Update and freeze pre-commit hooks

Closes #22

.gitignore

@@ -1,3 +1,2 @@
 .cache
-__pycache__
 build

.pre-commit-config.yaml

@@ -1,11 +1,11 @@
 repos:
   - repo: https://github.com/pre-commit/mirrors-clang-format
-    rev: 6d365699efc33b1b432eab5b4ae331a19e1857de # frozen: v18.1.2
+    rev: 6d365699efc33b1b432eab5b4ae331a19e1857de  # frozen: v18.1.2
     hooks:
       - id: clang-format
         exclude: ".*third_party/.*"
   - repo: https://github.com/cheshirekow/cmake-format-precommit
-    rev: e2c2116d86a80e72e7146a06e68b7c228afc6319 # frozen: v0.6.13
+    rev: e2c2116d86a80e72e7146a06e68b7c228afc6319  # frozen: v0.6.13
     hooks:
       - id: cmake-format
   - repo: local
@@ -13,7 +13,7 @@ repos:
       - id: debug verbose check
         name: disallow checking in DEBUG_VERBOSE=1
         description: disallow checking in DEBUG_VERBOSE=1
-        entry: "^#define DEBUG_VERBOSE 1$"
+        entry: '^#define DEBUG_VERBOSE 1$'
         language: pygrep
         types: [c++]
   - repo: local
@@ -21,14 +21,11 @@ repos:
       - id: debug verbose check
         name: disallow checking in SHOW_MEMORY=1
         description: disallow checking in SHOW_MEMORY=1
-        entry: "^#define SHOW_MEMORY 1$"
+        entry: '^#define SHOW_MEMORY 1$'
         language: pygrep
         types: [c++]
-  - repo: https://github.com/shellcheck-py/shellcheck-py
-    rev: a23f6b85d0fdd5bb9d564e2579e678033debbdff # frozen: v0.10.0.1
+  - repo: https://github.com/koalaman/shellcheck-precommit
+    rev: 2491238703a5d3415bb2b7ff11388bf775372f29  # frozen: v0.10.0
     hooks:
-      - id: shellcheck
-  - repo: https://github.com/psf/black
-    rev: 552baf822992936134cbd31a38f69c8cfe7c0f05 # frozen: 24.3.0
-    hooks:
-      - id: black
+    -   id: shellcheck
+#       args: ["--severity=warning"]  # Optionally only show errors and warnings

CMakeLists.txt

@@ -1,15 +1,13 @@
 cmake_minimum_required(VERSION 3.18)
 project(
   conflict-set
-  VERSION 0.0.5
+  VERSION 0.0.2
   DESCRIPTION
     "A data structure for optimistic concurrency control on ranges of bitwise-lexicographically-ordered keys."
   HOMEPAGE_URL "https://git.weaselab.dev/weaselab/conflict-set"
   LANGUAGES C CXX)
 set(CMAKE_CXX_STANDARD 20)
 
-file(WRITE ${CMAKE_BINARY_DIR}/version.txt ${PROJECT_VERSION})
-
 include(CMakePushCheckState)
 include(CheckCXXCompilerFlag)
 include(CheckIncludeFileCXX)
@@ -130,7 +128,7 @@ endif()
 if(APPLE)
   add_custom_command(
     TARGET ${PROJECT_NAME}-static
-    PRE_LINK
+    PRE_BUILD
     COMMAND ${CMAKE_SOURCE_DIR}/privatize_symbols_macos.sh
             $<TARGET_OBJECTS:${PROJECT_NAME}-object>)
 else()
@@ -138,8 +136,7 @@ else()
     TARGET ${PROJECT_NAME}-static
     POST_BUILD
     COMMAND
-      ${CMAKE_OBJCOPY}
-      --keep-global-symbols=${CMAKE_SOURCE_DIR}/symbol-exports.txt
+      ${CMAKE_OBJCOPY} --keep-global-symbols=${CMAKE_SOURCE_DIR}/symbols.txt
       $<TARGET_FILE:${PROJECT_NAME}-static> || echo
       "Proceeding with all symbols global in static library")
 endif()
@@ -261,22 +258,14 @@ if(BUILD_TESTING)
 
   # scripted tests. Written manually to fill in anything libfuzzer couldn't
   # find.
-  if(NOT CMAKE_CROSSCOMPILING)
-    find_package(Python3 REQUIRED COMPONENTS Interpreter)
-    set_property(
-      DIRECTORY
-      APPEND
-      PROPERTY CMAKE_CONFIGURE_DEPENDS ${CMAKE_SOURCE_DIR}/test_conflict_set.py)
-    execute_process(
-      COMMAND ${Python3_EXECUTABLE} ${CMAKE_SOURCE_DIR}/test_conflict_set.py
-              list OUTPUT_VARIABLE SCRIPT_TESTS)
-    foreach(TEST ${SCRIPT_TESTS})
-      add_test(
-        NAME script_test_${TEST}
-        COMMAND ${Python3_EXECUTABLE} ${CMAKE_SOURCE_DIR}/test_conflict_set.py
-                test ${TEST} --build-dir ${CMAKE_BINARY_DIR})
-    endforeach()
-  endif()
+  add_executable(script_test ScriptTest.cpp)
+  target_compile_options(script_test PRIVATE ${TEST_FLAGS})
+  target_link_libraries(script_test PRIVATE ${PROJECT_NAME})
+  file(GLOB SCRIPT_TESTS ${CMAKE_SOURCE_DIR}/script_tests/*)
+  foreach(TEST ${SCRIPT_TESTS})
+    get_filename_component(name ${TEST} NAME)
+    add_test(NAME conflict_set_script_${name} COMMAND script_test ${TEST})
+  endforeach()
 
   find_program(VALGRIND_EXE valgrind)
   if(VALGRIND_EXE AND NOT CMAKE_CROSSCOMPILING)
@@ -312,11 +301,7 @@ if(BUILD_TESTING)
       set(symbol_imports ${CMAKE_SOURCE_DIR}/apple-symbol-imports.txt)
     else()
       set(symbol_exports ${CMAKE_SOURCE_DIR}/symbol-exports.txt)
-      if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
-        set(symbol_imports ${CMAKE_SOURCE_DIR}/aarch64-symbol-imports.txt)
-      else()
-        set(symbol_imports ${CMAKE_SOURCE_DIR}/symbol-imports.txt)
-      endif()
+      set(symbol_imports ${CMAKE_SOURCE_DIR}/symbol-imports.txt)
     endif()
     add_test(
       NAME conflict_set_shared_symbols
@@ -343,7 +328,6 @@ endif()
 # packaging
 
 set(CPACK_PACKAGE_CONTACT andrew@weaselab.dev)
-set(CMAKE_INSTALL_DEFAULT_COMPONENT_NAME all)
 
 set(CPACK_PACKAGE_VENDOR "Weaselab")
 set(CPACK_RESOURCE_FILE_LICENSE "${CMAKE_CURRENT_SOURCE_DIR}/LICENSE")
@@ -357,26 +341,6 @@ set(CPACK_RPM_FILE_NAME RPM-DEFAULT)
 
 # deb
 set(CPACK_DEBIAN_FILE_NAME DEB-DEFAULT)
-# see *-imports.txt - dependency versions need to be synced with symbol versions
-if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
-  set(CPACK_DEBIAN_PACKAGE_DEPENDS "libc6 (>= 2.17)")
-else()
-  set(CPACK_DEBIAN_PACKAGE_DEPENDS "libc6 (>= 2.14)")
-endif()
-
-# macos
-set(CMAKE_OSX_DEPLOYMENT_TARGET 11.0)
-if(APPLE)
-  find_program(PANDOC_EXE pandoc)
-  if(PANDOC_EXE)
-    execute_process(COMMAND ${PANDOC_EXE} ${CMAKE_SOURCE_DIR}/README.md -o
-                            ${CMAKE_BINARY_DIR}/README.txt)
-    set(CPACK_RESOURCE_FILE_README ${CMAKE_BINARY_DIR}/README.txt)
-  endif()
-  configure_file(${CMAKE_SOURCE_DIR}/LICENSE ${CMAKE_BINARY_DIR}/LICENSE.txt
-                 COPYONLY)
-  set(CPACK_RESOURCE_FILE_LICENSE ${CMAKE_BINARY_DIR}/LICENSE.txt)
-endif()
 
 include(CPack)
 
@@ -402,11 +366,8 @@ install(
   ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
   LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
   RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
-
 install(DIRECTORY include/
         DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/${PROJECT_NAME})
 
 install(EXPORT ${PROJECT_NAME}Config
         DESTINATION ${CMAKE_INSTALL_DATAROOTDIR}/${PROJECT_NAME}/cmake)
-
-cpack_add_component(all)

ConflictSet.cpp

@@ -770,7 +770,7 @@ template <class NodeT> int getChildGeqSimd(NodeT *self, int child) {
 
   // cachegrind says the plain loop is fewer instructions and more mis-predicted
   // branches. Microbenchmark says plain loop is faster. It's written in this
-  // weird "generic" way though so that someday we can use the simd
+  // weird "generic" way though in case someday we can use the simd
   // implementation easily if we want.
   if constexpr (std::is_same_v<NodeT, Node3>) {
     Node3 *n = (Node3 *)self;
@@ -1146,14 +1146,8 @@ void freeAndMakeCapacityAtLeast(Node *&self, int capacity,
 // capacity.
 void maybeDecreaseCapacity(Node *&self, NodeAllocators *allocators,
                            ConflictSet::Impl *impl) {
-
   const int maxCapacity =
       (self->numChildren + int(self->entryPresent)) * (self->partialKeyLen + 1);
-#if DEBUG_VERBOSE && !defined(NDEBUG)
-  fprintf(stderr, "maybeDecreaseCapacity: current: %d, max: %d, key: %s\n",
-          self->getCapacity(), maxCapacity,
-          getSearchPathPrintable(self).c_str());
-#endif
   if (self->getCapacity() <= maxCapacity) {
     return;
   }
@@ -2395,11 +2389,7 @@ Iterator firstGeq(Node *n, const std::span<const uint8_t> key) {
       } else {
         n = nextSibling(n);
         if (n == nullptr) {
-          //  This line is genuinely unreachable from any entry point of the
-          //  final library, since we can't remove a key without introducing a
-          //  key after it, and the only production caller of firstGeq is for
-          //  resuming the setOldestVersion scan.
-          return {nullptr, 1}; // GCOVR_EXCL_LINE
+          return {nullptr, 1};
         }
         goto downLeftSpine;
       }
@@ -2520,10 +2510,6 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
   }
 
   explicit Impl(int64_t oldestVersion) : oldestVersion(oldestVersion) {
-#if DEBUG_VERBOSE
-    fprintf(stderr, "radix_tree: create\n");
-#endif
-
     // Insert ""
     root = allocators.node0.allocate(0);
     root->numChildren = 0;
@@ -2538,12 +2524,7 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     root->entry.pointVersion = oldestVersion;
     root->entry.rangeVersion = oldestVersion;
   }
-  ~Impl() {
-#if DEBUG_VERBOSE
-    fprintf(stderr, "radix_tree: destroy\n");
-#endif
-    destroyTree(root);
-  }
+  ~Impl() { destroyTree(root); }
 
   NodeAllocators allocators;
 
@@ -2597,16 +2578,17 @@ Node *&getInTree(Node *n, ConflictSet::Impl *impl) {
                               : getChildExists(n->parent, n->parentsIndex);
 }
 
-// Internal entry points. Public entry points should just delegate to these
+// ==================== END IMPLEMENTATION ====================
 
-void internal_check(ConflictSet::Impl *impl,
-                    const ConflictSet::ReadRange *reads,
-                    ConflictSet::Result *results, int count) {
-  impl->check(reads, results, count);
+// GCOVR_EXCL_START
+
+void ConflictSet::check(const ReadRange *reads, Result *results,
+                        int count) const {
+  return impl->check(reads, results, count);
 }
-void internal_addWrites(ConflictSet::Impl *impl,
-                        const ConflictSet::WriteRange *writes, int count,
-                        int64_t writeVersion) {
+
+void ConflictSet::addWrites(const WriteRange *writes, int count,
+                            int64_t writeVersion) {
   mallocBytesDelta = 0;
   impl->addWrites(writes, count, writeVersion);
   impl->totalBytes += mallocBytesDelta;
@@ -2617,7 +2599,7 @@ void internal_addWrites(ConflictSet::Impl *impl,
 #endif
 }
 
-void internal_setOldestVersion(ConflictSet::Impl *impl, int64_t oldestVersion) {
+void ConflictSet::setOldestVersion(int64_t oldestVersion) {
   mallocBytesDelta = 0;
   impl->setOldestVersion(oldestVersion);
   impl->totalBytes += mallocBytesDelta;
@@ -2627,47 +2609,19 @@ void internal_setOldestVersion(ConflictSet::Impl *impl, int64_t oldestVersion) {
   }
 #endif
 }
-ConflictSet::Impl *internal_create(int64_t oldestVersion) {
-  mallocBytesDelta = 0;
-  auto *result = new (safe_malloc(sizeof(ConflictSet::Impl)))
-      ConflictSet::Impl{oldestVersion};
-  result->totalBytes += mallocBytesDelta;
-  return result;
-}
 
-void internal_destroy(ConflictSet::Impl *impl) {
-  impl->~Impl();
-  safe_free(impl, sizeof(ConflictSet::Impl));
-}
-
-int64_t internal_getBytes(ConflictSet::Impl *impl) { return impl->totalBytes; }
-
-// ==================== END IMPLEMENTATION ====================
-
-// GCOVR_EXCL_START
-
-void ConflictSet::check(const ReadRange *reads, Result *results,
-                        int count) const {
-  internal_check(impl, reads, results, count);
-}
-
-void ConflictSet::addWrites(const WriteRange *writes, int count,
-                            int64_t writeVersion) {
-  internal_addWrites(impl, writes, count, writeVersion);
-}
-
-void ConflictSet::setOldestVersion(int64_t oldestVersion) {
-  internal_setOldestVersion(impl, oldestVersion);
-}
-
-int64_t ConflictSet::getBytes() const { return internal_getBytes(impl); }
+int64_t ConflictSet::getBytes() const { return impl->totalBytes; }
 
 ConflictSet::ConflictSet(int64_t oldestVersion)
-    : impl(internal_create(oldestVersion)) {}
+    : impl((mallocBytesDelta = 0,
+            new(safe_malloc(sizeof(Impl))) Impl{oldestVersion})) {
+  impl->totalBytes += mallocBytesDelta;
+}
 
 ConflictSet::~ConflictSet() {
   if (impl) {
-    internal_destroy(impl);
+    impl->~Impl();
+    safe_free(impl, sizeof(*impl));
   }
 }
 
@@ -2688,27 +2642,40 @@ extern "C" {
 __attribute__((__visibility__("default"))) void
 ConflictSet_check(void *cs, const ConflictSet_ReadRange *reads,
                   ConflictSet_Result *results, int count) {
-  internal_check((ConflictSet::Impl *)cs, reads, results, count);
+  ((ConflictSet::Impl *)cs)->check(reads, results, count);
 }
 __attribute__((__visibility__("default"))) void
 ConflictSet_addWrites(void *cs, const ConflictSet_WriteRange *writes, int count,
                       int64_t writeVersion) {
-  internal_addWrites((ConflictSet::Impl *)cs, writes, count, writeVersion);
+  auto *impl = ((ConflictSet::Impl *)cs);
+  mallocBytesDelta = 0;
+  impl->addWrites(writes, count, writeVersion);
+  impl->totalBytes += mallocBytesDelta;
 }
 __attribute__((__visibility__("default"))) void
 ConflictSet_setOldestVersion(void *cs, int64_t oldestVersion) {
-  internal_setOldestVersion((ConflictSet::Impl *)cs, oldestVersion);
+  auto *impl = ((ConflictSet::Impl *)cs);
+  mallocBytesDelta = 0;
+  impl->setOldestVersion(oldestVersion);
+  impl->totalBytes += mallocBytesDelta;
 }
 __attribute__((__visibility__("default"))) void *
 ConflictSet_create(int64_t oldestVersion) {
-  return internal_create(oldestVersion);
+  mallocBytesDelta = 0;
+  auto *result = new (safe_malloc(sizeof(ConflictSet::Impl)))
+      ConflictSet::Impl{oldestVersion};
+  result->totalBytes += mallocBytesDelta;
+  return result;
 }
 __attribute__((__visibility__("default"))) void ConflictSet_destroy(void *cs) {
-  internal_destroy((ConflictSet::Impl *)cs);
+  using Impl = ConflictSet::Impl;
+  ((Impl *)cs)->~Impl();
+  safe_free(cs, sizeof(Impl));
 }
 __attribute__((__visibility__("default"))) int64_t
 ConflictSet_getBytes(void *cs) {
-  return internal_getBytes((ConflictSet::Impl *)cs);
+  using Impl = ConflictSet::Impl;
+  return ((Impl *)cs)->totalBytes;
 }
 }
 
@@ -2944,6 +2911,10 @@ Iterator firstGeq(Node *n, std::string_view key) {
 
 } // namespace
 
+namespace std {
+void __throw_length_error(const char *) { __builtin_unreachable(); }
+} // namespace std
+
 #if SHOW_MEMORY
 
 int64_t nodeBytes = 0;
@@ -3024,7 +2995,6 @@ void removeKey(Node *n) {
 
 struct __attribute__((visibility("default"))) PeakPrinter {
   ~PeakPrinter() {
-    printf("--- radix_tree ---\n");
     printf("malloc bytes: %g\n", double(mallocBytes));
     printf("Peak malloc bytes: %g\n", double(peakMallocBytes));
     printf("Node bytes: %g\n", double(nodeBytes));

Internal.h

@@ -99,7 +99,8 @@ __attribute__((always_inline)) inline void safe_free(void *p, size_t s) {
   mallocBytesDelta -= s;
 #if SHOW_MEMORY
   mallocBytes -= s;
-#endif
+  free(p);
+#else
 #ifndef NDEBUG
   (char *&)p -= kMallocHeaderSize;
   size_t expected;
@@ -107,6 +108,7 @@ __attribute__((always_inline)) inline void safe_free(void *p, size_t s) {
   assert(s == expected);
 #endif
   free(p);
+#endif
 }
 
 // ==================== BEGIN ARENA IMPL ====================
@@ -255,65 +257,10 @@ template <class T> struct ArenaAlloc {
   void deallocate(T *, size_t) noexcept {}
 };
 
-template <class T> struct Vector {
-  static_assert(std::is_trivially_destructible_v<T>);
-  static_assert(std::is_trivially_copyable_v<T>);
-
-  explicit Vector(Arena *arena)
-      : arena(arena), t(nullptr), size_(0), capacity(0) {}
-
-  void append(std::span<const T> slice) {
-    if (size_ + int(slice.size()) > capacity) {
-      grow(std::max<int>(size_ + slice.size(), capacity * 2));
-    }
-    if (slice.size() > 0) {
-      memcpy(const_cast<std::remove_const_t<T> *>(t) + size_, slice.data(),
-             slice.size() * sizeof(T));
-    }
-    size_ += slice.size();
-  }
-
-  void push_back(const T &t) { append(std::span<const T>(&t, 1)); }
-
-  T *begin() { return t; }
-  T *end() { return t + size_; }
-  T *data() { return t; }
-  T &back() {
-    assert(size_ > 0);
-    return t[size_ - 1];
-  }
-  T &operator[](int i) {
-    assert(i >= 0 && i < size_);
-    return t[i];
-  }
-
-  void pop_back() {
-    assert(size_ > 0);
-    --size_;
-  }
-
-  int size() const { return size_; }
-
-  operator std::span<const T>() const { return std::span(t, size_); }
-
-private:
-  void grow(int newCapacity) {
-    capacity = newCapacity;
-    auto old = std::span<const T>(*this);
-    t = (T *)new (std::align_val_t(alignof(T)), *arena)
-        uint8_t[capacity * sizeof(T)];
-    size_ = 0;
-    append(old);
-  }
-
-  Arena *arena;
-  T *t;
-  int size_;
-  int capacity;
-};
-
-template <class T> auto vector(Arena &arena) { return Vector<T>(&arena); }
-
+template <class T> using Vector = std::vector<T, ArenaAlloc<T>>;
+template <class T> auto vector(Arena &arena) {
+  return Vector<T>(ArenaAlloc<T>(&arena));
+}
 template <class T, class C> using Set = std::set<T, C, ArenaAlloc<T>>;
 template <class T, class C = std::less<T>> auto set(Arena &arena) {
   return Set<T, C>(ArenaAlloc<T>(&arena));
@@ -578,18 +525,15 @@ template <class ConflictSetImpl> struct TestDriver {
   explicit TestDriver(const uint8_t *data, size_t size)
       : arbitrary({data, size}) {}
 
-  int64_t writeVersion = 100;
+  int64_t writeVersion = 0;
   int64_t oldestVersion = 0;
   ConflictSetImpl cs{oldestVersion};
   ReferenceImpl refImpl{oldestVersion};
 
-  constexpr static auto kMaxKeySuffixLen = 8;
+  constexpr static auto kMaxKeyLen = 8;
 
   bool ok = true;
 
-  const int prefixLen = arbitrary.bounded(512);
-  const int prefixByte = arbitrary.randT<uint8_t>();
-
   // Call until it returns true, for "done". Check internal invariants etc
   // between calls to next.
   bool next() {
@@ -600,7 +544,7 @@ template <class ConflictSetImpl> struct TestDriver {
     {
       int numPointWrites = arbitrary.bounded(100);
       int numRangeWrites = arbitrary.bounded(100);
-      int64_t v = (writeVersion += arbitrary.bounded(10));
+      int64_t v = ++writeVersion;
       auto *writes =
           new (arena) ConflictSet::WriteRange[numPointWrites + numRangeWrites];
       auto keys = set<std::string_view>(arena);
@@ -608,10 +552,9 @@ template <class ConflictSetImpl> struct TestDriver {
         if (!arbitrary.hasEntropy()) {
           return true;
         }
-        int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
+        int keyLen = arbitrary.bounded(kMaxKeyLen);
         auto *begin = new (arena) uint8_t[keyLen];
-        memset(begin, prefixByte, prefixLen);
-        arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
+        arbitrary.randomBytes(begin, keyLen);
         keys.insert(std::string_view((const char *)begin, keyLen));
       }
 
@@ -660,8 +603,8 @@ template <class ConflictSetImpl> struct TestDriver {
 
       refImpl.addWrites(writes, numPointWrites + numRangeWrites, v);
 
-      oldestVersion =
-          std::min(writeVersion - 10, oldestVersion + arbitrary.bounded(10));
+      oldestVersion = std::max<int64_t>(writeVersion - arbitrary.bounded(10),
+                                        oldestVersion);
       cs.setOldestVersion(oldestVersion);
       refImpl.setOldestVersion(oldestVersion);
     }
@@ -676,10 +619,9 @@ template <class ConflictSetImpl> struct TestDriver {
         if (!arbitrary.hasEntropy()) {
           return true;
         }
-        int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
+        int keyLen = arbitrary.bounded(kMaxKeyLen);
         auto *begin = new (arena) uint8_t[keyLen];
-        memset(begin, prefixByte, prefixLen);
-        arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
+        arbitrary.randomBytes(begin, keyLen);
         keys.insert(std::string_view((const char *)begin, keyLen));
       }
 

Jenkinsfile

@@ -111,9 +111,6 @@ pipeline {
                 gcovr -f ConflictSet.cpp --cobertura > build/coverage.xml
                 '''
                 cobertura autoUpdateHealth: false, autoUpdateStability: false, coberturaReportFile: 'build/coverage.xml', conditionalCoverageTargets: '70, 0, 0', failUnhealthy: false, failUnstable: false, lineCoverageTargets: '80, 0, 0', maxNumberOfBuilds: 0, methodCoverageTargets: '80, 0, 0', onlyStable: false, sourceEncoding: 'ASCII', zoomCoverageChart: false
-                sh '''
-                gcovr -f ConflictSet.cpp --fail-under-line 100 > /dev/null
-                '''
             }
         }
     }

README.md

@@ -58,27 +58,27 @@ Performance counters:
 
 ## Skip list
 
-|  ns/op |         op/s | err% | total | benchmark                           |
-| -----: | -----------: | ---: | ----: | :---------------------------------- |
-| 246.99 | 4,048,700.59 | 0.2% |  0.01 | `point reads`                       |
-| 260.16 | 3,843,784.65 | 0.1% |  0.01 | `prefix reads`                      |
-| 493.35 | 2,026,953.19 | 0.1% |  0.01 | `range reads`                       |
-| 462.05 | 2,164,289.23 | 0.6% |  0.01 | `point writes`                      |
-| 448.19 | 2,231,205.25 | 0.9% |  0.01 | `prefix writes`                     |
-| 255.83 | 3,908,845.72 | 1.5% |  0.02 | `range writes`                      |
-| 582.63 | 1,716,349.02 | 1.3% |  0.01 | `monotonic increasing point writes` |
+|               ns/op |                op/s |    err% |     total | benchmark
+|--------------------:|--------------------:|--------:|----------:|:----------
+|              246.99 |        4,048,700.59 |    0.2% |      0.01 | `point reads`
+|              260.16 |        3,843,784.65 |    0.1% |      0.01 | `prefix reads`
+|              493.35 |        2,026,953.19 |    0.1% |      0.01 | `range reads`
+|              462.05 |        2,164,289.23 |    0.6% |      0.01 | `point writes`
+|              448.19 |        2,231,205.25 |    0.9% |      0.01 | `prefix writes`
+|              255.83 |        3,908,845.72 |    1.5% |      0.02 | `range writes`
+|              582.63 |        1,716,349.02 |    1.3% |      0.01 | `monotonic increasing point writes`
 
 ## Radix tree (this implementation)
 
-|  ns/op |          op/s | err% | total | benchmark                           |
-| -----: | ------------: | ---: | ----: | :---------------------------------- |
-|  19.42 | 51,483,206.67 | 0.3% |  0.01 | `point reads`                       |
-|  58.43 | 17,115,612.57 | 0.1% |  0.01 | `prefix reads`                      |
-| 216.09 |  4,627,766.60 | 0.2% |  0.01 | `range reads`                       |
-|  28.35 | 35,267,567.72 | 0.2% |  0.01 | `point writes`                      |
-|  43.43 | 23,026,226.17 | 0.2% |  0.01 | `prefix writes`                     |
-|  50.00 | 20,000,000.00 | 0.0% |  0.01 | `range writes`                      |
-|  92.38 | 10,824,863.69 | 4.1% |  0.01 | `monotonic increasing point writes` |
+|               ns/op |                op/s |    err% |     total | benchmark
+|--------------------:|--------------------:|--------:|----------:|:----------
+|               19.42 |       51,483,206.67 |    0.3% |      0.01 | `point reads`
+|               58.43 |       17,115,612.57 |    0.1% |      0.01 | `prefix reads`
+|              216.09 |        4,627,766.60 |    0.2% |      0.01 | `range reads`
+|               28.35 |       35,267,567.72 |    0.2% |      0.01 | `point writes`
+|               43.43 |       23,026,226.17 |    0.2% |      0.01 | `prefix writes`
+|               50.00 |       20,000,000.00 |    0.0% |      0.01 | `range writes`
+|               92.38 |       10,824,863.69 |    4.1% |      0.01 | `monotonic increasing point writes`
 
 # "Real data" test
 

ScriptTest.cpp

@@ -0,0 +1,155 @@
+#include "Internal.h"
+#include <ConflictSet.h>
+
+#include <chrono>
+#include <cstring>
+#include <fcntl.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <vector>
+
+inline size_t getPageSize() {
+  static size_t kPageSize = sysconf(_SC_PAGESIZE);
+  return kPageSize;
+}
+
+/// Helper for rounding up to page size (or some other alignment)
+constexpr inline size_t rightAlign(size_t offset, size_t alignment) {
+  return offset % alignment == 0 ? offset
+                                 : ((offset / alignment) + 1) * alignment;
+}
+
+using StringView = std::basic_string_view<uint8_t>;
+
+inline StringView operator"" _v(const char *str, size_t size) {
+  return {reinterpret_cast<const uint8_t *>(str), size};
+}
+
+int main(int argc, const char **argv) {
+
+  ConflictSet cs{0};
+  ReferenceImpl ref{0};
+
+  for (int i = 1; i < argc; ++i) {
+    int fd = open(argv[i], O_RDONLY);
+    struct stat st;
+    if (fstat(fd, &st) == -1) {
+      int err = errno;
+      fprintf(stderr, "stat error %s - %s\n", argv[i], strerror(err));
+      fflush(stderr);
+      abort();
+    }
+
+    int64_t size = rightAlign(st.st_size, getPageSize());
+    const uint8_t *begin =
+        (uint8_t *)mmap(0, size, PROT_READ, MAP_PRIVATE, fd, 0);
+    madvise((void *)begin, size, MADV_SEQUENTIAL);
+    auto *const mapOriginal = begin;
+    const auto sizeOriginal = size;
+
+    StringView b;
+    StringView e;
+    int64_t v = 0;
+    int64_t lastWriteVersion = 0;
+    int64_t lastOldestVersion = 0;
+    std::vector<ConflictSet::WriteRange> writeRanges;
+    std::vector<ConflictSet::ReadRange> readRanges;
+    std::vector<ConflictSet::Result> results;
+
+    for (uint8_t *end = (uint8_t *)memchr(begin, '\n', size); end != nullptr;) {
+      StringView line{begin, static_cast<size_t>(end - begin)};
+      size -= end - begin + 1;
+      begin = end + 1;
+      end = (uint8_t *)memchr(begin, '\n', size);
+
+      if (line.starts_with("begin"_v)) {
+        b = line.substr("begin "_v.size(), line.size());
+        printf("b <- %.*s\n", int(b.size()), b.data());
+      } else if (line.starts_with("end"_v)) {
+        e = line.substr("end "_v.size(), line.size());
+        printf("e <- %.*s\n", int(e.size()), e.data());
+      } else if (line.starts_with("version"_v)) {
+        line = line.substr("version "_v.size(), line.size());
+        v = 0;
+        for (auto c : line) {
+          v = v * 10 + int(c) - int('0');
+        }
+        printf("v <- %" PRId64 "\n", v);
+      } else if (line.starts_with("pointread"_v)) {
+        printf("pointread\n");
+        ConflictSet::ReadRange r;
+        r.begin.p = b.data();
+        r.begin.len = b.size();
+        r.end.len = 0;
+        r.readVersion = v;
+        readRanges.push_back(r);
+      } else if (line.starts_with("pointwrite"_v)) {
+        printf("pointwrite\n");
+        assert(writeRanges.empty() ||
+               (writeRanges.back().end.len == 0 ? writeRanges.back().begin
+                                                : writeRanges.back().end) < b);
+        ConflictSet::WriteRange w;
+        w.begin.p = b.data();
+        w.begin.len = b.size();
+        w.end.len = 0;
+        writeRanges.push_back(w);
+      } else if (line.starts_with("rangeread"_v)) {
+        printf("rangeread\n");
+        ConflictSet::ReadRange r;
+        r.begin.p = b.data();
+        r.begin.len = b.size();
+        r.end.p = e.data();
+        r.end.len = e.size();
+        r.readVersion = v;
+        readRanges.push_back(r);
+      } else if (line.starts_with("rangewrite"_v)) {
+        printf("rangewrite\n");
+        assert(b < e);
+        assert(writeRanges.empty() ||
+               (writeRanges.back().end.len == 0 ? writeRanges.back().begin
+                                                : writeRanges.back().end) < b);
+        ConflictSet::WriteRange w;
+        w.begin.p = b.data();
+        w.begin.len = b.size();
+        w.end.p = e.data();
+        w.end.len = e.size();
+        writeRanges.push_back(w);
+      } else if (line.starts_with("check"_v)) {
+        printf("check\n");
+        Arena arena;
+        auto *expected = new (arena) ConflictSet::Result[readRanges.size()];
+        auto *actual = new (arena) ConflictSet::Result[readRanges.size()];
+        ref.check(readRanges.data(), expected, readRanges.size());
+        cs.check(readRanges.data(), actual, readRanges.size());
+        for (int i = 0; i < int(readRanges.size()); ++i) {
+          if (expected[i] != actual[i]) {
+            fprintf(stderr, "Expected %s, got %s at index %d\n",
+                    resultToStr(expected[i]), resultToStr(actual[i]), i);
+            return 1;
+          }
+        }
+        readRanges = {};
+      } else if (line.starts_with("addwrites"_v)) {
+        printf("addwrites\n");
+        assert(v > lastWriteVersion);
+        lastWriteVersion = v;
+        cs.addWrites(writeRanges.data(), writeRanges.size(), v);
+        ref.addWrites(writeRanges.data(), writeRanges.size(), v);
+        writeRanges = {};
+      } else if (line.starts_with("setoldest"_v)) {
+        printf("setoldest\n");
+        assert(v > lastOldestVersion);
+        lastOldestVersion = v;
+        cs.setOldestVersion(v);
+        ref.setOldestVersion(v);
+      } else if (line.empty() || line.starts_with(";"_v)) {
+        // skip
+      } else {
+        printf("Unrecognized line: %.*s\n", int(line.size()), line.data());
+      }
+    }
+    munmap((void *)mapOriginal, sizeOriginal);
+    close(fd);
+  }
+}

SkipList.cpp

@@ -16,8 +16,6 @@
  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  * See the License for the specific language governing permissions and
  * limitations under the License.
- *
- * This source code is modified to compile outside of FoundationDB
  */
 
 #include "ConflictSet.h"
@@ -270,21 +268,13 @@ public:
   }
 
   explicit SkipList(Version version = 0) {
-#if DEBUG_VERBOSE
-    fprintf(stderr, "skip_list: create\n");
-#endif
     header = Node::create(StringRef(), MaxLevels - 1);
     for (int l = 0; l < MaxLevels; l++) {
       header->setNext(l, nullptr);
       header->setMaxVersion(l, version);
     }
   }
-  ~SkipList() {
-#if DEBUG_VERBOSE
-    fprintf(stderr, "skip_list: destroy\n");
-#endif
-    destroy();
-  }
+  ~SkipList() { destroy(); }
   SkipList(SkipList &&other) noexcept : header(other.header) {
     other.header = nullptr;
   }
@@ -613,10 +603,6 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     for (int s = stripes - 1; s >= 0; s--) {
       for (int i = 0; i * 2 < ss; ++i) {
         const auto &w = writes[s * stripeSize / 2 + i];
-#if DEBUG_VERBOSE
-        printf("Write begin: %s\n", printable(w.begin).c_str());
-        fflush(stdout);
-#endif
         values[i * 2] = {w.begin.p, size_t(w.begin.len)};
         values[i * 2 + 1] = w.end.len > 0
                                 ? StringRef{w.end.p, size_t(w.end.len)}
@@ -651,16 +637,13 @@ private:
   SkipList skipList;
 };
 
-// Internal entry points. Public entry points should just delegate to these
-
-void internal_check(ConflictSet::Impl *impl,
-                    const ConflictSet::ReadRange *reads,
-                    ConflictSet::Result *results, int count) {
+void ConflictSet::check(const ReadRange *reads, Result *results,
+                        int count) const {
   impl->check(reads, results, count);
 }
-void internal_addWrites(ConflictSet::Impl *impl,
-                        const ConflictSet::WriteRange *writes, int count,
-                        int64_t writeVersion) {
+
+void ConflictSet::addWrites(const WriteRange *writes, int count,
+                            int64_t writeVersion) {
   mallocBytesDelta = 0;
   impl->addWrites(writes, count, writeVersion);
   impl->totalBytes += mallocBytesDelta;
@@ -671,7 +654,7 @@ void internal_addWrites(ConflictSet::Impl *impl,
 #endif
 }
 
-void internal_setOldestVersion(ConflictSet::Impl *impl, int64_t oldestVersion) {
+void ConflictSet::setOldestVersion(int64_t oldestVersion) {
   mallocBytesDelta = 0;
   impl->setOldestVersion(oldestVersion);
   impl->totalBytes += mallocBytesDelta;
@@ -681,43 +664,19 @@ void internal_setOldestVersion(ConflictSet::Impl *impl, int64_t oldestVersion) {
   }
 #endif
 }
-ConflictSet::Impl *internal_create(int64_t oldestVersion) {
-  mallocBytesDelta = 0;
-  auto *result = new (safe_malloc(sizeof(ConflictSet::Impl)))
-      ConflictSet::Impl{oldestVersion};
-  result->totalBytes += mallocBytesDelta;
-  return result;
-}
 
-void internal_destroy(ConflictSet::Impl *impl) {
-  impl->~Impl();
-  safe_free(impl, sizeof(ConflictSet::Impl));
-}
-
-int64_t internal_getBytes(ConflictSet::Impl *impl) { return impl->totalBytes; }
-
-void ConflictSet::check(const ReadRange *reads, Result *results,
-                        int count) const {
-  internal_check(impl, reads, results, count);
-}
-
-void ConflictSet::addWrites(const WriteRange *writes, int count,
-                            int64_t writeVersion) {
-  internal_addWrites(impl, writes, count, writeVersion);
-}
-
-void ConflictSet::setOldestVersion(int64_t oldestVersion) {
-  internal_setOldestVersion(impl, oldestVersion);
-}
-
-int64_t ConflictSet::getBytes() const { return internal_getBytes(impl); }
+int64_t ConflictSet::getBytes() const { return impl->totalBytes; }
 
 ConflictSet::ConflictSet(int64_t oldestVersion)
-    : impl(internal_create(oldestVersion)) {}
+    : impl((mallocBytesDelta = 0,
+            new(safe_malloc(sizeof(Impl))) Impl{oldestVersion})) {
+  impl->totalBytes += mallocBytesDelta;
+}
 
 ConflictSet::~ConflictSet() {
   if (impl) {
-    internal_destroy(impl);
+    impl->~Impl();
+    safe_free(impl, sizeof(Impl));
   }
 }
 
@@ -738,34 +697,37 @@ extern "C" {
 __attribute__((__visibility__("default"))) void
 ConflictSet_check(void *cs, const ConflictSet_ReadRange *reads,
                   ConflictSet_Result *results, int count) {
-  internal_check((ConflictSet::Impl *)cs, reads, results, count);
+  ((ConflictSet::Impl *)cs)->check(reads, results, count);
 }
 __attribute__((__visibility__("default"))) void
 ConflictSet_addWrites(void *cs, const ConflictSet_WriteRange *writes, int count,
                       int64_t writeVersion) {
-  internal_addWrites((ConflictSet::Impl *)cs, writes, count, writeVersion);
+  ((ConflictSet::Impl *)cs)->addWrites(writes, count, writeVersion);
 }
 __attribute__((__visibility__("default"))) void
 ConflictSet_setOldestVersion(void *cs, int64_t oldestVersion) {
-  internal_setOldestVersion((ConflictSet::Impl *)cs, oldestVersion);
+  ((ConflictSet::Impl *)cs)->setOldestVersion(oldestVersion);
 }
 __attribute__((__visibility__("default"))) void *
 ConflictSet_create(int64_t oldestVersion) {
-  return internal_create(oldestVersion);
+  return new (safe_malloc(sizeof(ConflictSet::Impl)))
+      ConflictSet::Impl{oldestVersion};
 }
 __attribute__((__visibility__("default"))) void ConflictSet_destroy(void *cs) {
-  internal_destroy((ConflictSet::Impl *)cs);
+  using Impl = ConflictSet::Impl;
+  ((Impl *)cs)->~Impl();
+  safe_free(cs, sizeof(Impl));
 }
 __attribute__((__visibility__("default"))) int64_t
 ConflictSet_getBytes(void *cs) {
-  return internal_getBytes((ConflictSet::Impl *)cs);
+  using Impl = ConflictSet::Impl;
+  return ((Impl *)cs)->totalBytes;
 }
 }
 
 #if SHOW_MEMORY
 struct __attribute__((visibility("default"))) PeakPrinter {
   ~PeakPrinter() {
-    printf("--- skip_list ---\n");
     printf("malloc bytes: %g\n", double(mallocBytes));
     printf("Peak malloc bytes: %g\n", double(peakMallocBytes));
   }

aarch64-symbol-imports.txt

@@ -1,8 +0,0 @@
-__stack_chk_fail@GLIBC_2.17
-__stack_chk_guard@GLIBC_2.17
-abort@GLIBC_2.17
-free@GLIBC_2.17
-malloc@GLIBC_2.17
-memcpy@GLIBC_2.17
-memmove@GLIBC_2.17
-memset@GLIBC_2.17

apple-symbol-imports.txt

@@ -4,5 +4,4 @@ _bzero
 _free
 _malloc
 _memcpy
-_memmove
-dyld_stub_binder
+_memmove

conflict_set.py

@@ -1,136 +0,0 @@
-import ctypes
-import enum
-import os
-
-from typing import Optional
-
-
-class _Key(ctypes.Structure):
-    _fields_ = [("p", ctypes.POINTER(ctypes.c_ubyte)), ("len", ctypes.c_int)]
-
-
-class ReadRange(ctypes.Structure):
-    _fields_ = [
-        ("begin", _Key),
-        ("end", _Key),
-        ("readVersion", ctypes.c_int64),
-    ]
-
-
-class WriteRange(ctypes.Structure):
-    _fields_ = [("begin", _Key), ("end", _Key)]
-
-
-class Result(enum.Enum):
-    COMMIT = 0
-    CONFLICT = 1
-    TOO_OLD = 2
-
-
-def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
-
-    if end is None:
-        e = (ctypes.c_ubyte * 0)()
-    else:
-        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
-    return WriteRange(_Key(b, len(b)), _Key(e, len(e)))
-
-
-def read(version: int, begin: bytes, end: Optional[bytes] = None) -> ReadRange:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
-    if end is None:
-        e = (ctypes.c_ubyte * 0)()
-    else:
-        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
-    return ReadRange(_Key(b, len(b)), _Key(e, len(e)), version)
-
-
-class ConflictSet:
-    def __init__(
-        self,
-        version: int = 0,
-        build_dir: Optional[str] = None,
-        implementation: Optional[str] = None,
-    ) -> None:
-        self._lib = None
-        if build_dir is None:
-            build_dir = os.path.dirname(__file__) + "/build"
-        if implementation is None:
-            implementation = "radix_tree"
-        for f in (
-            build_dir + "/" + implementation + "/libconflict-set.so.0",
-            os.path.dirname(__file__)
-            + "/build/"
-            + implementation
-            + "/libconflict-set.0.dylib",
-        ):
-            try:
-                self._lib = ctypes.cdll.LoadLibrary(f)
-            except:
-                pass
-
-        if self._lib is None:
-            import sys
-
-            print(
-                "Could not find libconflict-set implementation " + implementation,
-                file=sys.stderr,
-            )
-            sys.exit(1)
-
-        self._lib.ConflictSet_create.argtypes = (ctypes.c_int64,)
-        self._lib.ConflictSet_create.restype = ctypes.c_void_p
-
-        self._lib.ConflictSet_check.argtypes = (
-            ctypes.c_void_p,
-            ctypes.POINTER(ReadRange),
-            ctypes.POINTER(ctypes.c_int),
-            ctypes.c_int,
-        )
-
-        self._lib.ConflictSet_addWrites.argtypes = (
-            ctypes.c_void_p,
-            ctypes.POINTER(WriteRange),
-            ctypes.c_int,
-            ctypes.c_int64,
-        )
-
-        self._lib.ConflictSet_setOldestVersion.argtypes = (
-            ctypes.c_void_p,
-            ctypes.c_int64,
-        )
-
-        self._lib.ConflictSet_destroy.argtypes = (ctypes.c_void_p,)
-
-        self._lib.ConflictSet_getBytes.argtypes = (ctypes.c_void_p,)
-        self._lib.ConflictSet_getBytes.restype = ctypes.c_int64
-
-        self.p = self._lib.ConflictSet_create(version)
-
-    def addWrites(self, version: int, *writes: WriteRange):
-        self._lib.ConflictSet_addWrites(
-            self.p, (WriteRange * len(writes))(*writes), len(writes), version
-        )
-
-    def check(self, *reads: ReadRange) -> list[Result]:
-        r = (ctypes.c_int * len(reads))()
-        self._lib.ConflictSet_check(self.p, *reads, r, 1)
-        return [Result(x) for x in r]
-
-    def setOldestVersion(self, version: int) -> None:
-        self._lib.ConflictSet_setOldestVersion(self.p, version)
-
-    def getBytes(self) -> int:
-        return self._lib.ConflictSet_getBytes(self.p)
-
-    def __enter__(self):
-        return self
-
-    def close(self) -> None:
-        if self.p is not None:
-            self._lib.ConflictSet_destroy(self.p)
-            self.p = None
-
-    def __exit__(self, exception_type, exception_value, exception_traceback):
-        self.close()