singleton t, ns))) $ HM.toList m'' ------------------------------------------------------------------------ getContextsByNgramsUser :: HasDBid NodeType => CorpusId -> NgramsType -> Cmd err (HashMap NgramsTerm (Set ContextId)) getContextsByNgramsUser cId nt = HM.fromListWith (<>) <$> map (\(n,t) -> (NgramsTerm t, Set.singleton n)) <$> selectNgramsByContextUser cId nt where selectNgramsByContextUser :: HasDBid NodeType => CorpusId -> NgramsType -> Cmd err [(NodeId, Text)] selectNgramsByContextUser cId' nt' = runPGSQuery queryNgramsByContextUser ( cId' , toDBid NodeDocument , ngramsTypeId nt' -- , 100 :: Int -- limit -- , 0 :: Int -- offset ) queryNgramsByContextUser :: DPS.Query queryNgramsByContextUser = [sql| SELECT cng.context_id, ng.terms FROM context_node_ngrams cng JOIN ngrams ng ON cng.ngrams_id = ng.id JOIN nodes_contexts nc ON nc.context_id = cng.context_id JOIN contexts c ON nc.context_id = c.id WHERE nc.node_id = ? -- CorpusId AND c.typename = ? -- toDBid AND cng.ngrams_type = ? -- NgramsTypeId AND nc.category > 0 -- is not in Trash GROUP BY cng.context_id, ng.terms |] ------------------------------------------------------------------------ getOccByNgramsOnlyFast_withSample :: HasDBid NodeType => CorpusId -> Int -> NgramsType -> [NgramsTerm] -> Cmd err (HashMap NgramsTerm Int) getOccByNgramsOnlyFast_withSample cId int nt ngs = HM.fromListWith (+) <$> selectNgramsOccurrencesOnlyByContextUser_withSample cId int nt ngs getOccByNgramsOnlyFast :: CorpusId -> ListId -> NgramsType -> Cmd err (HashMap NgramsTerm [ContextId]) getOccByNgramsOnlyFast cId lId nt = do --HM.fromList <$> map (\(t,n) -> (NgramsTerm t, round n)) <$> run cId lId nt HM.fromList <$> map (\(t, ns) -> (NgramsTerm t, NodeId <$> DPST.fromPGArray ns)) <$> run cId lId nt where run :: CorpusId -> ListId -> NgramsType -> Cmd err [(Text, DPST.PGArray Int)] run cId' lId' nt' = runPGSQuery query ( cId' , lId' , ngramsTypeId nt' ) query :: DPS.Query query = [sql| WITH cnnv AS ( SELECT DISTINCT context_node_ngrams.context_id, context_node_ngrams.ngrams_id, nodes_contexts.node_id FROM nodes_contexts JOIN context_node_ngrams ON context_node_ngrams.context_id = nodes_contexts.context_id ), node_context_ids AS (SELECT context_id, ngrams_id, terms FROM cnnv JOIN ngrams ON cnnv.ngrams_id = ngrams.id WHERE node_id = ? ), ncids_agg AS (SELECT ngrams_id, terms, array_agg(DISTINCT context_id) AS agg FROM node_context_ids GROUP BY (ngrams_id, terms)), ns AS (SELECT ngrams_id, terms FROM node_stories JOIN ngrams ON ngrams_id = ngrams.id WHERE node_id = ? AND ngrams_type_id = ? ) SELECT ns.terms, CASE WHEN agg IS NULL THEN '{}' ELSE agg END FROM ns LEFT JOIN ncids_agg ON ns.ngrams_id = ncids_agg.ngrams_id |] -- query = [sql| -- WITH node_context_ids AS -- (select context_id, ngrams_id -- FROM context_node_ngrams_view -- WHERE node_id = ? -- ), ns AS -- (select ngrams_id FROM node_stories -- WHERE node_id = ? AND ngrams_type_id = ? -- ) -- SELECT ng.terms, -- ARRAY ( SELECT DISTINCT context_id -- FROM node_context_ids -- WHERE ns.ngrams_id = node_context_ids.ngrams_id -- ) -- AS context_ids -- FROM ngrams ng -- JOIN ns ON ng.id = ns.ngrams_id -- |] selectNgramsOccurrencesOnlyByContextUser_withSample :: HasDBid NodeType => CorpusId -> Int -> NgramsType -> [NgramsTerm] -> Cmd err [(NgramsTerm, Int)] selectNgramsOccurrencesOnlyByContextUser_withSample cId int nt tms = fmap (first NgramsTerm) <$> runPGSQuery queryNgramsOccurrencesOnlyByContextUser_withSample ( int , toDBid NodeDocument , cId , Values fields ((DPS.Only . unNgramsTerm) <$> (List.take 10000 tms)) , cId , ngramsTypeId nt ) where fields = [QualifiedIdentifier Nothing "text"] queryNgramsOccurrencesOnlyByContextUser_withSample :: DPS.Query queryNgramsOccurrencesOnlyByContextUser_withSample = [sql| WITH nodes_sample AS (SELECT n.id FROM contexts n TABLESAMPLE SYSTEM_ROWS (?) JOIN nodes_contexts nn ON n.id = nn.context_id WHERE n.typename = ? AND nn.node_id = ?), input_rows(terms) AS (?) SELECT ng.terms, COUNT(cng.context_id) FROM context_node_ngrams cng JOIN ngrams ng ON cng.ngrams_id = ng.id JOIN input_rows ir ON ir.terms = ng.terms JOIN nodes_contexts nn ON nn.context_id = cng.context_id JOIN nodes_sample n ON nn.context_id = n.id WHERE nn.node_id = ? -- CorpusId AND cng.ngrams_type = ? -- NgramsTypeId AND nn.category > 0 GROUP BY cng.node_id, ng.terms |] selectNgramsOccurrencesOnlyByContextUser_withSample' :: HasDBid NodeType => CorpusId -> Int -> NgramsType -> Cmd err [(NgramsTerm, Int)] selectNgramsOccurrencesOnlyByContextUser_withSample' cId int nt = fmap (first NgramsTerm) <$> runPGSQuery queryNgramsOccurrencesOnlyByContextUser_withSample ( int , toDBid NodeDocument , cId , cId , ngramsTypeId nt ) queryNgramsOccurrencesOnlyByContextUser_withSample' :: DPS.Query queryNgramsOccurrencesOnlyByContextUser_withSample' = [sql| WITH contexts_sample AS (SELECT c.id FROM contexts c TABLESAMPLE SYSTEM_ROWS (?) JOIN nodes_contexts nc ON c.id = nc.context_id WHERE c.typename = ? AND nc.node_id = ?) SELECT ng.terms, COUNT(cng.context_id) FROM context_node_ngrams cng JOIN ngrams ng ON cng.ngrams_id = ng.id JOIN node_stories ns ON ns.ngrams_id = ng.id JOIN nodes_contexts nc ON nc.context_id = cng.context_id JOIN contexts_sample c ON nc.context_id = c.id WHERE nc.node_id = ? -- CorpusId AND cng.ngrams_type = ? -- NgramsTypeId AND nc.category > 0 GROUP BY ng.id |] ------------------------------------------------------------------------ getContextsByNgramsOnlyUser :: HasDBid NodeType => CorpusId -> [ListId] -> NgramsType -> [NgramsTerm] -> Cmd err (HashMap NgramsTerm (Set NodeId)) getContextsByNgramsOnlyUser cId ls nt ngs = HM.unionsWith (<>) . map (HM.fromListWith (<>) . map (second Set.singleton)) <$> mapM (selectNgramsOnlyByContextUser cId ls nt) (splitEvery 1000 ngs) getNgramsByContextOnlyUser :: HasDBid NodeType => NodeId -> [ListId] -> NgramsType -> [NgramsTerm] -> Cmd err (Map NodeId (Set NgramsTerm)) getNgramsByContextOnlyUser cId ls nt ngs = Map.unionsWith (<>) . map ( Map.fromListWith (<>) . map (second Set.singleton) ) . map (map swap) <$> mapM (selectNgramsOnlyByContextUser cId ls nt) (splitEvery 1000 ngs) ------------------------------------------------------------------------ selectNgramsOnlyByContextUser :: HasDBid NodeType => CorpusId -> [ListId] -> NgramsType -> [NgramsTerm] -> Cmd err [(NgramsTerm, ContextId)] selectNgramsOnlyByContextUser cId ls nt tms = fmap (first NgramsTerm) <$> runPGSQuery queryNgramsOnlyByContextUser ( Values fields ((DPS.Only . unNgramsTerm) <$> tms) , Values [QualifiedIdentifier Nothing "int4"] (DPS.Only <$> (map (\(NodeId n) -> n) ls)) , cId , toDBid NodeDocument , ngramsTypeId nt ) where fields = [QualifiedIdentifier Nothing "text"] queryNgramsOnlyByContextUser :: DPS.Query queryNgramsOnlyByContextUser = [sql| WITH input_rows(terms) AS (?), input_list(id) AS (?) SELECT ng.terms, cng.context_id FROM context_node_ngrams cng JOIN ngrams ng ON cng.ngrams_id = ng.id JOIN input_rows ir ON ir.terms = ng.terms JOIN input_list il ON il.id = cng.node_id JOIN nodes_contexts nc ON nc.context_id = cng.context_id JOIN contexts c ON nc.context_id = c.id WHERE nc.node_id = ? -- CorpusId AND c.typename = ? -- toDBid (maybe not useful with context table) AND cng.ngrams_type = ? -- NgramsTypeId AND nc.category > 0 GROUP BY ng.terms, cng.context_id |] getNgramsByDocOnlyUser :: DocId -> [ListId] -> NgramsType -> [NgramsTerm] -> Cmd err (HashMap NgramsTerm (Set NodeId)) getNgramsByDocOnlyUser cId ls nt ngs = HM.unionsWith (<>) . map (HM.fromListWith (<>) . map (second Set.singleton)) <$> mapM (selectNgramsOnlyByDocUser cId ls nt) (splitEvery 1000 ngs) selectNgramsOnlyByDocUser :: DocId -> [ListId] -> NgramsType -> [NgramsTerm] -> Cmd err [(NgramsTerm, NodeId)] selectNgramsOnlyByDocUser dId ls nt tms = fmap (first NgramsTerm) <$> runPGSQuery queryNgramsOnlyByDocUser ( Values fields ((DPS.Only . unNgramsTerm) <$> tms) , Values [QualifiedIdentifier Nothing "int4"] (DPS.Only <$> (map (\(NodeId n) -> n) ls)) , dId , ngramsTypeId nt ) where fields = [QualifiedIdentifier Nothing "text"] queryNgramsOnlyByDocUser :: DPS.Query queryNgramsOnlyByDocUser = [sql| WITH input_rows(terms) AS (?), input_list(id) AS (?) SELECT ng.terms, cng.node_id FROM context_node_ngrams cng JOIN ngrams ng ON cng.ngrams_id = ng.id JOIN input_rows ir ON ir.terms = ng.terms JOIN input_list il ON il.id = cng.context_id WHERE cng.node_id = ? -- DocId AND cng.ngrams_type = ? -- NgramsTypeId GROUP BY ng.terms, cng.node_id |] ------------------------------------------------------------------------ -- | TODO filter by language, database, any social field getContextsByNgramsMaster :: HasDBid NodeType => UserCorpusId -> MasterCorpusId -> Cmd err (HashMap Text (Set NodeId)) getContextsByNgramsMaster ucId mcId = unionsWith (<>) . map (HM.fromListWith (<>) . map (\(n,t) -> (t, Set.singleton n))) -- . takeWhile (not . List.null) -- . takeWhile (\l -> List.length l > 3) <$> mapM (selectNgramsByContextMaster 1000 ucId mcId) [0,500..10000] selectNgramsByContextMaster :: HasDBid NodeType => Int -> UserCorpusId -> MasterCorpusId -> Int -> Cmd err [(NodeId, Text)] selectNgramsByContextMaster n ucId mcId p = runPGSQuery queryNgramsByContextMaster' ( ucId , ngramsTypeId NgramsTerms , toDBid NodeDocument , p , toDBid NodeDocument , p , n , mcId , toDBid NodeDocument , ngramsTypeId NgramsTerms ) -- | TODO fix context_node_ngrams relation queryNgramsByContextMaster' :: DPS.Query queryNgramsByContextMaster' = [sql| WITH contextsByNgramsUser AS ( SELECT n.id, ng.terms FROM contexts n JOIN nodes_contexts nn ON n.id = nn.context_id JOIN context_node_ngrams cng ON cng.context_id = n.id JOIN ngrams ng ON cng.ngrams_id = ng.id WHERE nn.node_id = ? -- UserCorpusId -- AND n.typename = ? -- toDBid AND cng.ngrams_type = ? -- NgramsTypeId AND nn.category > 0 AND node_pos(n.id,?) >= ? AND node_pos(n.id,?) < ? GROUP BY n.id, ng.terms ), contextsByNgramsMaster AS ( SELECT n.id, ng.terms FROM contexts n TABLESAMPLE SYSTEM_ROWS(?) JOIN context_node_ngrams cng ON n.id = cng.context_id JOIN ngrams ng ON ng.id = cng.ngrams_id WHERE n.parent_id = ? -- Master Corpus toDBid AND n.typename = ? -- toDBid AND cng.ngrams_type = ? -- NgramsTypeId GROUP BY n.id, ng.terms ) SELECT m.id, m.terms FROM nodesByNgramsMaster m RIGHT JOIN contextsByNgramsUser u ON u.id = m.id |] -- | Refreshes the \"context_node_ngrams_view\" materialized view. -- This function will be run : -- - periodically -- - at reindex stage -- - at the end of each text flow -- refreshNgramsMaterialized :: Cmd err () -- refreshNgramsMaterialized = void $ execPGSQuery refreshNgramsMaterializedQuery () -- where -- refreshNgramsMaterializedQuery :: DPS.Query -- refreshNgramsMaterializedQuery = -- [sql| REFRESH MATERIALIZED VIEW CONCURRENTLY context_node_ngrams_view; |]