@nxtedition/rocksdb 15.1.5 → 15.2.1

package/deps/re2/re2/testing/backtrack.cc

@@ -0,0 +1,274 @@
+// Copyright 2008 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Tested by search_test.cc, exhaustive_test.cc, tester.cc
+//
+// Prog::UnsafeSearchBacktrack is a backtracking regular expression search,
+// except that it remembers where it has been, trading a lot of
+// memory for a lot of time. It exists only for testing purposes.
+//
+// Let me repeat that.
+//
+// THIS CODE SHOULD NEVER BE USED IN PRODUCTION:
+//   - It uses a ton of memory.
+//   - It uses a ton of stack.
+//   - It uses ABSL_CHECK() and ABSL_LOG(FATAL).
+//   - It implements unanchored search by repeated anchored search.
+//
+// On the other hand, it is very simple and a good reference
+// implementation for the more complicated regexp packages.
+//
+// In BUILD, this file is linked into the ":testing" library,
+// not the main library, in order to make it harder to pick up
+// accidentally.
+
+#include <stddef.h>
+#include <stdint.h>
+#include <string.h>
+
+#include "absl/base/macros.h"
+#include "absl/log/absl_check.h"
+#include "absl/log/absl_log.h"
+#include "absl/strings/string_view.h"
+#include "re2/pod_array.h"
+#include "re2/prog.h"
+#include "re2/regexp.h"
+
+namespace re2 {
+
+// Backtracker holds the state for a backtracking search.
+//
+// Excluding the search parameters, the main search state
+// is just the "capture registers", which record, for the
+// current execution, the string position at which each
+// parenthesis was passed.  cap_[0] and cap_[1] are the
+// left and right parenthesis in $0, cap_[2] and cap_[3] in $1, etc.
+//
+// To avoid infinite loops during backtracking on expressions
+// like (a*)*, the visited_[] bitmap marks the (state, string-position)
+// pairs that have already been explored and are thus not worth
+// re-exploring if we get there via another path.  Modern backtracking
+// libraries engineer their program representation differently, to make
+// such infinite loops possible to avoid without keeping a giant visited_
+// bitmap, but visited_ works fine for a reference implementation
+// and it has the nice benefit of making the search run in linear time.
+class Backtracker {
+ public:
+  explicit Backtracker(Prog* prog);
+
+  bool Search(absl::string_view text, absl::string_view context, bool anchored,
+              bool longest, absl::string_view* submatch, int nsubmatch);
+
+ private:
+  // Explores from instruction id at string position p looking for a match.
+  // Returns true if found (so that caller can stop trying other possibilities).
+  bool Visit(int id, const char* p);
+
+  // Tries instruction id at string position p.
+  // Returns true if a match is found.
+  bool Try(int id, const char* p);
+
+  // Search parameters
+  Prog* prog_;                   // program being run
+  absl::string_view text_;       // text being searched
+  absl::string_view context_;    // greater context of text being searched
+  bool anchored_;                // whether search is anchored at text.begin()
+  bool longest_;                 // whether search wants leftmost-longest match
+  bool endmatch_;                // whether search must end at text.end()
+  absl::string_view* submatch_;  // submatches to fill in
+  int nsubmatch_;                // # of submatches to fill in
+
+  // Search state
+  const char* cap_[64];         // capture registers
+  PODArray<uint32_t> visited_;  // bitmap: (Inst*, char*) pairs visited
+
+  Backtracker(const Backtracker&) = delete;
+  Backtracker& operator=(const Backtracker&) = delete;
+};
+
+Backtracker::Backtracker(Prog* prog)
+  : prog_(prog),
+    anchored_(false),
+    longest_(false),
+    endmatch_(false),
+    submatch_(NULL),
+    nsubmatch_(0) {
+}
+
+// Runs a backtracking search.
+bool Backtracker::Search(absl::string_view text, absl::string_view context,
+                         bool anchored, bool longest,
+                         absl::string_view* submatch, int nsubmatch) {
+  text_ = text;
+  context_ = context;
+  if (context_.data() == NULL)
+    context_ = text;
+  if (prog_->anchor_start() && BeginPtr(text) > BeginPtr(context_))
+    return false;
+  if (prog_->anchor_end() && EndPtr(text) < EndPtr(context_))
+    return false;
+  anchored_ = anchored | prog_->anchor_start();
+  longest_ = longest | prog_->anchor_end();
+  endmatch_ = prog_->anchor_end();
+  submatch_ = submatch;
+  nsubmatch_ = nsubmatch;
+  ABSL_CHECK_LT(2*nsubmatch_, static_cast<int>(ABSL_ARRAYSIZE(cap_)));
+  memset(cap_, 0, sizeof cap_);
+
+  // We use submatch_[0] for our own bookkeeping,
+  // so it had better exist.
+  absl::string_view sp0;
+  if (nsubmatch < 1) {
+    submatch_ = &sp0;
+    nsubmatch_ = 1;
+  }
+  submatch_[0] = absl::string_view();
+
+  // Allocate new visited_ bitmap -- size is proportional
+  // to text, so have to reallocate on each call to Search.
+  int nvisited = prog_->size() * static_cast<int>(text.size()+1);
+  nvisited = (nvisited + 31) / 32;
+  visited_ = PODArray<uint32_t>(nvisited);
+  memset(visited_.data(), 0, nvisited*sizeof visited_[0]);
+
+  // Anchored search must start at text.begin().
+  if (anchored_) {
+    cap_[0] = text.data();
+    return Visit(prog_->start(), text.data());
+  }
+
+  // Unanchored search, starting from each possible text position.
+  // Notice that we have to try the empty string at the end of
+  // the text, so the loop condition is p <= text.end(), not p < text.end().
+  for (const char* p = text.data(); p <= text.data() + text.size(); p++) {
+    cap_[0] = p;
+    if (Visit(prog_->start(), p))  // Match must be leftmost; done.
+      return true;
+    // Avoid invoking undefined behavior (arithmetic on a null pointer)
+    // by simply not continuing the loop.
+    if (p == NULL)
+      break;
+  }
+  return false;
+}
+
+// Explores from instruction id at string position p looking for a match.
+// Return true if found (so that caller can stop trying other possibilities).
+bool Backtracker::Visit(int id, const char* p) {
+  // Check bitmap.  If we've already explored from here,
+  // either it didn't match or it did but we're hoping for a better match.
+  // Either way, don't go down that road again.
+  ABSL_CHECK(p <= text_.data() + text_.size());
+  int n = id * static_cast<int>(text_.size()+1) +
+          static_cast<int>(p-text_.data());
+  ABSL_CHECK_LT(n/32, visited_.size());
+  if (visited_[n/32] & (1 << (n&31)))
+    return false;
+  visited_[n/32] |= 1 << (n&31);
+
+  Prog::Inst* ip = prog_->inst(id);
+  if (Try(id, p)) {
+    if (longest_ && !ip->last())
+      Visit(id+1, p);
+    return true;
+  }
+  if (!ip->last())
+    return Visit(id+1, p);
+  return false;
+}
+
+// Tries instruction id at string position p.
+// Returns true if a match is found.
+bool Backtracker::Try(int id, const char* p) {
+  // Pick out byte at current position.  If at end of string,
+  // have to explore in hope of finishing a match.  Use impossible byte -1.
+  int c = -1;
+  if (p < text_.data() + text_.size())
+    c = *p & 0xFF;
+
+  Prog::Inst* ip = prog_->inst(id);
+  switch (ip->opcode()) {
+    default:
+      ABSL_LOG(FATAL) << "Unexpected opcode: " << ip->opcode();
+      return false;  // not reached
+
+    case kInstAltMatch:
+      // Ignored.
+      return false;
+
+    case kInstByteRange:
+      if (ip->Matches(c))
+        return Visit(ip->out(), p+1);
+      return false;
+
+    case kInstCapture:
+      if (0 <= ip->cap() &&
+          ip->cap() < static_cast<int>(ABSL_ARRAYSIZE(cap_))) {
+        // Capture p to register, but save old value.
+        const char* q = cap_[ip->cap()];
+        cap_[ip->cap()] = p;
+        bool ret = Visit(ip->out(), p);
+        // Restore old value as we backtrack.
+        cap_[ip->cap()] = q;
+        return ret;
+      }
+      return Visit(ip->out(), p);
+
+    case kInstEmptyWidth:
+      if (ip->empty() & ~Prog::EmptyFlags(context_, p))
+        return false;
+      return Visit(ip->out(), p);
+
+    case kInstNop:
+      return Visit(ip->out(), p);
+
+    case kInstMatch:
+      // We found a match.  If it's the best so far, record the
+      // parameters in the caller's submatch_ array.
+      if (endmatch_ && p != context_.data() + context_.size())
+        return false;
+      cap_[1] = p;
+      if (submatch_[0].data() == NULL ||
+          (longest_ && p > submatch_[0].data() + submatch_[0].size())) {
+        // First match so far - or better match.
+        for (int i = 0; i < nsubmatch_; i++)
+          submatch_[i] = absl::string_view(
+              cap_[2 * i], static_cast<size_t>(cap_[2 * i + 1] - cap_[2 * i]));
+      }
+      return true;
+
+    case kInstFail:
+      return false;
+  }
+}
+
+// Runs a backtracking search.
+bool Prog::UnsafeSearchBacktrack(absl::string_view text,
+                                 absl::string_view context, Anchor anchor,
+                                 MatchKind kind, absl::string_view* match,
+                                 int nmatch) {
+  // If full match, we ask for an anchored longest match
+  // and then check that match[0] == text.
+  // So make sure match[0] exists.
+  absl::string_view sp0;
+  if (kind == kFullMatch) {
+    anchor = kAnchored;
+    if (nmatch < 1) {
+      match = &sp0;
+      nmatch = 1;
+    }
+  }
+
+  // Run the search.
+  Backtracker b(this);
+  bool anchored = anchor == kAnchored;
+  bool longest = kind != kFirstMatch;
+  if (!b.Search(text, context, anchored, longest, match, nmatch))
+    return false;
+  if (kind == kFullMatch && EndPtr(match[0]) != EndPtr(text))
+    return false;
+  return true;
+}
+
+}  // namespace re2

package/deps/re2/re2/testing/charclass_test.cc

@@ -0,0 +1,228 @@
+// Copyright 2006 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Test character class manipulations.
+
+#include <stdio.h>
+
+#include "absl/base/macros.h"
+#include "absl/strings/str_format.h"
+#include "gtest/gtest.h"
+#include "re2/regexp.h"
+#include "util/utf.h"
+
+namespace re2 {
+
+struct CCTest {
+  struct {
+    Rune lo;
+    Rune hi;
+  } add[10];
+  int remove;
+  struct {
+    Rune lo;
+    Rune hi;
+  } final[10];
+};
+
+static CCTest tests[] = {
+  { { { 10, 20 }, {-1} }, -1,
+    { { 10, 20 }, {-1} } },
+
+  { { { 10, 20 }, { 20, 30 }, {-1} }, -1,
+    { { 10, 30 }, {-1} } },
+
+  { { { 10, 20 }, { 30, 40 }, { 20, 30 }, {-1} }, -1,
+    { { 10, 40 }, {-1} } },
+
+  { { { 0, 50 }, { 20, 30 }, {-1} }, -1,
+    { { 0, 50 }, {-1} } },
+
+  { { { 10, 11 }, { 13, 14 }, { 16, 17 }, { 19, 20 }, { 22, 23 }, {-1} }, -1,
+    { { 10, 11 }, { 13, 14 }, { 16, 17 }, { 19, 20 }, { 22, 23 }, {-1} } },
+
+  { { { 13, 14 }, { 10, 11 }, { 22, 23 }, { 19, 20 }, { 16, 17 }, {-1} }, -1,
+    { { 10, 11 }, { 13, 14 }, { 16, 17 }, { 19, 20 }, { 22, 23 }, {-1} } },
+
+  { { { 13, 14 }, { 10, 11 }, { 22, 23 }, { 19, 20 }, { 16, 17 }, {-1} }, -1,
+    { { 10, 11 }, { 13, 14 }, { 16, 17 }, { 19, 20 }, { 22, 23 }, {-1} } },
+
+  { { { 13, 14 }, { 10, 11 }, { 22, 23 }, { 19, 20 }, { 16, 17 }, { 5, 25 }, {-1} }, -1,
+    { { 5, 25 }, {-1} } },
+
+  { { { 13, 14 }, { 10, 11 }, { 22, 23 }, { 19, 20 }, { 16, 17 }, { 12, 21 }, {-1} }, -1,
+    { { 10, 23 }, {-1} } },
+
+  // These check boundary cases during negation.
+  { { { 0, Runemax }, {-1} }, -1,
+    { { 0, Runemax }, {-1} } },
+
+  { { { 0, 50 }, {-1} }, -1,
+    { { 0, 50 }, {-1} } },
+
+  { { { 50, Runemax }, {-1} }, -1,
+    { { 50, Runemax }, {-1} } },
+
+  // Check RemoveAbove.
+  { { { 50, Runemax }, {-1} }, 255,
+    { { 50, 255 }, {-1} } },
+
+  { { { 50, Runemax }, {-1} }, 65535,
+    { { 50, 65535 }, {-1} } },
+
+  { { { 50, Runemax }, {-1} }, Runemax,
+    { { 50, Runemax }, {-1} } },
+
+  { { { 50, 60 }, { 250, 260 }, { 350, 360 }, {-1} }, 255,
+    { { 50, 60 }, { 250, 255 }, {-1} } },
+
+  { { { 50, 60 }, {-1} }, 255,
+    { { 50, 60 }, {-1} } },
+
+  { { { 350, 360 }, {-1} }, 255,
+    { {-1} } },
+
+  { { {-1} }, 255,
+    { {-1} } },
+};
+
+template <typename CharClass>
+static void Broke(const char *desc, const CCTest* t, CharClass* cc) {
+  if (t == NULL) {
+    absl::PrintF("\t%s:", desc);
+  } else {
+    absl::PrintF("\n");
+    absl::PrintF("CharClass added: [%s]", desc);
+    for (int k = 0; t->add[k].lo >= 0; k++)
+      absl::PrintF(" %d-%d", t->add[k].lo, t->add[k].hi);
+    absl::PrintF("\n");
+    if (t->remove >= 0)
+      absl::PrintF("Removed > %d\n", t->remove);
+    absl::PrintF("\twant:");
+    for (int k = 0; t->final[k].lo >= 0; k++)
+      absl::PrintF(" %d-%d", t->final[k].lo, t->final[k].hi);
+    absl::PrintF("\n");
+    absl::PrintF("\thave:");
+  }
+
+  for (typename CharClass::iterator it = cc->begin(); it != cc->end(); ++it)
+    absl::PrintF(" %d-%d", it->lo, it->hi);
+  absl::PrintF("\n");
+}
+
+bool ShouldContain(CCTest *t, int x) {
+  for (int j = 0; t->final[j].lo >= 0; j++)
+    if (t->final[j].lo <= x && x <= t->final[j].hi)
+      return true;
+  return false;
+}
+
+// Helpers to make templated CorrectCC work with both CharClass and CharClassBuilder.
+
+CharClass* Negate(CharClass *cc) {
+  return cc->Negate();
+}
+
+void Delete(CharClass* cc) {
+  cc->Delete();
+}
+
+CharClassBuilder* Negate(CharClassBuilder* cc) {
+  CharClassBuilder* ncc = cc->Copy();
+  ncc->Negate();
+  return ncc;
+}
+
+void Delete(CharClassBuilder* cc) {
+  delete cc;
+}
+
+template <typename CharClass>
+bool CorrectCC(CharClass *cc, CCTest *t, const char *desc) {
+  typename CharClass::iterator it = cc->begin();
+  int size = 0;
+  for (int j = 0; t->final[j].lo >= 0; j++, ++it) {
+    if (it == cc->end() ||
+        it->lo != t->final[j].lo ||
+        it->hi != t->final[j].hi) {
+      Broke(desc, t, cc);
+      return false;
+    }
+    size += it->hi - it->lo + 1;
+  }
+  if (it != cc->end()) {
+    Broke(desc, t, cc);
+    return false;
+  }
+  if (cc->size() != size) {
+    Broke(desc, t, cc);
+    absl::PrintF("wrong size: want %d have %d\n", size, cc->size());
+    return false;
+  }
+
+  for (int j = 0; j < 101; j++) {
+    if (j == 100)
+      j = Runemax;
+    if (ShouldContain(t, j) != cc->Contains(j)) {
+      Broke(desc, t, cc);
+      absl::PrintF("want contains(%d)=%d, got %d\n",
+                   j, ShouldContain(t, j), cc->Contains(j));
+      return false;
+    }
+  }
+
+  CharClass* ncc = Negate(cc);
+  for (int j = 0; j < 101; j++) {
+    if (j == 100)
+      j = Runemax;
+    if (ShouldContain(t, j) == ncc->Contains(j)) {
+      Broke(desc, t, cc);
+      Broke("ncc", NULL, ncc);
+      absl::PrintF("want ncc contains(%d)!=%d, got %d\n",
+                   j, ShouldContain(t, j), ncc->Contains(j));
+      Delete(ncc);
+      return false;
+    }
+    if (ncc->size() != Runemax+1 - cc->size()) {
+      Broke(desc, t, cc);
+      Broke("ncc", NULL, ncc);
+      absl::PrintF("ncc size should be %d is %d\n",
+                   Runemax+1 - cc->size(), ncc->size());
+      Delete(ncc);
+      return false;
+    }
+  }
+  Delete(ncc);
+  return true;
+}
+
+TEST(TestCharClassBuilder, Adds) {
+  int nfail = 0;
+  for (size_t i = 0; i < ABSL_ARRAYSIZE(tests); i++) {
+    CharClassBuilder ccb;
+    CCTest* t = &tests[i];
+    for (int j = 0; t->add[j].lo >= 0; j++)
+      ccb.AddRange(t->add[j].lo, t->add[j].hi);
+    if (t->remove >= 0)
+      ccb.RemoveAbove(t->remove);
+    if (!CorrectCC(&ccb, t, "before copy (CharClassBuilder)"))
+      nfail++;
+    CharClass* cc = ccb.GetCharClass();
+    if (!CorrectCC(cc, t, "before copy (CharClass)"))
+      nfail++;
+    cc->Delete();
+
+    CharClassBuilder *ccb1 = ccb.Copy();
+    if (!CorrectCC(ccb1, t, "after copy (CharClassBuilder)"))
+      nfail++;
+    cc = ccb.GetCharClass();
+    if (!CorrectCC(cc, t, "after copy (CharClass)"))
+      nfail++;
+    cc->Delete();
+    delete ccb1;
+  }
+  EXPECT_EQ(nfail, 0);
+}
+
+}  // namespace re2