Python: Add LocalSourceNode and flowsTo

This fixes the major performance problem with type tracking on
some (pathological) databases.

The interface could probably be improved a bit. In particular, I'm
thinking that we might want to have `DataFlow::exprNode` return a
`LocalSourceNode` so that a cast isn't necessary in order to use
`flowsTo`.

I have added two `cached` annotations. The one on `flowsTo` is
crucial, as performance regresses without it. The one on
`simpleLocalFlowStep` may not be needed, but Java has a similar
annotation, and to me it makes sense to have this relation cached.

python/ql/src/semmle/python/Concepts.qll

@@ -336,7 +336,7 @@ module HTTP {
         /** Gets the URL pattern for this route, if it can be statically determined. */
         string getUrlPattern() {
           exists(StrConst str |
-            DataFlow::localFlow(DataFlow::exprNode(str), this.getUrlPatternArg()) and
+            DataFlow::exprNode(str).(DataFlow::LocalSourceNode).flowsTo(this.getUrlPatternArg()) and
             result = str.getText()
           )
         }
@@ -403,7 +403,9 @@ module HTTP {
         /** Gets the mimetype of this HTTP response, if it can be statically determined. */
         string getMimetype() {
           exists(StrConst str |
-            DataFlow::localFlow(DataFlow::exprNode(str), this.getMimetypeOrContentTypeArg()) and
+            DataFlow::exprNode(str)
+                .(DataFlow::LocalSourceNode)
+                .flowsTo(this.getMimetypeOrContentTypeArg()) and
             result = str.getText().splitAt(";", 0)
           )
           or

python/ql/src/semmle/python/dataflow/new/TypeTracker.qll

@@ -46,7 +46,7 @@ class StepSummary extends TStepSummary {
 
 module StepSummary {
   cached
-  predicate step(Node nodeFrom, Node nodeTo, StepSummary summary) {
+  predicate step(LocalSourceNode nodeFrom, Node nodeTo, StepSummary summary) {
     exists(Node mid | typePreservingStep*(nodeFrom, mid) and smallstep(mid, nodeTo, summary))
   }
 
@@ -70,9 +70,8 @@ module StepSummary {
 
 /** Holds if it's reasonable to expect the data flow step from `nodeFrom` to `nodeTo` to preserve types. */
 private predicate typePreservingStep(Node nodeFrom, Node nodeTo) {
-  EssaFlow::essaFlowStep(nodeFrom, nodeTo) or
-  jumpStep(nodeFrom, nodeTo) or
-  nodeFrom = nodeTo.(PostUpdateNode).getPreUpdateNode()
+  simpleLocalFlowStep(nodeFrom, nodeTo) or
+  jumpStep(nodeFrom, nodeTo)
 }
 
 /** Holds if `nodeFrom` steps to `nodeTo` by being passed as a parameter in a call. */
@@ -115,11 +114,11 @@ predicate returnStep(ReturnNode nodeFrom, Node nodeTo) {
  * function. This means we will track the fact that `x.attr` can have the type of `y` into the
  * assignment to `z` inside `bar`, even though this attribute write happens _after_ `bar` is called.
  */
-predicate basicStoreStep(Node nodeFrom, Node nodeTo, string attr) {
+predicate basicStoreStep(Node nodeFrom, LocalSourceNode nodeTo, string attr) {
   exists(AttrWrite a |
     a.mayHaveAttributeName(attr) and
     nodeFrom = a.getValue() and
-    simpleLocalFlowStep*(nodeTo, a.getObject())
+    nodeTo.flowsTo(a.getObject())
   )
 }
 

python/ql/src/semmle/python/dataflow/new/internal/Attributes.qll

@@ -30,8 +30,8 @@ abstract class AttrRef extends Node {
   predicate mayHaveAttributeName(string attrName) {
     attrName = this.getAttributeName()
     or
-    exists(Node nodeFrom |
-      localFlow(nodeFrom, this.getAttributeNameExpr()) and
+    exists(LocalSourceNode nodeFrom |
+      nodeFrom.flowsTo(this.getAttributeNameExpr()) and
       attrName = nodeFrom.asExpr().(StrConst).getText()
     )
   }

python/ql/src/semmle/python/dataflow/new/internal/DataFlowPrivate.qll

@@ -181,6 +181,7 @@ module EssaFlow {
  * data flow. It is a strict subset of the `localFlowStep` predicate, as it
  * excludes SSA flow through instance fields.
  */
+cached
 predicate simpleLocalFlowStep(Node nodeFrom, Node nodeTo) {
   // If there is ESSA-flow out of a node `node`, we want flow
   // both out of `node` and any post-update node of `node`.

python/ql/src/semmle/python/dataflow/new/internal/DataFlowPublic.qll

@@ -351,6 +351,19 @@ class BarrierGuard extends GuardNode {
   }
 }
 
+/**
+ * A data flow node that is a source of local flow. This includes things like
+ * - Expressions
+ * - Function parameters
+ */
+class LocalSourceNode extends Node {
+  LocalSourceNode() { not simpleLocalFlowStep(_, this) }
+
+  /** Holds if this `LocalSourceNode` can flow to `nodeTo` in one or more local flow steps. */
+  cached
+  predicate flowsTo(Node nodeTo) { simpleLocalFlowStep*(this, nodeTo) }
+}
+
 /**
  * A reference contained in an object. This is either a field or a property.
  */

python/ql/src/semmle/python/frameworks/Django.qll

@@ -1641,7 +1641,7 @@ private module Django {
 
     DjangoRouteRegex() {
       this instanceof StrConst and
-      DataFlow::localFlow(DataFlow::exprNode(this), rePathCall.getUrlPatternArg())
+      DataFlow::exprNode(this).(DataFlow::LocalSourceNode).flowsTo(rePathCall.getUrlPatternArg())
     }
 
     DjangoRegexRouteSetup getRouteSetup() { result = rePathCall }

python/ql/src/semmle/python/frameworks/Flask.qll

@@ -319,9 +319,9 @@ private module FlaskModel {
     }
 
     override Function getARouteHandler() {
-      exists(DataFlow::Node view_func_arg, DataFlow::Node func_src |
+      exists(DataFlow::Node view_func_arg, DataFlow::LocalSourceNode func_src |
         view_func_arg.asCfgNode() in [node.getArg(2), node.getArgByName("view_func")] and
-        DataFlow::localFlow(func_src, view_func_arg) and
+        func_src.flowsTo(view_func_arg) and
         func_src.asExpr().(CallableExpr) = result.getDefinition()
       )
     }