Query splitter complete

relational/optimizer.py

@@ -25,6 +25,7 @@
 
 from relational import optimizations
 from relational import parser
+from relational import querysplit
 from relational.maintenance import UserInterface
 
 
@@ -57,8 +58,8 @@ def optimize_program(code, rels):
         parsed = parser.tree(query)
         optimizations.replace_leaves(parsed, context)
         context[res] = parsed
-    result = optimize_all(context[last_res], rels)
+    node = optimize_all(context[last_res], rels, tostr=False)
-    return result
+    return querysplit.split(node, rels)
 
 
 def optimize_all(expression, rels, specific=True, general=True, debug=None,tostr=True):

relational/querysplit.py

@@ -19,6 +19,52 @@
 # This module splits a query into a program.
 
 
+from relational import parser
+
+
+class Program:
+    def __init__(self, rels):
+        self.queries = []
+        self.dictionary = {} # Key is the query, value is the relation
+        self.vgen = vargen(rels, 'optm_')
+
+    def __str__(self):
+        r = ''
+        for q in self.queries:
+            r += '%s = %s' % (q[0], q[1]) + '\n'
+        return r.rstrip()
+
+    def append_query(self, node):
+        strnode = str(node)
+
+        rel = self.dictionary.get(strnode)
+        if rel:
+            return rel
+
+        qname = next(self.vgen)
+        self.queries.append((qname, node))
+        n = parser.Node()
+        n.kind = parser.RELATION
+        n.name = qname
+        self.dictionary[strnode] = n
+        return n
+
+def _separate(node, program):
+    if node.kind == parser.UNARY and node.child.kind != parser.RELATION:
+        _separate(node.child, program)
+        rel = program.append_query(node.child)
+        node.child = rel
+    elif node.kind == parser.BINARY:
+        if node.left.kind != parser.RELATION:
+            _separate(node.left, program)
+            rel = program.append_query(node.left)
+            node.left = rel
+        if node.right.kind != parser.RELATION:
+            _separate(node.right, program)
+            rel = program.append_query(node.right)
+            node.right = rel
+    program.append_query(node)
+
 def vargen(avoid, prefix=''):
     '''
     Generates temp variables.
@@ -40,3 +86,14 @@ def vargen(avoid, prefix=''):
         if r not in avoid:
             yield r
         count += 1
+
+def split(node, rels):
+    '''
+    Split a query into a program.
+
+    The idea is that if there are duplicated subdtrees they
+    get executed only once.
+    '''
+    p = Program(rels)
+    _separate(node, p)
+    return str(p)