Python: Support unpacking of keyword arguments.

yoff · yoff · commit 30d048f9d419 · 2020-09-30T11:55:27.000+02:00
diff --git a/python/ql/src/experimental/dataflow/internal/DataFlowPrivate.qll b/python/ql/src/experimental/dataflow/internal/DataFlowPrivate.qll
@@ -294,6 +294,12 @@ module ArgumentPassing {
         n = -2 and
         result = TKwOverflowNode(call, callable)
       )
+      or
+      // argument unpacked from dict
+      exists(string name |
+        call_unpacks(call, callable, name, n) and
+        result = TKwUnpacked(call, callable, name)
+      )
     )
   }
 
@@ -320,6 +326,22 @@ module ArgumentPassing {
       result = call.getArgByName(key)
     )
   }
+
+  /**
+   * Holds if `call` unpacks a dictionary argument in order to pass it via `name`.
+   * It will then be passed to the `n`th parameter of `callable`.
+   */
+  predicate call_unpacks(CallNode call, CallableValue callable, string name, int n) {
+    exists(Function f |
+      f = callable.getScope() and
+      not exists(call.getArg(n)) and // no positional arguement available
+      name = f.getArgName(n) and
+      not exists(call.getArgByName(name)) and // no keyword argument available
+      n >= 0 and
+      n < f.getPositionalParameterCount() + f.getKeywordOnlyParameterCount() and
+      exists(call.getNode().getKwargs()) // dict argument available
+    )
+  }
 }
 
 import ArgumentPassing
@@ -760,6 +782,8 @@ predicate readStep(Node nodeFrom, Content c, Node nodeTo) {
   comprehensionReadStep(nodeFrom, c, nodeTo)
   or
   attributeReadStep(nodeFrom, c, nodeTo)
+  or
+  kwUnpackReadStep(nodeFrom, c, nodeTo)
 }
 
 /** Data flows from a sequence to a subscript of the sequence. */
@@ -864,6 +888,19 @@ predicate attributeReadStep(CfgNode nodeFrom, AttributeContent c, CfgNode nodeTo
   )
 }
 
+/**
+ * Holds if `nodeFrom` is a dictionary argument being unpacked and `nodeTo` is the
+ * synthezised unpacked argument with the name indicated by `c`.
+ */
+predicate kwUnpackReadStep(CfgNode nodeFrom, DictionaryElementContent c, Node nodeTo) {
+  exists(CallNode call, CallableValue callable, string name, int n |
+    call_unpacks(call, callable, name, n) and
+    nodeFrom.asCfgNode() = call.getNode().getKwargs().getAFlowNode() and
+    nodeTo = TKwUnpacked(call, callable, name) and
+    name = c.getKey()
+  )
+}
+
 /**
  * Holds if values stored inside content `c` are cleared at node `n`. For example,
  * any value stored inside `f` is cleared at the pre-update node associated with `x`
diff --git a/python/ql/src/experimental/dataflow/internal/DataFlowPublic.qll b/python/ql/src/experimental/dataflow/internal/DataFlowPublic.qll
@@ -36,6 +36,10 @@ newtype TNode =
   /** A node representing the overflow keyword arguments to a call. */
   TKwOverflowNode(CallNode call, CallableValue callable) {
     exists(getKeywordOverflowArg(call, callable, _))
+  } or
+  /** A node representing an unpacked element of a dictionary argument. */
+  TKwUnpacked(CallNode call, CallableValue callable, string name) {
+    call_unpacks(call, callable, name, _)
   }
 
 /**
diff --git a/python/ql/test/experimental/dataflow/coverage/test.py b/python/ql/test/experimental/dataflow/coverage/test.py
@@ -378,7 +378,7 @@ def test_call_unpack_iterable():
 
 
 def test_call_unpack_mapping():
-    SINK(second(NONSOURCE, **{"b": SOURCE}))  # Flow missing
+    SINK(second(NONSOURCE, **{"b": SOURCE}))
 
 
 def f_extra_pos(a, *b):
@@ -474,7 +474,7 @@ def test_lambda_unpack_mapping():
     def second(a, b):
         return b
 
-    SINK(second(NONSOURCE, **{"b": SOURCE}))  # Flow missing
+    SINK(second(NONSOURCE, **{"b": SOURCE}))
 
 
 def test_lambda_extra_pos():
