From: Ralph Ronnquist <ralph.ronnquist@gmail.com>
Date: Tue, 5 Jul 2022 11:55:26 +0000 (+1000)
Subject: Add "relation" implementation based on hashvector
X-Git-Url: https://git.rrq.au/?a=commitdiff_plain;h=c08a0975036f23f972a16a07efb0c07113c96d2e;p=rrq%2Frrqmisc.git

Add "relation" implementation based on hashvector
---

diff --git a/vector/Makefile b/vector/Makefile
index 7c5268d..d5c718b 100644
--- a/vector/Makefile
+++ b/vector/Makefile
@@ -1,6 +1,6 @@
 LIBRARY = libvector.a
 LIBOBJS = vector.o hashvector.o
-LIBOBJS += integeritem.o  stringitem.o  tupleitem.o
+LIBOBJS += integeritem.o  stringitem.o  tupleitem.o relation.o
 
 default: $(LIBRARY)
 
diff --git a/vector/relation.c b/vector/relation.c
new file mode 100644
index 0000000..330b3de
--- /dev/null
+++ b/vector/relation.c
@@ -0,0 +1,100 @@
+#include <stdlib.h>
+#include <relation.h>
+
+#define COPYA(T,P,N) (T*) memcpy( malloc( N * sizeof(T) ), P, N * sizeof( T ) )
+#define COPY(T,P) COPYA(T,P,1)
+
+// Add an indexing hashvector to the relation using the nominated
+// column indexes being the value part. the key must be a clone of the
+// relation columns but with some columns reset.
+int relation_addindex(relation *r,tupleschema *key) {
+    tupleschema *primary = r->columns;
+    if ( primary->arity != key->arity ) {
+	return -1;
+    }
+    int i = 0;
+    for ( ; i < primary->arity; i++ ) {
+	if ( key->columns[i] && primary->columns[i] != key->columns[i] ) {
+	    return -1;
+	}
+    }
+    i = (int) r->indexes.size;
+    vector_append( &r->indexes, hashvector_create( 1, (itemkeyfun*) key ) );
+    return i;
+}
+
+relation *relation_create(int arity,tupleschema *schema) {
+    relation *r = (relation *) malloc( sizeof( relation ) );
+    (*r) = (relation) {
+	.indexes = {
+	.variant = 2,
+	.size = 0,
+	.entries = 0
+	},
+	.columns = schema
+    };
+    return r;
+}
+
+// Iteration context for adding or querying a relation
+typedef struct {
+    itemkeyfun *columns;
+    vector knockouts;
+    void *key;
+} knockoutdata;
+
+// Determine matches to ((knockoutdata*)data)->key in
+// (hashvector*)item, optionally using ((knockoutdata*)data)->columns
+// for ignoring full matches to the key tuple.
+static int knockout_check(vector_index index,void *item,void *data) {
+    knockoutdata *kod = (knockoutdata*) data;
+    void *old = 0;
+    if ( hashvector_find( (hashvector*) item, kod->key, &old ) ) {
+	if ( kod->columns == 0 ||
+	     kod->columns->haskey( kod->columns, old, kod->key ) == 0 ) {
+	    if ( vector_find( &kod->knockouts, old ) >= kod->knockouts.size ) {
+		vector_append( &kod->knockouts, old );
+	    }
+	}
+    }
+    return 0;
+}
+
+// delete the (tuple*)item from the (hashvector*)data
+static int knockout_delete_item(vector_index index,void *item,void *data) {
+    hashvector_delete( (hashvector*)data, item );
+    return 0;
+}
+
+// delete the tuples of (vector*)data from the (hashvector*)item
+static int knockout_delete(vector_index index,void *item,void *data) {
+    vector_iterate( (vector*)data, 0, knockout_delete_item, item );
+    return 0;
+}
+
+// add the (tuple*)data to the (hashvector*)item
+static int knockout_add(vector_index index,void *item,void *data) {
+    hashvector_add( (hashvector*)item, data );
+    return 0;
+}
+
+// add a tuple to a relation and return a vector of knocked out
+// tuples, if any, or 0 otherwise.
+vector *relation_add(relation *r,tuple *x) {
+    knockoutdata data = {
+	.columns = &(r->columns->functions),
+	.knockouts = { .variant = 2, .size = 0, .entries = 0 },
+	.key = x
+    };
+    // Find all knockouts
+    vector_iterate( &r->indexes, 0, knockout_check, &data );
+    // Delete them
+    vector_iterate( &r->indexes, 0, knockout_delete, &data.knockouts );
+    // Add the new tuple
+    vector_iterate( &r->indexes, 0, knockout_add, x );
+    if ( data.knockouts.size == 0 ) {
+	return 0;
+    }
+    return vector_clone( 3, &data.knockouts );
+}
+
diff --git a/vector/relation.h b/vector/relation.h
new file mode 100644
index 0000000..b0996c1
--- /dev/null
+++ b/vector/relation.h
@@ -0,0 +1,40 @@
+#ifndef relation_H
+#define relation_H
+
+#include <hashvector.h>
+#include <tupleitem.h>
+
+/**
+ * A relation is an implementation of a tuple set with a primary key
+ * through a hashvector whose "type" defines the key. Additional
+ * hashvectors may be added to impose additional key restrictions and
+ * the final relation is the intersection of them all.
+ *
+ * The relation record itself holds the column schema whereas the
+ * indexes hold key schemas that in practice are copies of the column
+ * schema with some columns blanked out.
+ */
+typedef struct {
+    vector indexes; // one or more indexes over the tuple collection
+    tupleschema *columns;
+} relation;
+
+/**
+ * Create a relation of the given arity and tupleschema.
+ */
+extern relation *relation_create(int arity,tupleschema *schema);
+
+/**
+ * Add a a key index to the relation by identifying the value part for
+ * this index.
+ */
+extern int relation_addindex(relation *r,tupleschema *ts);
+
+
+/**
+ * Add a tuple to a relation.
+ * \returns a vector of all knocked out tuples.
+ */
+extern vector *relation_add(relation *r,tuple *t);
+
+#endif