Merge pull request #8149 from asgerf/shared/use-shared-access-path-syntax

Shared: use shared access path syntax to parse arguments in CSV rows

java/ql/lib/semmle/code/java/dataflow/internal/AccessPathSyntax.qll

@@ -6,6 +6,15 @@
  * (which does not use the shared data flow libraries).
  */
 
+/**
+ * Convenience-predicate for extracting two capture groups at once.
+ */
+bindingset[input, regexp]
+private predicate regexpCaptureTwo(string input, string regexp, string capture1, string capture2) {
+  capture1 = input.regexpCapture(regexp, 1) and
+  capture2 = input.regexpCapture(regexp, 2)
+}
+
 /** Companion module to the `AccessPath` class. */
 module AccessPath {
   /** A string that should be parsed as an access path. */
@@ -13,6 +22,95 @@ module AccessPath {
     bindingset[this]
     Range() { any() }
   }
+
+  /**
+   * Parses an integer constant `n` or interval `n1..n2` (inclusive) and gets the value
+   * of the constant or any value contained in the interval.
+   */
+  bindingset[arg]
+  int parseInt(string arg) {
+    result = arg.toInt()
+    or
+    // Match "n1..n2"
+    exists(string lo, string hi |
+      regexpCaptureTwo(arg, "(-?\\d+)\\.\\.(-?\\d+)", lo, hi) and
+      result = [lo.toInt() .. hi.toInt()]
+    )
+  }
+
+  /**
+   * Parses a lower-bounded interval `n..` and gets the lower bound.
+   */
+  bindingset[arg]
+  private int parseLowerBound(string arg) {
+    result = arg.regexpCapture("(-?\\d+)\\.\\.", 1).toInt()
+  }
+
+  /**
+   * Parses an integer constant or interval (bounded or unbounded) that explicitly
+   * references the arity, such as `N-1` or `N-3..N-1`.
+   *
+   * Note that expressions of form `N-x` will never resolve to a negative index,
+   * even if `N` is zero (it will have no result in that case).
+   */
+  bindingset[arg, arity]
+  private int parseIntWithExplicitArity(string arg, int arity) {
+    result >= 0 and // do not allow N-1 to resolve to a negative index
+    exists(string lo |
+      // N-x
+      lo = arg.regexpCapture("N-(\\d+)", 1) and
+      result = arity - lo.toInt()
+      or
+      // N-x..
+      lo = arg.regexpCapture("N-(\\d+)\\.\\.", 1) and
+      result = [arity - lo.toInt(), arity - 1]
+    )
+    or
+    exists(string lo, string hi |
+      // x..N-y
+      regexpCaptureTwo(arg, "(-?\\d+)\\.\\.N-(\\d+)", lo, hi) and
+      result = [lo.toInt() .. arity - hi.toInt()]
+      or
+      // N-x..N-y
+      regexpCaptureTwo(arg, "N-(\\d+)\\.\\.N-(\\d+)", lo, hi) and
+      result = [arity - lo.toInt() .. arity - hi.toInt()] and
+      result >= 0
+      or
+      // N-x..y
+      regexpCaptureTwo(arg, "N-(\\d+)\\.\\.(\\d+)", lo, hi) and
+      result = [arity - lo.toInt() .. hi.toInt()] and
+      result >= 0
+    )
+  }
+
+  /**
+   * Parses an integer constant or interval (bounded or unbounded) and gets any
+   * of the integers contained within (of which there may be infinitely many).
+   *
+   * Has no result for arguments involving an explicit arity, such as `N-1`.
+   */
+  bindingset[arg, result]
+  int parseIntUnbounded(string arg) {
+    result = parseInt(arg)
+    or
+    result >= parseLowerBound(arg)
+  }
+
+  /**
+   * Parses an integer constant or interval (bounded or unbounded) that
+   * may reference the arity of a call, such as `N-1` or `N-3..N-1`.
+   *
+   * Note that expressions of form `N-x` will never resolve to a negative index,
+   * even if `N` is zero (it will have no result in that case).
+   */
+  bindingset[arg, arity]
+  int parseIntWithArity(string arg, int arity) {
+    result = parseInt(arg)
+    or
+    result in [parseLowerBound(arg) .. arity - 1]
+    or
+    result = parseIntWithExplicitArity(arg, arity)
+  }
 }
 
 /** Gets the `n`th token on the access path as a string. */

java/ql/lib/semmle/code/java/dataflow/internal/FlowSummaryImplSpecific.qll

@@ -200,21 +200,10 @@ predicate interpretInputSpecific(string c, InterpretNode mid, InterpretNode n) {
   )
 }
 
-bindingset[s]
-private int parsePosition(string s) {
-  result = s.regexpCapture("([-0-9]+)", 1).toInt()
-  or
-  exists(int n1, int n2 |
-    s.regexpCapture("([-0-9]+)\\.\\.([0-9]+)", 1).toInt() = n1 and
-    s.regexpCapture("([-0-9]+)\\.\\.([0-9]+)", 2).toInt() = n2 and
-    result in [n1 .. n2]
-  )
-}
-
 /** Gets the argument position obtained by parsing `X` in `Parameter[X]`. */
 bindingset[s]
-ArgumentPosition parseParamBody(string s) { result = parsePosition(s) }
+ArgumentPosition parseParamBody(string s) { result = AccessPath::parseInt(s) }
 
 /** Gets the parameter position obtained by parsing `X` in `Argument[X]`. */
 bindingset[s]
-ParameterPosition parseArgBody(string s) { result = parsePosition(s) }
+ParameterPosition parseArgBody(string s) { result = AccessPath::parseInt(s) }