C++: Unify four implementations of repeatStars. Note that the recursive approach is faster for very large strings (well over 100 stars), while the concat approach appears to be faster for short strings and does not require an upper bound.

geoffw0 · geoffw0 · commit 76780d74d96e · 2024-03-25T18:10:21.000Z
diff --git a/cpp/ql/lib/semmle/code/cpp/dataflow/internal/FlowSummaryImpl.qll b/cpp/ql/lib/semmle/code/cpp/dataflow/internal/FlowSummaryImpl.qll
@@ -12,14 +12,6 @@ private import semmle.code.cpp.dataflow.ExternalFlow
 private import semmle.code.cpp.ir.IR
 
 module Input implements InputSig<DataFlowImplSpecific::CppDataFlow> {
-  /**
-   * Gets a string representing a level of indirection, for example for
-   * `indirection = 2`, the result is `**`.
-   */
-  private bindingset[indirection] string indirectionString(int indirection) {
-    result = concat(int i | i in [1 .. indirection] | "*")
-  }
-
   class SummarizedCallableBase = Function;
 
   ArgumentPosition callbackSelfParameterPosition() { result = TDirectPosition(-1) }
@@ -33,15 +25,15 @@ module Input implements InputSig<DataFlowImplSpecific::CppDataFlow> {
   string encodeReturn(ReturnKind rk, string arg) {
     rk != getStandardReturnValueKind() and
     result = "ReturnValue" and
-    arg = indirectionString(rk.(NormalReturnKind).getIndirectionIndex())
+    arg = repeatStars(rk.(NormalReturnKind).getIndirectionIndex())
   }
 
   string encodeContent(ContentSet cs, string arg) {
     exists(FieldContent c |
       cs.isSingleton(c) and
       // FieldContent indices have 0 for the address, 1 for content, so we need to subtract one.
       result = "Field" and
-      arg = indirectionString(c.getIndirectionIndex() - 1) + c.getField().getName()
+      arg = repeatStars(c.getIndirectionIndex() - 1) + c.getField().getName()
     )
   }
 
@@ -58,7 +50,7 @@ module Input implements InputSig<DataFlowImplSpecific::CppDataFlow> {
    */
   private bindingset[argString] TPosition decodePosition(string argString) {
     exists(int indirection, string posString, int pos |
-      argString = indirectionString(indirection) + posString and
+      argString = repeatStars(indirection) + posString and
       pos = AccessPath::parseInt(posString) and
       (
         pos >= 0 and indirection = 0 and result = TDirectPosition(pos)
@@ -98,7 +90,7 @@ module Input implements InputSig<DataFlowImplSpecific::CppDataFlow> {
       result.isSingleton(c) and
       token.getName() = c.getField().getName() and
       // FieldContent indices have 0 for the address, 1 for content, so we need to subtract one.
-      token.getAnArgument() = indirectionString(c.getIndirectionIndex() - 1)
+      token.getAnArgument() = repeatStars(c.getIndirectionIndex() - 1)
     )
   }
 }
diff --git a/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowPrivate.qll b/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowPrivate.qll
@@ -80,15 +80,6 @@ module NodeStars {
     result = n.(FinalParameterNode).getIndirectionIndex()
   }
 
-  private int maxNumberOfIndirections() { result = max(getNumberOfIndirections(_)) }
-
-  private string repeatStars(int n) {
-    n = 0 and result = ""
-    or
-    n = [1 .. maxNumberOfIndirections()] and
-    result = "*" + repeatStars(n - 1)
-  }
-
   /**
    * Gets the number of stars (i.e., `*`s) needed to produce the `toString`
    * output for `n`.
diff --git a/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll b/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowUtil.qll
@@ -2363,16 +2363,15 @@ class Content extends TContent {
   abstract predicate impliesClearOf(Content c);
 }
 
-private module ContentStars {
-  private int maxNumberOfIndirections() { result = max(any(Content c).getIndirectionIndex()) }
-
-  private string repeatStars(int n) {
-    n = 0 and result = ""
-    or
-    n = [1 .. maxNumberOfIndirections()] and
-    result = "*" + repeatStars(n - 1)
-  }
+/**
+ * Gets a string consisting of `n` star characters ("*"), where n >= 0. This is
+ * used to represent indirection.
+ */
+bindingset[n] string repeatStars(int n) {
+  result = concat(int i | i in [1 .. n] | "*")
+}
 
+private module ContentStars {
   /**
    * Gets the number of stars (i.e., `*`s) needed to produce the `toString`
    * output for `c`.
diff --git a/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/SsaInternals.qll b/cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/SsaInternals.qll
@@ -19,15 +19,6 @@ private module SourceVariables {
       ind = [0 .. countIndirectionsForCppType(base.getLanguageType()) + 1]
     }
 
-  private int maxNumberOfIndirections() { result = max(SourceVariable sv | | sv.getIndirection()) }
-
-  private string repeatStars(int n) {
-    n = 0 and result = ""
-    or
-    n = [1 .. maxNumberOfIndirections()] and
-    result = "*" + repeatStars(n - 1)
-  }
-
   class SourceVariable extends TSourceVariable {
     BaseSourceVariable base;
     int ind;
