JS: Make the edges of API-graphs into IPA types

javascript/ql/lib/semmle/javascript/Constants.qll

@@ -3,10 +3,12 @@
  */
 
 import javascript
+private import semmle.javascript.internal.CachedStages
 
 /**
  * An expression that evaluates to a constant primitive value.
  */
+cached
 abstract class ConstantExpr extends Expr { }
 
 /**
@@ -16,15 +18,19 @@ module SyntacticConstants {
   /**
    * An expression that evaluates to a constant value according to a bottom-up syntactic analysis.
    */
+  cached
   abstract class SyntacticConstant extends ConstantExpr { }
 
   /**
    * A literal primitive expression.
    *
    * Note that `undefined`, `NaN` and `Infinity` are global variables, and are not covered by this class.
    */
+  cached
   class PrimitiveLiteralConstant extends SyntacticConstant {
+    cached
     PrimitiveLiteralConstant() {
+      Stages::Ast::ref() and
       this instanceof NumberLiteral
       or
       this instanceof StringLiteral
@@ -43,19 +49,27 @@ module SyntacticConstants {
   /**
    * A literal null expression.
    */
-  class NullConstant extends SyntacticConstant, NullLiteral { }
+  cached
+  class NullConstant extends SyntacticConstant, NullLiteral {
+    cached
+    NullConstant() { Stages::Ast::ref() and this = this }
+  }
 
   /**
    * A unary operation on a syntactic constant.
    */
+  cached
   class UnaryConstant extends SyntacticConstant, UnaryExpr {
+    cached
     UnaryConstant() { getOperand() instanceof SyntacticConstant }
   }
 
   /**
    * A binary operation on syntactic constants.
    */
+  cached
   class BinaryConstant extends SyntacticConstant, BinaryExpr {
+    cached
     BinaryConstant() {
       getLeftOperand() instanceof SyntacticConstant and
       getRightOperand() instanceof SyntacticConstant
@@ -65,7 +79,9 @@ module SyntacticConstants {
   /**
    * A conditional expression on syntactic constants.
    */
+  cached
   class ConditionalConstant extends SyntacticConstant, ConditionalExpr {
+    cached
     ConditionalConstant() {
       getCondition() instanceof SyntacticConstant and
       getConsequent() instanceof SyntacticConstant and
@@ -76,7 +92,9 @@ module SyntacticConstants {
   /**
    * A use of the global variable `undefined` or `void e`.
    */
+  cached
   class UndefinedConstant extends SyntacticConstant {
+    cached
     UndefinedConstant() {
       this.(GlobalVarAccess).getName() = "undefined" or
       this instanceof VoidExpr
@@ -86,21 +104,27 @@ module SyntacticConstants {
   /**
    * A use of the global variable `NaN`.
    */
+  cached
   class NaNConstant extends SyntacticConstant {
+    cached
     NaNConstant() { this.(GlobalVarAccess).getName() = "NaN" }
   }
 
   /**
    * A use of the global variable `Infinity`.
    */
+  cached
   class InfinityConstant extends SyntacticConstant {
+    cached
     InfinityConstant() { this.(GlobalVarAccess).getName() = "Infinity" }
   }
 
   /**
    * An expression that wraps the syntactic constant it evaluates to.
    */
+  cached
   class WrappedConstant extends SyntacticConstant {
+    cached
     WrappedConstant() { getUnderlyingValue() instanceof SyntacticConstant }
   }
 
@@ -123,6 +147,8 @@ module SyntacticConstants {
 /**
  * An expression that evaluates to a constant string.
  */
+cached
 class ConstantString extends ConstantExpr {
+  cached
   ConstantString() { exists(getStringValue()) }
 }

javascript/ql/lib/semmle/javascript/Expr.qll

@@ -89,7 +89,8 @@ class ExprOrType extends @expr_or_type, Documentable {
    *
    * Also see `getUnderlyingReference` and `stripParens`.
    */
-  Expr getUnderlyingValue() { result = this }
+  cached
+  Expr getUnderlyingValue() { Stages::Ast::ref() and result = this }
 }
 
 /**
@@ -274,7 +275,11 @@ private DataFlow::Node getCatchParameterFromStmt(Stmt stmt) {
  */
 class Identifier extends @identifier, ExprOrType {
   /** Gets the name of this identifier. */
-  string getName() { literals(result, _, this) }
+  cached
+  string getName() {
+    Stages::Ast::ref() and
+    literals(result, _, this)
+  }
 
   override string getAPrimaryQlClass() { result = "Identifier" }
 }

javascript/ql/lib/semmle/javascript/MembershipCandidates.qll

@@ -165,6 +165,7 @@ module MembershipCandidate {
     EnumerationRegExp enumeration;
     boolean polarity;
 
+    pragma[nomagic]
     RegExpEnumerationCandidate() {
       exists(DataFlow::MethodCallNode mcn, DataFlow::Node base, string m, DataFlow::Node firstArg |
         (

javascript/ql/lib/semmle/javascript/dataflow/Configuration.qll

@@ -939,18 +939,21 @@ private predicate basicFlowStepNoBarrier(
  * This predicate is field insensitive (it does not distinguish between `x` and `x.p`)
  * and hence should only be used for purposes of approximation.
  */
-pragma[inline]
+pragma[noinline]
 private predicate exploratoryFlowStep(
   DataFlow::Node pred, DataFlow::Node succ, DataFlow::Configuration cfg
 ) {
-  basicFlowStepNoBarrier(pred, succ, _, cfg) or
-  exploratoryLoadStep(pred, succ, cfg) or
-  isAdditionalLoadStoreStep(pred, succ, _, _, cfg) or
-  // the following three disjuncts taken together over-approximate flow through
-  // higher-order calls
-  exploratoryCallbackStep(pred, succ) or
-  succ = pred.(DataFlow::FunctionNode).getAParameter() or
-  exploratoryBoundInvokeStep(pred, succ)
+  isRelevantForward(pred, cfg) and
+  (
+    basicFlowStepNoBarrier(pred, succ, _, cfg) or
+    exploratoryLoadStep(pred, succ, cfg) or
+    isAdditionalLoadStoreStep(pred, succ, _, _, cfg) or
+    // the following three disjuncts taken together over-approximate flow through
+    // higher-order calls
+    exploratoryCallbackStep(pred, succ) or
+    succ = pred.(DataFlow::FunctionNode).getAParameter() or
+    exploratoryBoundInvokeStep(pred, succ)
+  )
 }
 
 /**
@@ -1024,6 +1027,7 @@ private string getAPropertyUsedInLoadStore(DataFlow::Configuration cfg) {
  * Holds if there exists a store-step from `pred` to `succ` under configuration `cfg`,
  * and somewhere in the program there exists a load-step that could possibly read the stored value.
  */
+pragma[noinline]
 private predicate exploratoryForwardStoreStep(
   DataFlow::Node pred, DataFlow::Node succ, DataFlow::Configuration cfg
 ) {
@@ -1075,8 +1079,10 @@ private string getABackwardsRelevantStoreProperty(DataFlow::Configuration cfg) {
 private predicate isRelevantForward(DataFlow::Node nd, DataFlow::Configuration cfg) {
   isSource(nd, cfg, _) and isLive()
   or
-  exists(DataFlow::Node mid | isRelevantForward(mid, cfg) |
-    exploratoryFlowStep(mid, nd, cfg) or
+  exists(DataFlow::Node mid |
+    exploratoryFlowStep(mid, nd, cfg)
+    or
+    isRelevantForward(mid, cfg) and
     exploratoryForwardStoreStep(mid, nd, cfg)
   )
 }
@@ -1098,11 +1104,10 @@ private predicate isRelevant(DataFlow::Node nd, DataFlow::Configuration cfg) {
 private predicate isRelevantBackStep(
   DataFlow::Node mid, DataFlow::Node nd, DataFlow::Configuration cfg
 ) {
+  exploratoryFlowStep(nd, mid, cfg)
+  or
   isRelevantForward(nd, cfg) and
-  (
-    exploratoryFlowStep(nd, mid, cfg) or
-    exploratoryBackwardStoreStep(nd, mid, cfg)
-  )
+  exploratoryBackwardStoreStep(nd, mid, cfg)
 }
 
 /**
@@ -1273,23 +1278,30 @@ private predicate parameterPropRead(
   DataFlow::Node arg, string prop, DataFlow::Node succ, DataFlow::Configuration cfg,
   PathSummary summary
 ) {
-  exists(Function f, DataFlow::Node read, DataFlow::Node invk |
+  exists(Function f, DataFlow::Node read, DataFlow::Node invk, DataFlow::Node parm |
+    reachesReturn(f, read, cfg, summary) and
+    parameterPropReadStep(parm, read, prop, cfg, arg, invk, f, succ)
+  )
+}
+
+// all the non-recursive parts of parameterPropRead outlined into a precomputed predicate
+pragma[noinline]
+private predicate parameterPropReadStep(
+  DataFlow::SourceNode parm, DataFlow::Node read, string prop, DataFlow::Configuration cfg,
+  DataFlow::Node arg, DataFlow::Node invk, Function f, DataFlow::Node succ
+) {
+  (
     not f.isAsyncOrGenerator() and invk = succ
     or
     // load from an immediately awaited function call
     f.isAsync() and
     invk = getAwaitOperand(succ)
-  |
-    exists(DataFlow::SourceNode parm |
-      callInputStep(f, invk, arg, parm, cfg) and
-      (
-        reachesReturn(f, read, cfg, summary) and
-        read = parm.getAPropertyRead(prop)
-        or
-        reachesReturn(f, read, cfg, summary) and
-        exists(DataFlow::Node use | parm.flowsTo(use) | isAdditionalLoadStep(use, read, prop, cfg))
-      )
-    )
+  ) and
+  callInputStep(f, invk, arg, parm, cfg) and
+  (
+    read = parm.getAPropertyRead(prop)
+    or
+    exists(DataFlow::Node use | parm.flowsTo(use) | isAdditionalLoadStep(use, read, prop, cfg))
   )
 }
 

javascript/ql/lib/semmle/javascript/dataflow/TaintTracking.qll

@@ -160,10 +160,12 @@ module TaintTracking {
    * of the standard library. Override `Configuration::isSanitizerGuard`
    * for analysis-specific taint sanitizer guards.
    */
+  cached
   abstract class AdditionalSanitizerGuardNode extends SanitizerGuardNode {
     /**
      * Holds if this guard applies to the flow in `cfg`.
      */
+    cached
     abstract predicate appliesTo(Configuration cfg);
   }
 
@@ -1127,7 +1129,7 @@ module TaintTracking {
         idx = astNode.getAnOperand() and
         idx.getPropertyNameExpr() = x and
         // and the other one is guaranteed to be `undefined`
-        forex(InferredType tp | tp = undef.getAType() | tp = TTUndefined())
+        unique(InferredType tp | tp = pragma[only_bind_into](undef.getAType())) = TTUndefined()
       )
     }
 

javascript/ql/lib/semmle/javascript/frameworks/D3.qll

@@ -23,7 +23,7 @@ module D3 {
     or
     result = API::moduleImport("d3-node").getInstance().getMember("d3")
     or
-    result = API::root().getASuccessor(any(D3GlobalEntry i))
+    result = any(D3GlobalEntry i).getNode()
   }
 
   /**

javascript/ql/lib/semmle/javascript/frameworks/History.qll

@@ -17,7 +17,7 @@ module History {
    * Gets a reference to the [`history`](https://npmjs.org/package/history) library.
    */
   private API::Node history() {
-    result = [API::moduleImport("history"), API::root().getASuccessor(any(HistoryGlobalEntry h))]
+    result = [API::moduleImport("history"), any(HistoryGlobalEntry h).getNode()]
   }
 
   /**

javascript/ql/lib/semmle/javascript/frameworks/Immutable.qll

@@ -27,7 +27,7 @@ private module Immutable {
   API::Node immutableImport() {
     result = API::moduleImport("immutable")
     or
-    result = API::root().getASuccessor(any(ImmutableGlobalEntry i))
+    result = any(ImmutableGlobalEntry i).getNode()
   }
 
   /**

javascript/ql/lib/semmle/javascript/frameworks/Logging.qll

@@ -45,7 +45,7 @@ private module Console {
    */
   private API::Node console() {
     result = API::moduleImport("console") or
-    result = API::root().getASuccessor(any(ConsoleGlobalEntry e))
+    result = any(ConsoleGlobalEntry e).getNode()
   }
 
   /**

javascript/ql/lib/semmle/javascript/frameworks/Nest.qll

@@ -151,7 +151,7 @@ module NestJS {
   private API::Node validationPipe() {
     result = nestjs().getMember("ValidationPipe")
     or
-    result = API::root().getASuccessor(any(ValidationNodeEntry e))
+    result = any(ValidationNodeEntry e).getNode()
   }
 
   /**

javascript/ql/lib/semmle/javascript/frameworks/Redux.qll

@@ -1111,9 +1111,7 @@ module Redux {
 
     /** A heuristic call to `connect`, recognized by it taking arguments named `mapStateToProps` and `mapDispatchToProps`. */
     private class HeuristicConnectFunction extends ConnectCall {
-      HeuristicConnectFunction() {
-        this = API::root().getASuccessor(any(HeuristicConnectEntryPoint e)).getACall()
-      }
+      HeuristicConnectFunction() { this = any(HeuristicConnectEntryPoint e).getNode().getACall() }
 
       override API::Node getMapStateToProps() {
         result = getAParameter() and

javascript/ql/lib/semmle/javascript/frameworks/Vue.qll

@@ -35,7 +35,7 @@ module Vue {
   API::Node vueLibrary() {
     result = API::moduleImport("vue")
     or
-    result = API::root().getASuccessor(any(GlobalVueEntryPoint e))
+    result = any(GlobalVueEntryPoint e).getNode()
   }
 
   /**
@@ -51,7 +51,7 @@ module Vue {
     or
     result = vueLibrary().getMember("component").getReturn()
     or
-    result = API::root().getASuccessor(any(VueFileImportEntryPoint e))
+    result = any(VueFileImportEntryPoint e).getNode()
   }
 
   /**

javascript/ql/lib/semmle/javascript/internal/CachedStages.qll

@@ -69,6 +69,14 @@ module Stages {
       exists(any(Expr e).getStringValue())
       or
       any(ASTNode node).isAmbient()
+      or
+      exists(any(Identifier e).getName())
+      or
+      exists(any(ExprOrType e).getUnderlyingValue())
+      or
+      exists(ConstantExpr e)
+      or
+      exists(SyntacticConstants::NullConstant n)
     }
   }
 
@@ -233,6 +241,43 @@ module Stages {
     }
   }
 
+  /**
+   * The `APIStage` stage.
+   */
+  cached
+  module APIStage {
+    /**
+     * Always holds.
+     * Ensures that a predicate is evaluated as part of the APIStage stage.
+     */
+    cached
+    predicate ref() { 1 = 1 }
+
+    /**
+     * DONT USE!
+     * Contains references to each predicate that use the above `ref` predicate.
+     */
+    cached
+    predicate backref() {
+      1 = 1
+      or
+      exists(
+        API::moduleImport("foo")
+            .getMember("bar")
+            .getUnknownMember()
+            .getAMember()
+            .getAParameter()
+            .getPromised()
+            .getReturn()
+            .getParameter(2)
+            .getUnknownMember()
+            .getInstance()
+            .getReceiver()
+            .getPromisedError()
+      )
+    }
+  }
+
   /**
    * The `taint` stage.
    */
@@ -262,6 +307,20 @@ module Stages {
       exists(Exports::getALibraryInputParameter())
       or
       any(RegExpTerm t).isUsedAsRegExp()
+      or
+      any(TaintTracking::AdditionalSanitizerGuardNode e).appliesTo(_)
+    }
+
+    cached
+    class DummySanitizer extends TaintTracking::AdditionalSanitizerGuardNode {
+      cached
+      DummySanitizer() { none() }
+
+      cached
+      override predicate appliesTo(TaintTracking::Configuration cfg) { none() }
+
+      cached
+      override predicate sanitizes(boolean outcome, Expr e) { none() }
     }
   }
 }

javascript/ql/lib/semmle/javascript/security/dataflow/ExternalAPIUsedWithUntrustedDataCustomizations.qll

@@ -211,11 +211,9 @@ module ExternalAPIUsedWithUntrustedData {
         node = getNamedParameter(base.getAParameter(), paramName) and
         result = basename + ".[callback].[param '" + paramName + "']"
         or
-        exists(string callbackName, string index |
-          node =
-            getNamedParameter(base.getASuccessor("parameter " + index).getMember(callbackName),
-              paramName) and
-          index != "-1" and // ignore receiver
+        exists(string callbackName, int index |
+          node = getNamedParameter(base.getParameter(index).getMember(callbackName), paramName) and
+          index != -1 and // ignore receiver
           result =
             basename + ".[callback " + index + " '" + callbackName + "'].[param '" + paramName +
               "']"

javascript/ql/lib/semmle/javascript/security/dataflow/RemoteFlowSources.qll

@@ -117,16 +117,20 @@ private class RemoteFlowSourceAccessPath extends JSONString {
   string getSourceType() { result = sourceType }
 
   /** Gets the `i`th component of the access path specifying this remote flow source. */
-  string getComponent(int i) {
+  API::Label::ApiLabel getComponent(int i) {
     exists(string raw | raw = this.getValue().splitAt(".", i + 1) |
       i = 0 and
-      result = "ExternalRemoteFlowSourceSpec " + raw
+      result =
+        API::Label::entryPoint(any(ExternalRemoteFlowSourceSpecEntryPoint e | e.getName() = raw))
       or
       i > 0 and
-      result = API::EdgeLabel::member(raw)
+      result = API::Label::member(raw)
     )
   }
 
+  /** Gets the first part of this access path. E.g. for "window.user.name" the result is "window". */
+  string getRootPath() { result = this.getValue().splitAt(".", 1) }
+
   /** Gets the index of the last component of this access path. */
   int getMaxComponentIndex() { result = max(int i | exists(this.getComponent(i))) }
 
@@ -154,10 +158,12 @@ private class ExternalRemoteFlowSourceSpecEntryPoint extends API::EntryPoint {
   string name;
 
   ExternalRemoteFlowSourceSpecEntryPoint() {
-    this = any(RemoteFlowSourceAccessPath s).getComponent(0) and
+    name = any(RemoteFlowSourceAccessPath s).getRootPath() and
     this = "ExternalRemoteFlowSourceSpec " + name
   }
 
+  string getName() { result = name }
+
   override DataFlow::SourceNode getAUse() { result = DataFlow::globalVarRef(name) }
 
   override DataFlow::Node getARhs() { none() }

javascript/ql/src/meta/ApiGraphs/ApiGraphEdges.ql

@@ -12,4 +12,4 @@ import javascript
 import meta.MetaMetrics
 
 select projectRoot(),
-  count(API::Node pred, string lbl, API::Node succ | succ = pred.getASuccessor(lbl))
+  count(API::Node pred, API::Label::ApiLabel lbl, API::Node succ | succ = pred.getASuccessor(lbl))

javascript/ql/test/ApiGraphs/VerifyAssertions.qll

@@ -62,7 +62,9 @@ class Assertion extends Comment {
     i = this.getPathLength() and
     result = API::root()
     or
-    result = this.lookup(i + 1).getASuccessor(this.getEdgeLabel(i))
+    result =
+      this.lookup(i + 1)
+          .getASuccessor(any(API::Label::ApiLabel label | label.toString() = this.getEdgeLabel(i)))
   }
 
   predicate isNegative() { polarity = "!" }
@@ -79,7 +81,11 @@ class Assertion extends Comment {
       then
         suffix =
           "it does have outgoing edges labelled " +
-            concat(string lbl | exists(nd.getASuccessor(lbl)) | lbl, ", ") + "."
+            concat(string lbl |
+              exists(nd.getASuccessor(any(API::Label::ApiLabel label | label.toString() = lbl)))
+            |
+              lbl, ", "
+            ) + "."
       else suffix = "it has no outgoing edges at all."
     |
       result = prefix + " " + suffix