Cleanup regex compile optimize functions

cpp/src/strings/regex/regcomp.cpp

@@ -21,7 +21,9 @@
 
 #include <algorithm>
 #include <array>
-#include <cstring>
+#include <numeric>
+#include <stack>
+#include <string>
 
 namespace cudf {
 namespace strings {
@@ -862,8 +864,7 @@ class regex_compiler {
       ;  // "unmatched left paren";
     /* points to first and only operand */
     m_prog.set_start_inst(andstack[andstack.size() - 1].id_first);
-    m_prog.optimize1();
-    m_prog.optimize2();
+    m_prog.finalize();
     m_prog.check_for_errors();
     m_prog.set_groups_count(cursubid);
   }
@@ -880,81 +881,78 @@ reprog reprog::create_from(std::string_view pattern, regex_flags const flags)
   return rtn;
 }
 
-//
-void reprog::optimize1()
+void reprog::finalize()
 {
-  // Treat non-capturing LBRAs/RBRAs as NOOP
-  for (int i = 0; i < static_cast<int>(_insts.size()); i++) {
-    if (_insts[i].type == LBRA || _insts[i].type == RBRA) {
-      if (_insts[i].u1.subid < 1) { _insts[i].type = NOP; }
+  collapse_nops();
+  build_start_ids();
+}
+
+void reprog::collapse_nops()
+{
+  // treat non-capturing LBRAs/RBRAs as NOP
+  std::transform(_insts.begin(), _insts.end(), _insts.begin(), [](auto inst) {
+    if ((inst.type == LBRA || inst.type == RBRA) && (inst.u1.subid < 1)) { inst.type = NOP; }
+    return inst;
+  });
+
+  // functor for finding the next valid op
+  auto find_next_op = [insts = _insts](int id) {
+    while (insts[id].type == NOP) {
+      id = insts[id].u2.next_id;
     }
-  }
+    return id;
+  };
 
-  // get rid of NOP chains
-  for (int i = 0; i < insts_count(); i++) {
-    if (_insts[i].type != NOP) {
-      {
-        int target_id = _insts[i].u2.next_id;
-        while (_insts[target_id].type == NOP)
-          target_id = _insts[target_id].u2.next_id;
-        _insts[i].u2.next_id = target_id;
-      }
-      if (_insts[i].type == OR) {
-        int target_id = _insts[i].u1.right_id;
-        while (_insts[target_id].type == NOP)
-          target_id = _insts[target_id].u2.next_id;
-        _insts[i].u1.right_id = target_id;
-      }
+  // create new routes around NOP chains
+  std::transform(_insts.begin(), _insts.end(), _insts.begin(), [find_next_op](auto inst) {
+    if (inst.type != NOP) {
+      inst.u2.next_id = find_next_op(inst.u2.next_id);
+      if (inst.type == OR) { inst.u1.right_id = find_next_op(inst.u1.right_id); }
     }
-  }
-  // skip NOPs from the beginning
-  {
-    int target_id = _startinst_id;
-    while (_insts[target_id].type == NOP)
-      target_id = _insts[target_id].u2.next_id;
-    _startinst_id = target_id;
-  }
-  // actually remove the no-ops
+    return inst;
+  });
+
+  // find starting op
+  _startinst_id = find_next_op(_startinst_id);
+
+  // build a map of op ids
+  // these are used to fix up the ids after the NOPs are removed
   std::vector<int> id_map(insts_count());
-  int j = 0;  // compact the ops (non no-ops)
-  for (int i = 0; i < insts_count(); i++) {
-    id_map[i] = j;
-    if (_insts[i].type != NOP) {
-      _insts[j] = _insts[i];
-      j++;
-    }
-  }
-  _insts.resize(j);
-  // fix up the ORs
-  for (int i = 0; i < insts_count(); i++) {
-    {
-      int target_id        = _insts[i].u2.next_id;
-      _insts[i].u2.next_id = id_map[target_id];
-    }
-    if (_insts[i].type == OR) {
-      int target_id         = _insts[i].u1.right_id;
-      _insts[i].u1.right_id = id_map[target_id];
-    }
-  }
-  // set the new start id
+  std::transform_exclusive_scan(
+    _insts.begin(), _insts.end(), id_map.begin(), 0, std::plus<int>{}, [](auto inst) {
+      return static_cast<int>(inst.type != NOP);
+    });
+
+  // remove the NOP instructions
+  auto end = std::remove_if(_insts.begin(), _insts.end(), [](auto i) { return i.type == NOP; });
+  _insts.resize(std::distance(_insts.begin(), end));
+
+  // fix up the ids on the remaining instructions using the id_map
+  std::transform(_insts.begin(), _insts.end(), _insts.begin(), [id_map](auto inst) {
+    inst.u2.next_id = id_map[inst.u2.next_id];
+    if (inst.type == OR) { inst.u1.right_id = id_map[inst.u1.right_id]; }
+    return inst;
+  });
+
+  // fix up the start instruction id too
   _startinst_id = id_map[_startinst_id];
 }
 
 // expand leading ORs to multiple startinst_ids
-void reprog::optimize2()
+void reprog::build_start_ids()
 {
   _startinst_ids.clear();
-  std::vector<int> stack;
-  stack.push_back(_startinst_id);
-  while (!stack.empty()) {
-    int id = stack.back();
-    stack.pop_back();
+  std::stack<int> ids;
+  ids.push(_startinst_id);
+  while (!ids.empty()) {
+    int id = ids.top();
+    ids.pop();
     const reinst& inst = _insts[id];
     if (inst.type == OR) {
       if (inst.u2.left_id != id)  // prevents infinite while-loop here
-        stack.push_back(inst.u2.left_id);
+        ids.push(inst.u2.left_id);
       if (inst.u1.right_id != id)  // prevents infinite while-loop here
-        stack.push_back(inst.u1.right_id);
+        ids.push(inst.u1.right_id);
     } else {
       _startinst_ids.push_back(id);
     }

cpp/src/strings/regex/regcomp.h

@@ -114,8 +114,7 @@ class reprog {
   void set_start_inst(int32_t id);
   [[nodiscard]] int32_t get_start_inst() const;
 
-  void optimize1();
-  void optimize2();
+  void finalize();
   void check_for_errors();
 #ifndef NDEBUG
   void print(regex_flags const flags);
@@ -128,6 +127,8 @@ class reprog {
   std::vector<int32_t> _startinst_ids;  // short-cut to speed-up ORs
   int32_t _num_capturing_groups{};
 
+  void collapse_nops();
+  void build_start_ids();
   void check_for_errors(int32_t id, int32_t next_id);
 };