func (n *Node) Parents() ([]lattice.Node, error) { if n.pat.Items.Size() == 0 { return []lattice.Node{}, nil } else if n.pat.Items.Size() == 1 { return []lattice.Node{n.dt.empty}, nil } i := setToInt32s(n.pat.Items) if has, err := n.dt.ParentCount.Has(i); err != nil { return nil, err } else if has { return n.cached(n.dt.Parents, i) } parents := make([]*set.SortedSet, 0, n.pat.Items.Size()) for item, next := n.pat.Items.Items()(); next != nil; item, next = next() { parent := n.pat.Items.Copy() parent.Delete(item) parents = append(parents, parent) } nodes := make([]lattice.Node, 0, 10) for _, items := range parents { if node, err := TryLoadNode(setToInt32s(items), n.dt); err != nil { return nil, err } else if node != nil { nodes = append(nodes, node) continue } ctxs := int32sToSet(n.txs) var txs types.Set for item, next := items.Items()(); next != nil; item, next = next() { mytxs := set.NewSortedSet(len(n.txs) + 10) for _, tx := range n.dt.InvertedIndex[item.(types.Int32)] { if !ctxs.Has(types.Int32(tx)) { mytxs.Add(types.Int32(tx)) } } var err error if txs == nil { txs = mytxs } else { txs, err = txs.Intersect(mytxs) if err != nil { return nil, err } } } txs, err := txs.Union(ctxs) if err != nil { return nil, err } stxs := make([]int32, 0, txs.Size()) for item, next := txs.Items()(); next != nil; item, next = next() { stxs = append(stxs, int32(item.(types.Int32))) } node := &Node{Pattern{items}, n.dt, stxs} err = node.Save() if err != nil { return nil, err } nodes = append(nodes, node) } err := n.cache(n.dt.ParentCount, n.dt.Parents, i, nodes) if err != nil { return nil, err } return nodes, nil }
func jaccardSetSimilarity(a, b types.Set) float64 { i, err := a.Intersect(b) exc.ThrowOnError(err) inter := float64(i.Size()) return 1.0 - (inter / (float64(a.Size()) + float64(b.Size()) - inter)) }
// unique extensions and supported embeddings func ExtsAndEmbs(dt *Digraph, pattern *subgraph.SubGraph, patternOverlap []map[int]bool, unsupExts types.Set, unsupEmbs map[subgraph.VrtEmb]bool, mode Mode, debug bool) (int, []*subgraph.Extension, []*subgraph.Embedding, []map[int]bool, subgraph.VertexEmbeddings, error) { if !debug { if has, support, exts, embs, overlap, unsupEmbs, err := loadCachedExtsEmbs(dt, pattern); err != nil { return 0, nil, nil, nil, nil, err } else if has { if false { errors.Logf("LOAD-DEBUG", "Loaded cached %v exts %v embs %v", pattern, len(exts), len(embs)) } return support, exts, embs, overlap, unsupEmbs, nil } } if CACHE_DEBUG || debug { errors.Logf("CACHE-DEBUG", "ExtsAndEmbs %v", pattern.Pretty(dt.Labels)) } // compute the embeddings var seen map[int]bool = nil var ei subgraph.EmbIterator var dropped *subgraph.VertexEmbeddings switch { case mode&(MNI|FIS) != 0: ei, dropped = pattern.IterEmbeddings( dt.EmbSearchStartPoint, dt.Indices, unsupEmbs, patternOverlap, nil) case mode&(GIS) == GIS: seen = make(map[int]bool) ei, dropped = pattern.IterEmbeddings( dt.EmbSearchStartPoint, dt.Indices, unsupEmbs, patternOverlap, func(ids *subgraph.IdNode) bool { for c := ids; c != nil; c = c.Prev { if _, has := seen[c.Id]; has { for c := ids; c != nil; c = c.Prev { seen[c.Id] = true } return true } } return false }) default: return 0, nil, nil, nil, nil, errors.Errorf("Unknown support counting strategy %v", mode) } // find the actual embeddings and compute the extensions // the extensions are stored in exts // the embeddings are stored in sets var exts types.Set var fisEmbs []*subgraph.Embedding var sets []*hashtable.LinearHash var overlap []map[int]bool var total int if mode&ExtFromEmb == ExtFromEmb && len(pattern.E) > 0 { // add the supported embeddings to the vertex sets // add the extensions to the extensions set total, overlap, fisEmbs, sets, exts = extensionsFromEmbeddings(dt, pattern, ei, seen) if total == 0 { // return 0, nil, nil, nil, nil, errors.Errorf("could not find any embedding of %v", pattern) // because we are extending from frequent edges for vertices this // is ok. return 0, nil, nil, nil, nil, nil } } else if mode&ExtFromFreqEdges == ExtFromFreqEdges || len(pattern.E) <= 0 { total, overlap, fisEmbs, sets, exts = extensionsFromFreqEdges(dt, pattern, ei, seen) if total < dt.Support() { return 0, nil, nil, nil, nil, nil } } else { return 0, nil, nil, nil, nil, errors.Errorf("Unknown extension strategy %v", mode) } // construct the extensions output slice extensions := make([]*subgraph.Extension, 0, exts.Size()) for i, next := exts.Items()(); next != nil; i, next = next() { ext := i.(*subgraph.Extension) if unsupExts != nil && unsupExts.Has(ext) { continue } extensions = append(extensions, ext) } if mode&EmbeddingPruning == EmbeddingPruning && unsupEmbs != nil { // for i, next := unsupEmbs.Items()(); next != nil; i, next = next() { for emb, ok := range unsupEmbs { if ok { *dropped = append(*dropped, &emb) } } } var embeddings []*subgraph.Embedding if mode&(MNI|GIS) != 0 { // compute the minimally supported vertex arg, size := stats.Min(stats.RandomPermutation(len(sets)), func(i int) float64 { return float64(sets[i].Size()) }) // construct the embeddings output slice embeddings = make([]*subgraph.Embedding, 0, int(size)+1) for i, next := sets[arg].Values()(); next != nil; i, next = next() { emb := i.(*subgraph.Embedding) embeddings = append(embeddings, emb) } } else if mode&(FIS) == FIS { embeddings = fisEmbs } else { return 0, nil, nil, nil, nil, errors.Errorf("Unknown support counting strategy %v", mode) } if CACHE_DEBUG || debug { errors.Logf("CACHE-DEBUG", "Caching exts %v embs %v total-embs %v : %v", len(extensions), len(embeddings), total, pattern.Pretty(dt.Labels)) } if !debug { err := cacheExtsEmbs(dt, pattern, len(embeddings), extensions, embeddings, overlap, *dropped) if err != nil { return 0, nil, nil, nil, nil, err } } return len(embeddings), extensions, embeddings, overlap, *dropped, nil }