StringDictionary Class Reference

#include <StringDictionary.h>

Collaboration diagram for StringDictionary:

Classes
struct	compare_cache_value_t
struct	PayloadString
struct	StringIdxEntry

Public Member Functions
	StringDictionary (const std::string &folder, const bool isTemp, const bool recover, const bool materializeHashes=false, size_t initial_capacity=256)
	StringDictionary (const LeafHostInfo &host, const DictRef dict_ref)
	~StringDictionary () noexcept
int32_t	getOrAdd (const std::string &str) noexcept
template<class T , class String >
void	getOrAddBulk (const std::vector< String > &string_vec, T *encoded_vec)
template<class T , class String >
void	getOrAddBulkParallel (const std::vector< String > &string_vec, T *encoded_vec)
template<class String >
void	getOrAddBulkArray (const std::vector< std::vector< String >> &string_array_vec, std::vector< std::vector< int32_t >> &ids_array_vec)
int32_t	getIdOfString (const std::string &str) const
std::string	getString (int32_t string_id) const
std::pair< char *, size_t >	getStringBytes (int32_t string_id) const noexcept
size_t	storageEntryCount () const
std::vector< int32_t >	getLike (const std::string &pattern, const bool icase, const bool is_simple, const char escape, const size_t generation) const
std::vector< int32_t >	getCompare (const std::string &pattern, const std::string &comp_operator, const size_t generation)
std::vector< int32_t >	getRegexpLike (const std::string &pattern, const char escape, const size_t generation) const
std::shared_ptr< const std::vector< std::string > >	copyStrings () const
bool	checkpoint () noexcept

Static Public Member Functions
static void	populate_string_ids (std::vector< int32_t > &dest_ids, StringDictionary *dest_dict, const std::vector< int32_t > &source_ids, const StringDictionary *source_dict, const std::map< int32_t, std::string > transient_mapping={})
	Populates provided dest_ids vector with string ids corresponding to given source strings. More...
static void	populate_string_array_ids (std::vector< std::vector< int32_t >> &dest_array_ids, StringDictionary *dest_dict, const std::vector< std::vector< int32_t >> &source_array_ids, const StringDictionary *source_dict)

Static Public Attributes
static constexpr int32_t	INVALID_STR_ID = -1
static constexpr size_t	MAX_STRLEN = (1 << 15) - 1
static constexpr size_t	MAX_STRCOUNT = (1U << 31) - 1

Private Member Functions
void	processDictionaryFutures (std::vector< std::future< std::vector< std::pair< uint32_t, unsigned int >>>> &dictionary_futures)
size_t	getNumStringsFromStorage (const size_t storage_slots) const noexcept
bool	fillRateIsHigh (const size_t num_strings) const noexcept
void	increaseCapacity () noexcept
template<class String >
void	increaseCapacityFromStorageAndMemory (const size_t storage_high_water_mark, const std::vector< String > &input_strings, const std::vector< size_t > &string_memory_ids, const std::vector< uint32_t > &input_strings_rk_hashes) noexcept
int32_t	getOrAddImpl (const std::string &str) noexcept
template<class String >
void	hashStrings (const std::vector< String > &string_vec, std::vector< uint32_t > &hashes) const noexcept
template<class T , class String >
void	getOrAddBulkRemote (const std::vector< String > &string_vec, T *encoded_vec)
int32_t	getUnlocked (const std::string &str) const noexcept
std::string	getStringUnlocked (int32_t string_id) const noexcept
std::string	getStringChecked (const int string_id) const noexcept
std::pair< char *, size_t >	getStringBytesChecked (const int string_id) const noexcept
template<class String >
uint32_t	computeBucket (const uint32_t hash, const String &str, const std::vector< int32_t > &data) const noexcept
template<class String >
uint32_t	computeBucketFromStorageAndMemory (const uint32_t input_string_rk_hash, const String &input_string, const std::vector< int32_t > &string_id_hash_table, const size_t storage_high_water_mark, const std::vector< String > &input_strings, const std::vector< size_t > &string_memory_ids) const noexcept
uint32_t	computeUniqueBucketWithHash (const uint32_t hash, const std::vector< int32_t > &data) noexcept
void	checkAndConditionallyIncreasePayloadCapacity (const size_t write_length)
void	checkAndConditionallyIncreaseOffsetCapacity (const size_t write_length)
template<class String >
void	appendToStorage (String str) noexcept
template<class String >
void	appendToStorageBulk (const std::vector< String > &input_strings, const std::vector< size_t > &string_memory_ids, const size_t sum_new_strings_lengths) noexcept
PayloadString	getStringFromStorage (const int string_id) const noexcept
std::string_view	getStringFromStorageFast (const int string_id) const noexcept
void	addPayloadCapacity (const size_t min_capacity_requested=0) noexcept
void	addOffsetCapacity (const size_t min_capacity_requested=0) noexcept
size_t	addStorageCapacity (int fd, const size_t min_capacity_requested=0) noexcept
void *	addMemoryCapacity (void *addr, size_t &mem_size, const size_t min_capacity_requested=0) noexcept
void	invalidateInvertedIndex () noexcept
std::vector< int32_t >	getEquals (std::string pattern, std::string comp_operator, size_t generation)
void	buildSortedCache ()
void	insertInSortedCache (std::string str, int32_t str_id)
void	sortCache (std::vector< int32_t > &cache)
void	mergeSortedCache (std::vector< int32_t > &temp_sorted_cache)
compare_cache_value_t *	binary_search_cache (const std::string &pattern) const

Private Attributes
size_t	str_count_
size_t	collisions_
std::vector< int32_t >	string_id_hash_table_
std::vector< uint32_t >	rk_hashes_
std::vector< int32_t >	sorted_cache
bool	isTemp_
bool	materialize_hashes_
std::string	offsets_path_
int	payload_fd_
int	offset_fd_
StringIdxEntry *	offset_map_
char *	payload_map_
size_t	offset_file_size_
size_t	payload_file_size_
size_t	payload_file_off_
mapd_shared_mutex	rw_mutex_
std::map< std::tuple< std::string, bool, bool, char >, std::vector< int32_t > >	like_cache_
std::map< std::pair< std::string, char >, std::vector< int32_t > >	regex_cache_
std::map< std::string, int32_t >	equal_cache_
DictionaryCache< std::string, compare_cache_value_t >	compare_cache_
std::shared_ptr< std::vector< std::string > >	strings_cache_
std::unique_ptr< StringDictionaryClient >	client_
std::unique_ptr< StringDictionaryClient >	client_no_timeout_
char *	CANARY_BUFFER {nullptr}
size_t	canary_buffer_size = 0

Detailed Description

Definition at line 42 of file StringDictionary.h.

Constructor & Destructor Documentation

StringDictionary() [1/2]

StringDictionary::StringDictionary	(	const std::string &	folder,
		const bool	isTemp,
		const bool	recover,
		const bool	materializeHashes = false,
		size_t	initial_capacity = 256
	)

Definition at line 111 of file StringDictionary.cpp.

References addOffsetCapacity(), addPayloadCapacity(), CHECK_EQ, anonymous_namespace{StringDictionary.cpp}::checked_mmap(), anonymous_namespace{StringDictionary.cpp}::checked_open(), collisions_, anonymous_namespace{StringDictionary.cpp}::file_size(), getNumStringsFromStorage(), getStringFromStorage(), INVALID_STR_ID, isTemp_, LOG, materialize_hashes_, offset_fd_, offset_file_size_, offset_map_, offsets_path_, payload_fd_, payload_file_size_, payload_map_, processDictionaryFutures(), anonymous_namespace{StringDictionary.cpp}::rk_hash(), rk_hashes_, anonymous_namespace{StringDictionary.cpp}::round_up_p2(), rw_mutex_, str_count_, string_id_hash_table_, VLOG, and logger::WARNING.

    : str_count_(0)
    , string_id_hash_table_(initial_capacity, INVALID_STR_ID)
    , rk_hashes_(initial_capacity)
    , isTemp_(isTemp)
    , materialize_hashes_(materializeHashes)
    , payload_fd_(-1)
    , offset_fd_(-1)
    , offset_map_(nullptr)
    , payload_map_(nullptr)
    , offset_file_size_(0)
    , payload_file_size_(0)
    , payload_file_off_(0)
    , strings_cache_(nullptr) {
  if (!isTemp && folder.empty()) {
    return;
  }

  // initial capacity must be a power of two for efficient bucket computation
  CHECK_EQ(size_t(0), (initial_capacity & (initial_capacity - 1)));
  if (!isTemp_) {
    boost::filesystem::path storage_path(folder);
    offsets_path_ = (storage_path / boost::filesystem::path("DictOffsets")).string();
    const auto payload_path =
        (storage_path / boost::filesystem::path("DictPayload")).string();
    payload_fd_ = checked_open(payload_path.c_str(), recover);
    offset_fd_ = checked_open(offsets_path_.c_str(), recover);
    payload_file_size_ = file_size(payload_fd_);
    offset_file_size_ = file_size(offset_fd_);
  }
  bool storage_is_empty = false;
  if (payload_file_size_ == 0) {
    storage_is_empty = true;
    addPayloadCapacity();
  }
  if (offset_file_size_ == 0) {
    addOffsetCapacity();
  }
  if (!isTemp_) {  // we never mmap or recover temp dictionaries
    payload_map_ = reinterpret_cast<char*>(checked_mmap(payload_fd_, payload_file_size_));
    offset_map_ =
        reinterpret_cast<StringIdxEntry*>(checked_mmap(offset_fd_, offset_file_size_));
    if (recover) {
      const size_t bytes = file_size(offset_fd_);
      if (bytes % sizeof(StringIdxEntry) != 0) {
        LOG(WARNING) << "Offsets " << offsets_path_ << " file is truncated";
      }
      const uint64_t str_count =
          storage_is_empty ? 0 : getNumStringsFromStorage(bytes / sizeof(StringIdxEntry));
      collisions_ = 0;
      // at this point we know the size of the StringDict we need to load
      // so lets reallocate the vector to the correct size
      const uint64_t max_entries =
          std::max(round_up_p2(str_count * 2 + 1),
                   round_up_p2(std::max(initial_capacity, static_cast<size_t>(1))));
      std::vector<int32_t> new_str_ids(max_entries, INVALID_STR_ID);
      string_id_hash_table_.swap(new_str_ids);
      if (materialize_hashes_) {
        std::vector<uint32_t> new_rk_hashes(max_entries / 2);
        rk_hashes_.swap(new_rk_hashes);
      }
      // Bail early if we know we don't have strings to add (i.e. a new or empty
      // dictionary)
      if (str_count == 0) {
        return;
      }

      unsigned string_id = 0;
      mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);

      uint32_t thread_inits = 0;
      const auto thread_count = std::thread::hardware_concurrency();
      const uint32_t items_per_thread = std::max<uint32_t>(
          2000, std::min<uint32_t>(200000, (str_count / thread_count) + 1));
      std::vector<std::future<std::vector<std::pair<uint32_t, unsigned int>>>>
          dictionary_futures;
      for (string_id = 0; string_id < str_count; string_id += items_per_thread) {
        dictionary_futures.emplace_back(std::async(
            std::launch::async, [string_id, str_count, items_per_thread, this] {
              std::vector<std::pair<uint32_t, unsigned int>> hashVec;
              for (uint32_t curr_id = string_id;
                   curr_id < string_id + items_per_thread && curr_id < str_count;
                   curr_id++) {
                const auto recovered = getStringFromStorage(curr_id);
                if (recovered.canary) {
                  // hit the canary, recovery finished
                  break;
                } else {
                  std::string temp(recovered.c_str_ptr, recovered.size);
                  hashVec.emplace_back(std::make_pair(rk_hash(temp), temp.size()));
                }
              }
              return hashVec;
            }));
        thread_inits++;
        if (thread_inits % thread_count == 0) {
          processDictionaryFutures(dictionary_futures);
        }
      }
      // gather last few threads
      if (dictionary_futures.size() != 0) {
        processDictionaryFutures(dictionary_futures);
      }
      VLOG(1) << "Opened string dictionary " << folder << " # Strings: " << str_count_
              << " Hash table size: " << string_id_hash_table_.size() << " Fill rate: "
              << static_cast<double>(str_count_) * 100.0 / string_id_hash_table_.size()
              << "% Collisions: " << collisions_;
    }
  }
}

Here is the call graph for this function:

StringDictionary() [2/2]

StringDictionary::StringDictionary	(	const LeafHostInfo &	host,
		const DictRef	dict_ref
	)

Definition at line 275 of file StringDictionary.cpp.

    : strings_cache_(nullptr)
    , client_(new StringDictionaryClient(host, dict_ref, true))
    , client_no_timeout_(new StringDictionaryClient(host, dict_ref, false)) {}

~StringDictionary()

StringDictionary::~StringDictionary ( )

noexcept

Definition at line 280 of file StringDictionary.cpp.

References CANARY_BUFFER, CHECK, CHECK_GE, anonymous_namespace{StringDictionary.cpp}::checked_munmap(), client_, File_Namespace::close(), isTemp_, offset_fd_, offset_file_size_, offset_map_, payload_fd_, payload_file_size_, and payload_map_.

                                             {
  free(CANARY_BUFFER);
  if (client_) {
    return;
  }
  if (payload_map_) {
    if (!isTemp_) {
      CHECK(offset_map_);
      checked_munmap(payload_map_, payload_file_size_);
      checked_munmap(offset_map_, offset_file_size_);
      CHECK_GE(payload_fd_, 0);
      close(payload_fd_);
      CHECK_GE(offset_fd_, 0);
      close(offset_fd_);
    } else {
      CHECK(offset_map_);
      free(payload_map_);
      free(offset_map_);
    }
  }
}

Here is the call graph for this function:

Member Function Documentation

addMemoryCapacity()

void * StringDictionary::addMemoryCapacity ( void *  addr, size_t &  mem_size, const size_t  min_capacity_requested = 0  )

privatenoexcept

Definition at line 1348 of file StringDictionary.cpp.

References CANARY_BUFFER, canary_buffer_size, CHECK, and anonymous_namespace{StringDictionary.cpp}::SYSTEM_PAGE_SIZE.

Referenced by addOffsetCapacity(), and addPayloadCapacity().

                                                                                        {
  const size_t canary_buff_size_to_add =
      std::max(static_cast<size_t>(1024 * SYSTEM_PAGE_SIZE),
               (min_capacity_requested / SYSTEM_PAGE_SIZE + 1) * SYSTEM_PAGE_SIZE);
  if (canary_buffer_size != canary_buff_size_to_add) {
    CANARY_BUFFER =
        reinterpret_cast<char*>(realloc(CANARY_BUFFER, canary_buff_size_to_add));
    canary_buffer_size = canary_buff_size_to_add;
  }
  CHECK(CANARY_BUFFER);
  memset(CANARY_BUFFER, 0xff, canary_buff_size_to_add);
  void* new_addr = realloc(addr, mem_size + canary_buff_size_to_add);
  CHECK(new_addr);
  void* write_addr = reinterpret_cast<void*>(static_cast<char*>(new_addr) + mem_size);
  CHECK(memcpy(write_addr, CANARY_BUFFER, canary_buff_size_to_add));
  mem_size += canary_buff_size_to_add;
  return new_addr;
}

Here is the caller graph for this function:

addOffsetCapacity()

void StringDictionary::addOffsetCapacity ( const size_t  min_capacity_requested = 0 )

privatenoexcept

Definition at line 1318 of file StringDictionary.cpp.

References addMemoryCapacity(), addStorageCapacity(), isTemp_, offset_fd_, offset_file_size_, and offset_map_.

Referenced by checkAndConditionallyIncreaseOffsetCapacity(), and StringDictionary().

                                                                                     {
  if (!isTemp_) {
    offset_file_size_ += addStorageCapacity(offset_fd_, min_capacity_requested);
  } else {
    offset_map_ = static_cast<StringIdxEntry*>(
        addMemoryCapacity(offset_map_, offset_file_size_, min_capacity_requested));
  }
}

Here is the call graph for this function:

Here is the caller graph for this function:

addPayloadCapacity()

void StringDictionary::addPayloadCapacity ( const size_t  min_capacity_requested = 0 )

privatenoexcept

Definition at line 1309 of file StringDictionary.cpp.

References addMemoryCapacity(), addStorageCapacity(), isTemp_, payload_fd_, payload_file_size_, and payload_map_.

Referenced by checkAndConditionallyIncreasePayloadCapacity(), and StringDictionary().

                                                                                      {
  if (!isTemp_) {
    payload_file_size_ += addStorageCapacity(payload_fd_, min_capacity_requested);
  } else {
    payload_map_ = static_cast<char*>(
        addMemoryCapacity(payload_map_, payload_file_size_, min_capacity_requested));
  }
}

Here is the call graph for this function:

Here is the caller graph for this function:

addStorageCapacity()

size_t StringDictionary::addStorageCapacity ( int  fd, const size_t  min_capacity_requested = 0  )

privatenoexcept

Definition at line 1327 of file StringDictionary.cpp.

References CANARY_BUFFER, canary_buffer_size, CHECK, CHECK_NE, anonymous_namespace{StringDictionary.cpp}::SYSTEM_PAGE_SIZE, and File_Namespace::write().

Referenced by addOffsetCapacity(), and addPayloadCapacity().

                                                  {
  const size_t canary_buff_size_to_add =
      std::max(static_cast<size_t>(1024 * SYSTEM_PAGE_SIZE),
               (min_capacity_requested / SYSTEM_PAGE_SIZE + 1) * SYSTEM_PAGE_SIZE);

  if (canary_buffer_size != canary_buff_size_to_add) {
    CANARY_BUFFER = static_cast<char*>(realloc(CANARY_BUFFER, canary_buff_size_to_add));
    canary_buffer_size = canary_buff_size_to_add;
  }
  CHECK(CANARY_BUFFER);
  memset(CANARY_BUFFER, 0xff, canary_buff_size_to_add);

  CHECK_NE(lseek(fd, 0, SEEK_END), -1);
  ssize_t write_return = write(fd, CANARY_BUFFER, canary_buff_size_to_add);
  CHECK(write_return > 0 &&
        (static_cast<size_t>(write_return) == canary_buff_size_to_add));
  return canary_buff_size_to_add;
}

Here is the call graph for this function:

Here is the caller graph for this function:

appendToStorage()

template<class String >

void StringDictionary::appendToStorage ( String  str )

privatenoexcept

Definition at line 1254 of file StringDictionary.cpp.

References checkAndConditionallyIncreaseOffsetCapacity(), checkAndConditionallyIncreasePayloadCapacity(), offset_map_, payload_file_off_, payload_map_, StringDictionary::StringIdxEntry::size, and str_count_.

Referenced by getOrAddBulk(), and getOrAddImpl().

                                                          {
  // write the payload
  checkAndConditionallyIncreasePayloadCapacity(str.size());
  memcpy(payload_map_ + payload_file_off_, str.data(), str.size());

  // write the offset and length
  StringIdxEntry str_meta{static_cast<uint64_t>(payload_file_off_), str.size()};
  payload_file_off_ += str.size();  // Need to increment after we've defined str_meta

  checkAndConditionallyIncreaseOffsetCapacity(sizeof(str_meta));
  memcpy(offset_map_ + str_count_, &str_meta, sizeof(str_meta));
}

Here is the call graph for this function:

Here is the caller graph for this function:

appendToStorageBulk()

template<class String >

void StringDictionary::appendToStorageBulk ( const std::vector< String > &  input_strings, const std::vector< size_t > &  string_memory_ids, const size_t  sum_new_strings_lengths  )

privatenoexcept

Definition at line 1268 of file StringDictionary.cpp.

References checkAndConditionallyIncreaseOffsetCapacity(), checkAndConditionallyIncreasePayloadCapacity(), offset_map_, payload_file_off_, payload_map_, and str_count_.

Referenced by getOrAddBulkParallel().

                                                   {
  const size_t num_strings = string_memory_ids.size();

  checkAndConditionallyIncreasePayloadCapacity(sum_new_strings_lengths);
  checkAndConditionallyIncreaseOffsetCapacity(sizeof(StringIdxEntry) * num_strings);

  for (size_t i = 0; i < num_strings; ++i) {
    const size_t string_idx = string_memory_ids[i];
    const String str = input_strings[string_idx];
    const size_t str_size(str.size());
    memcpy(payload_map_ + payload_file_off_, str.data(), str_size);
    StringIdxEntry str_meta{static_cast<uint64_t>(payload_file_off_), str_size};
    payload_file_off_ += str_size;  // Need to increment after we've defined str_meta
    memcpy(offset_map_ + str_count_ + i, &str_meta, sizeof(str_meta));
  }
}

Here is the call graph for this function:

Here is the caller graph for this function:

binary_search_cache()

compare_cache_value_t* StringDictionary::binary_search_cache ( const std::string &  pattern ) const

private

buildSortedCache()

void StringDictionary::buildSortedCache ( )

private

Definition at line 1399 of file StringDictionary.cpp.

References mergeSortedCache(), sortCache(), sorted_cache, and str_count_.

Referenced by getCompare().

                                        {
  // This method is not thread-safe.
  const auto cur_cache_size = sorted_cache.size();
  std::vector<int32_t> temp_sorted_cache;
  for (size_t i = cur_cache_size; i < str_count_; i++) {
    temp_sorted_cache.push_back(i);
  }
  sortCache(temp_sorted_cache);
  mergeSortedCache(temp_sorted_cache);
}

Here is the call graph for this function:

Here is the caller graph for this function:

checkAndConditionallyIncreaseOffsetCapacity()

void StringDictionary::checkAndConditionallyIncreaseOffsetCapacity ( const size_t  write_length )

private

Definition at line 1233 of file StringDictionary.cpp.

References addOffsetCapacity(), CHECK, CHECK_GE, anonymous_namespace{StringDictionary.cpp}::checked_mmap(), anonymous_namespace{StringDictionary.cpp}::checked_munmap(), isTemp_, offset_fd_, offset_file_size_, offset_map_, and str_count_.

Referenced by appendToStorage(), and appendToStorageBulk().

                               {
  size_t offset_file_off = str_count_ * sizeof(StringIdxEntry);
  if (offset_file_off + write_length >= offset_file_size_) {
    const size_t min_capacity_needed =
        write_length - (offset_file_size_ - offset_file_off);
    if (!isTemp_) {
      CHECK_GE(offset_fd_, 0);
      checked_munmap(offset_map_, offset_file_size_);
      addOffsetCapacity(min_capacity_needed);
      CHECK(offset_file_off + write_length <= offset_file_size_);
      offset_map_ =
          reinterpret_cast<StringIdxEntry*>(checked_mmap(offset_fd_, offset_file_size_));
    } else {
      addOffsetCapacity(min_capacity_needed);
      CHECK(offset_file_off + write_length <= offset_file_size_);
    }
  }
}

Here is the call graph for this function:

Here is the caller graph for this function:

checkAndConditionallyIncreasePayloadCapacity()

void StringDictionary::checkAndConditionallyIncreasePayloadCapacity ( const size_t  write_length )

private

Definition at line 1214 of file StringDictionary.cpp.

References addPayloadCapacity(), CHECK, CHECK_GE, anonymous_namespace{StringDictionary.cpp}::checked_mmap(), anonymous_namespace{StringDictionary.cpp}::checked_munmap(), isTemp_, payload_fd_, payload_file_off_, payload_file_size_, and payload_map_.

Referenced by appendToStorage(), and appendToStorageBulk().

                               {
  if (payload_file_off_ + write_length > payload_file_size_) {
    const size_t min_capacity_needed =
        write_length - (payload_file_size_ - payload_file_off_);
    if (!isTemp_) {
      CHECK_GE(payload_fd_, 0);
      checked_munmap(payload_map_, payload_file_size_);
      addPayloadCapacity(min_capacity_needed);
      CHECK(payload_file_off_ + write_length <= payload_file_size_);
      payload_map_ =
          reinterpret_cast<char*>(checked_mmap(payload_fd_, payload_file_size_));
    } else {
      addPayloadCapacity(min_capacity_needed);
      CHECK(payload_file_off_ + write_length <= payload_file_size_);
    }
  }
}

Here is the call graph for this function:

Here is the caller graph for this function:

checkpoint()

bool StringDictionary::checkpoint ( )

noexcept

Definition at line 1382 of file StringDictionary.cpp.

References CHECK, client_, isTemp_, offset_fd_, offset_file_size_, offset_map_, payload_fd_, payload_file_size_, and payload_map_.

                                           {
  if (client_) {
    try {
      return client_->checkpoint();
    } catch (...) {
      return false;
    }
  }
  CHECK(!isTemp_);
  bool ret = true;
  ret = ret && (msync((void*)offset_map_, offset_file_size_, MS_SYNC) == 0);
  ret = ret && (msync((void*)payload_map_, payload_file_size_, MS_SYNC) == 0);
  ret = ret && (fsync(offset_fd_) == 0);
  ret = ret && (fsync(payload_fd_) == 0);
  return ret;
}

computeBucket()

template<class String >

uint32_t StringDictionary::computeBucket ( const uint32_t  hash, const String &  str, const std::vector< int32_t > &  data  ) const

privatenoexcept

Definition at line 1117 of file StringDictionary.cpp.

References getStringFromStorageFast(), INVALID_STR_ID, materialize_hashes_, and rk_hashes_.

Referenced by getOrAddBulk(), getOrAddImpl(), and getUnlocked().

             {
  auto bucket = hash & (data.size() - 1);
  while (true) {
    const int32_t candidate_string_id = data[bucket];
    if (candidate_string_id ==
        INVALID_STR_ID) {  // In this case it means the slot is available for use
      break;
    }
    if (!materialize_hashes_ ||
        (materialize_hashes_ && hash == rk_hashes_[candidate_string_id])) {
      const auto old_str = getStringFromStorageFast(candidate_string_id);
      if (str.size() == old_str.size() &&
          !memcmp(str.data(), old_str.data(), str.size())) {
        // found the string
        break;
      }
    }
    // wrap around
    if (++bucket == data.size()) {
      bucket = 0;
    }
  }
  return bucket;
}

Here is the call graph for this function:

Here is the caller graph for this function:

computeBucketFromStorageAndMemory()

template<class String >

uint32_t StringDictionary::computeBucketFromStorageAndMemory ( const uint32_t  input_string_rk_hash, const String &  input_string, const std::vector< int32_t > &  string_id_hash_table, const size_t  storage_high_water_mark, const std::vector< String > &  input_strings, const std::vector< size_t > &  string_memory_ids  ) const

privatenoexcept

memcmp(input_string.data(), candidate_storage_string.c_str_ptr, input_string.size())) {

Definition at line 1146 of file StringDictionary.cpp.

References getStringFromStorageFast(), INVALID_STR_ID, materialize_hashes_, and rk_hashes_.

Referenced by getOrAddBulkParallel().

                                                               {
  auto bucket = input_string_rk_hash & (string_id_hash_table.size() - 1);
  while (true) {
    const int32_t candidate_string_id = string_id_hash_table[bucket];
    if (candidate_string_id ==
        INVALID_STR_ID) {  // In this case it means the slot is available for use
      break;
    }
    if (!materialize_hashes_ ||
        (input_string_rk_hash == rk_hashes_[candidate_string_id])) {
      if (candidate_string_id > 0 &&
          static_cast<size_t>(candidate_string_id) >= storage_high_water_mark) {
        // The candidate string is not in storage yet but in our string_memory_ids temp
        // buffer
        size_t memory_offset =
            static_cast<size_t>(candidate_string_id - storage_high_water_mark);
        const String candidate_string = input_strings[string_memory_ids[memory_offset]];
        if (input_string.size() == candidate_string.size() &&
            !memcmp(input_string.data(), candidate_string.data(), input_string.size())) {
          // found the string in the temp memory buffer
          break;
        }
      } else {
        // The candidate string is in storage, need to fetch it for comparison
        const auto candidate_storage_string =
            getStringFromStorageFast(candidate_string_id);
        if (input_string.size() == candidate_storage_string.size() &&
            !memcmp(input_string.data(),
                    candidate_storage_string.data(),
                    input_string.size())) {
          // found the string in storage
          break;
        }
      }
    }
    if (++bucket == string_id_hash_table.size()) {
      bucket = 0;
    }
  }
  return bucket;
}

Here is the call graph for this function:

Here is the caller graph for this function:

computeUniqueBucketWithHash()

uint32_t StringDictionary::computeUniqueBucketWithHash ( const uint32_t  hash, const std::vector< int32_t > &  data  )

privatenoexcept

Definition at line 1196 of file StringDictionary.cpp.

References collisions_, and INVALID_STR_ID.

Referenced by increaseCapacity(), increaseCapacityFromStorageAndMemory(), and processDictionaryFutures().

                                             {
  auto bucket = hash & (data.size() - 1);
  while (true) {
    if (data[bucket] ==
        INVALID_STR_ID) {  // In this case it means the slot is available for use
      break;
    }
    collisions_++;
    // wrap around
    if (++bucket == data.size()) {
      bucket = 0;
    }
  }
  return bucket;
}

Here is the caller graph for this function:

copyStrings()

std::shared_ptr< const std::vector< std::string > > StringDictionary::copyStrings

(

)

const

Definition at line 950 of file StringDictionary.cpp.

References CHECK_EQ, CHECK_GT, CHECK_LE, client_, cpu_threads(), getStringUnlocked(), rw_mutex_, str_count_, and strings_cache_.

                                                                              {
  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);
  if (client_) {
    // TODO(miyu): support remote string dictionary
    throw std::runtime_error(
        "copying dictionaries from remote server is not supported yet.");
  }

  if (strings_cache_) {
    return strings_cache_;
  }

  strings_cache_ = std::make_shared<std::vector<std::string>>();
  strings_cache_->reserve(str_count_);
  const bool multithreaded = str_count_ > 10000;
  const auto worker_count =
      multithreaded ? static_cast<size_t>(cpu_threads()) : size_t(1);
  CHECK_GT(worker_count, 0UL);
  std::vector<std::vector<std::string>> worker_results(worker_count);
  auto copy = [this](std::vector<std::string>& str_list,
                     const size_t start_id,
                     const size_t end_id) {
    CHECK_LE(start_id, end_id);
    str_list.reserve(end_id - start_id);
    for (size_t string_id = start_id; string_id < end_id; ++string_id) {
      str_list.push_back(getStringUnlocked(string_id));
    }
  };
  if (multithreaded) {
    std::vector<std::future<void>> workers;
    const auto stride = (str_count_ + (worker_count - 1)) / worker_count;
    for (size_t worker_idx = 0, start = 0, end = std::min(start + stride, str_count_);
         worker_idx < worker_count && start < str_count_;
         ++worker_idx, start += stride, end = std::min(start + stride, str_count_)) {
      workers.push_back(std::async(
          std::launch::async, copy, std::ref(worker_results[worker_idx]), start, end));
    }
    for (auto& worker : workers) {
      worker.get();
    }
  } else {
    CHECK_EQ(worker_results.size(), size_t(1));
    copy(worker_results[0], 0, str_count_);
  }

  for (const auto& worker_result : worker_results) {
    strings_cache_->insert(
        strings_cache_->end(), worker_result.begin(), worker_result.end());
  }
  return strings_cache_;
}

Here is the call graph for this function:

fillRateIsHigh()

bool StringDictionary::fillRateIsHigh ( const size_t  num_strings ) const

privatenoexcept

Definition at line 1002 of file StringDictionary.cpp.

References string_id_hash_table_.

Referenced by getOrAddBulk(), getOrAddBulkParallel(), and getOrAddImpl().

                                                                             {
  return string_id_hash_table_.size() <= num_strings * 2;
}

Here is the caller graph for this function:

getCompare()

std::vector< int32_t > StringDictionary::getCompare	(	const std::string &	pattern,
		const std::string &	comp_operator,
		const size_t	generation
	)

Definition at line 742 of file StringDictionary.cpp.

References buildSortedCache(), client_, compare_cache_, getEquals(), getStringFromStorage(), rw_mutex_, sorted_cache, str_count_, string_eq(), and string_lt().

                                                                           {
  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);
  if (client_) {
    return client_->get_compare(pattern, comp_operator, generation);
  }
  std::vector<int32_t> ret;
  if (str_count_ == 0) {
    return ret;
  }
  if (sorted_cache.size() < str_count_) {
    if (comp_operator == "=" || comp_operator == "<>") {
      return getEquals(pattern, comp_operator, generation);
    }

    buildSortedCache();
  }
  auto cache_index = compare_cache_.get(pattern);

  if (!cache_index) {
    cache_index = std::make_shared<StringDictionary::compare_cache_value_t>();
    const auto cache_itr = std::lower_bound(
        sorted_cache.begin(),
        sorted_cache.end(),
        pattern,
        [this](decltype(sorted_cache)::value_type const& a, decltype(pattern)& b) {
          auto a_str = this->getStringFromStorage(a);
          return string_lt(a_str.c_str_ptr, a_str.size, b.c_str(), b.size());
        });

    if (cache_itr == sorted_cache.end()) {
      cache_index->index = sorted_cache.size() - 1;
      cache_index->diff = 1;
    } else {
      const auto cache_str = getStringFromStorage(*cache_itr);
      if (!string_eq(
              cache_str.c_str_ptr, cache_str.size, pattern.c_str(), pattern.size())) {
        cache_index->index = cache_itr - sorted_cache.begin() - 1;
        cache_index->diff = 1;
      } else {
        cache_index->index = cache_itr - sorted_cache.begin();
        cache_index->diff = 0;
      }
    }

    compare_cache_.put(pattern, cache_index);
  }

  // since we have a cache in form of vector of ints which is sorted according to
  // corresponding strings in the dictionary all we need is the index of the element
  // which equal to the pattern that we are trying to match or the index of “biggest”
  // element smaller than the pattern, to perform all the comparison operators over
  // string. The search function guarantees we have such index so now it is just the
  // matter to include all the elements in the result vector.

  // For < operator if the index that we have points to the element which is equal to
  // the pattern that we are searching for we simply get all the elements less than the
  // index. If the element pointed by the index is not equal to the pattern we are
  // comparing with we also need to include that index in result vector, except when the
  // index points to 0 and the pattern is lesser than the smallest value in the string
  // dictionary.

  if (comp_operator == "<") {
    size_t idx = cache_index->index;
    if (cache_index->diff) {
      idx = cache_index->index + 1;
      if (cache_index->index == 0 && cache_index->diff > 0) {
        idx = cache_index->index;
      }
    }
    for (size_t i = 0; i < idx; i++) {
      ret.push_back(sorted_cache[i]);
    }

    // For <= operator if the index that we have points to the element which is equal to
    // the pattern that we are searching for we want to include the element pointed by
    // the index in the result set. If the element pointed by the index is not equal to
    // the pattern we are comparing with we just want to include all the ids with index
    // less than the index that is cached, except when pattern that we are searching for
    // is smaller than the smallest string in the dictionary.

  } else if (comp_operator == "<=") {
    size_t idx = cache_index->index + 1;
    if (cache_index == 0 && cache_index->diff > 0) {
      idx = cache_index->index;
    }
    for (size_t i = 0; i < idx; i++) {
      ret.push_back(sorted_cache[i]);
    }

    // For > operator we want to get all the elements with index greater than the index
    // that we have except, when the pattern we are searching for is lesser than the
    // smallest string in the dictionary we also want to include the id of the index
    // that we have.

  } else if (comp_operator == ">") {
    size_t idx = cache_index->index + 1;
    if (cache_index->index == 0 && cache_index->diff > 0) {
      idx = cache_index->index;
    }
    for (size_t i = idx; i < sorted_cache.size(); i++) {
      ret.push_back(sorted_cache[i]);
    }

    // For >= operator when the indexed element that we have points to element which is
    // equal to the pattern we are searching for we want to include that in the result
    // vector. If the index that we have does not point to the string which is equal to
    // the pattern we are searching we don’t want to include that id into the result
    // vector except when the index is 0.

  } else if (comp_operator == ">=") {
    size_t idx = cache_index->index;
    if (cache_index->diff) {
      idx = cache_index->index + 1;
      if (cache_index->index == 0 && cache_index->diff > 0) {
        idx = cache_index->index;
      }
    }
    for (size_t i = idx; i < sorted_cache.size(); i++) {
      ret.push_back(sorted_cache[i]);
    }
  } else if (comp_operator == "=") {
    if (!cache_index->diff) {
      ret.push_back(sorted_cache[cache_index->index]);
    }

    // For <> operator it is simple matter of not including id of string which is equal
    // to pattern we are searching for.
  } else if (comp_operator == "<>") {
    if (!cache_index->diff) {
      size_t idx = cache_index->index;
      for (size_t i = 0; i < idx; i++) {
        ret.push_back(sorted_cache[i]);
      }
      ++idx;
      for (size_t i = idx; i < sorted_cache.size(); i++) {
        ret.push_back(sorted_cache[i]);
      }
    } else {
      for (size_t i = 0; i < sorted_cache.size(); i++) {
        ret.insert(ret.begin(), sorted_cache.begin(), sorted_cache.end());
      }
    }

  } else {
    std::runtime_error("Unsupported string comparison operator");
  }
  return ret;
}

Here is the call graph for this function:

getEquals()

std::vector< int32_t > StringDictionary::getEquals ( std::string  pattern, std::string  comp_operator, size_t  generation  )

private

Definition at line 682 of file StringDictionary.cpp.

References CHECK, CHECK_GT, CHECK_LE, cpu_threads(), equal_cache_, getStringUnlocked(), MAX_STRLEN, run_benchmark_import::result, and str_count_.

Referenced by getCompare().

                                                                    {
  std::vector<int32_t> result;
  auto eq_id_itr = equal_cache_.find(pattern);
  int32_t eq_id = MAX_STRLEN + 1;
  int32_t cur_size = str_count_;
  if (eq_id_itr != equal_cache_.end()) {
    auto eq_id = eq_id_itr->second;
    if (comp_operator == "=") {
      result.push_back(eq_id);
    } else {
      for (int32_t idx = 0; idx <= cur_size; idx++) {
        if (idx == eq_id) {
          continue;
        }
        result.push_back(idx);
      }
    }
  } else {
    std::vector<std::thread> workers;
    int worker_count = cpu_threads();
    CHECK_GT(worker_count, 0);
    std::vector<std::vector<int32_t>> worker_results(worker_count);
    CHECK_LE(generation, str_count_);
    for (int worker_idx = 0; worker_idx < worker_count; ++worker_idx) {
      workers.emplace_back(
          [&worker_results, &pattern, generation, worker_idx, worker_count, this]() {
            for (size_t string_id = worker_idx; string_id < generation;
                 string_id += worker_count) {
              const auto str = getStringUnlocked(string_id);
              if (str == pattern) {
                worker_results[worker_idx].push_back(string_id);
              }
            }
          });
    }
    for (auto& worker : workers) {
      worker.join();
    }
    for (const auto& worker_result : worker_results) {
      result.insert(result.end(), worker_result.begin(), worker_result.end());
    }
    if (result.size() > 0) {
      const auto it_ok = equal_cache_.insert(std::make_pair(pattern, result[0]));
      CHECK(it_ok.second);
      eq_id = result[0];
    }
    if (comp_operator == "<>") {
      for (int32_t idx = 0; idx <= cur_size; idx++) {
        if (idx == eq_id) {
          continue;
        }
        result.push_back(idx);
      }
    }
  }
  return result;
}

Here is the call graph for this function:

Here is the caller graph for this function:

getIdOfString()

int32_t StringDictionary::getIdOfString

(

const std::string &

str

)

const

Definition at line 557 of file StringDictionary.cpp.

References client_, getUnlocked(), and rw_mutex_.

                                                                  {
  mapd_shared_lock<mapd_shared_mutex> read_lock(rw_mutex_);
  if (client_) {
    return client_->get(str);
  }
  return getUnlocked(str);
}

Here is the call graph for this function:

getLike()

std::vector< int32_t > StringDictionary::getLike	(	const std::string &	pattern,
		const bool	icase,
		const bool	is_simple,
		const char	escape,
		const size_t	generation
	)		const

Definition at line 629 of file StringDictionary.cpp.

References CHECK, CHECK_GT, CHECK_LE, client_, cpu_threads(), getStringUnlocked(), anonymous_namespace{StringDictionary.cpp}::is_like(), like_cache_, run_benchmark_import::result, rw_mutex_, and str_count_.

                                                                              {
  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);
  if (client_) {
    return client_->get_like(pattern, icase, is_simple, escape, generation);
  }
  const auto cache_key = std::make_tuple(pattern, icase, is_simple, escape);
  const auto it = like_cache_.find(cache_key);
  if (it != like_cache_.end()) {
    return it->second;
  }
  std::vector<int32_t> result;
  std::vector<std::thread> workers;
  int worker_count = cpu_threads();
  CHECK_GT(worker_count, 0);
  std::vector<std::vector<int32_t>> worker_results(worker_count);
  CHECK_LE(generation, str_count_);
  for (int worker_idx = 0; worker_idx < worker_count; ++worker_idx) {
    workers.emplace_back([&worker_results,
                          &pattern,
                          generation,
                          icase,
                          is_simple,
                          escape,
                          worker_idx,
                          worker_count,
                          this]() {
      for (size_t string_id = worker_idx; string_id < generation;
           string_id += worker_count) {
        const auto str = getStringUnlocked(string_id);
        if (is_like(str, pattern, icase, is_simple, escape)) {
          worker_results[worker_idx].push_back(string_id);
        }
      }
    });
  }
  for (auto& worker : workers) {
    worker.join();
  }
  for (const auto& worker_result : worker_results) {
    result.insert(result.end(), worker_result.begin(), worker_result.end());
  }
  // place result into cache for reuse if similar query
  const auto it_ok = like_cache_.insert(std::make_pair(cache_key, result));

  CHECK(it_ok.second);

  return result;
}

Here is the call graph for this function:

getNumStringsFromStorage()

size_t StringDictionary::getNumStringsFromStorage ( const size_t  storage_slots ) const

privatenoexcept

Method to retrieve number of strings in storage via a binary search for the first canary

Parameters

storage_slots

number of storage entries we should search to find the minimum canary

Returns

number of strings in storage

Definition at line 252 of file StringDictionary.cpp.

References CHECK_GE, and getStringFromStorage().

Referenced by StringDictionary().

             {
  if (storage_slots == 0) {
    return 0;
  }
  // Must use signed integers since final binary search step can wrap to max size_t value
  // if dictionary is empty
  int64_t min_bound = 0;
  int64_t max_bound = storage_slots - 1;
  int64_t guess{0};
  while (min_bound <= max_bound) {
    guess = (max_bound + min_bound) / 2;
    CHECK_GE(guess, 0);
    if (getStringFromStorage(guess).canary) {
      max_bound = guess - 1;
    } else {
      min_bound = guess + 1;
    }
  }
  CHECK_GE(guess + (min_bound > guess ? 1 : 0), 0);
  return guess + (min_bound > guess ? 1 : 0);
}

Here is the call graph for this function:

Here is the caller graph for this function:

getOrAdd()

int32_t StringDictionary::getOrAdd ( const std::string &  str )

noexcept

Definition at line 302 of file StringDictionary.cpp.

References CHECK_EQ, client_, and getOrAddImpl().

                                                                {
  if (client_) {
    std::vector<int32_t> string_ids;
    client_->get_or_add_bulk(string_ids, std::vector<std::string>{str});
    CHECK_EQ(size_t(1), string_ids.size());
    return string_ids.front();
  }
  return getOrAddImpl(str);
}

Here is the call graph for this function:

getOrAddBulk()

template<class T , class String >

template void StringDictionary::getOrAddBulk	(	const std::vector< String > &	string_vec,
		T *	encoded_vec
	)

Definition at line 362 of file StringDictionary.cpp.

References appendToStorage(), CHECK, CHECK_LT, client_no_timeout_, computeBucket(), fillRateIsHigh(), g_enable_stringdict_parallel, getOrAddBulkParallel(), getOrAddBulkRemote(), increaseCapacity(), INVALID_STR_ID, invalidateInvertedIndex(), materialize_hashes_, MAX_STRCOUNT, MAX_STRLEN, offsets_path_, anonymous_namespace{StringDictionary.cpp}::rk_hash(), rk_hashes_, rw_mutex_, str_count_, and string_id_hash_table_.

Referenced by foreign_storage::ParquetStringEncoder< V >::appendData(), ArrowCsvForeignStorage::createDictionaryEncodedColumn(), getOrAddBulkArray(), getOrAddBulkParallel(), and populate_string_ids().

                                                          {
  if (g_enable_stringdict_parallel) {
    getOrAddBulkParallel(input_strings, output_string_ids);
    return;
  }
  // Single-thread path.
  if (client_no_timeout_) {
    getOrAddBulkRemote(input_strings, output_string_ids);
    return;
  }
  size_t out_idx{0};
  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);

  for (const auto& str : input_strings) {
    if (str.empty()) {
      output_string_ids[out_idx++] = inline_int_null_value<T>();
      continue;
    }
    CHECK(str.size() <= MAX_STRLEN);
    uint32_t bucket;
    const uint32_t hash = rk_hash(str);
    bucket = computeBucket(hash, str, string_id_hash_table_);
    if (string_id_hash_table_[bucket] != INVALID_STR_ID) {
      output_string_ids[out_idx++] = string_id_hash_table_[bucket];
      continue;
    }
    // need to add record to dictionary
    // check there is room
    if (str_count_ == static_cast<size_t>(max_valid_int_value<T>())) {
      log_encoding_error<T>(str);
      output_string_ids[out_idx++] = inline_int_null_value<T>();
      continue;
    }
    if (string_id_hash_table_[bucket] == INVALID_STR_ID) {
      CHECK_LT(str_count_, MAX_STRCOUNT)
          << "Maximum number (" << str_count_
          << ") of Dictionary encoded Strings reached for this column, offset path "
             "for column is  "
          << offsets_path_;
      if (fillRateIsHigh(str_count_)) {
        // resize when more than 50% is full
        increaseCapacity();
        bucket = computeBucket(hash, str, string_id_hash_table_);
      }
      appendToStorage(str);

      string_id_hash_table_[bucket] = static_cast<int32_t>(str_count_);
      if (materialize_hashes_) {
        rk_hashes_[str_count_] = hash;
      }
      ++str_count_;
    }
    output_string_ids[out_idx++] = string_id_hash_table_[bucket];
  }
  invalidateInvertedIndex();
}

Here is the call graph for this function:

Here is the caller graph for this function:

getOrAddBulkArray()

template<class String >

template void StringDictionary::getOrAddBulkArray	(	const std::vector< std::vector< String >> &	string_array_vec,
		std::vector< std::vector< int32_t >> &	ids_array_vec
	)

Definition at line 324 of file StringDictionary.cpp.

References getOrAddBulk().

                                                  {
  ids_array_vec.resize(string_array_vec.size());
  for (size_t i = 0; i < string_array_vec.size(); i++) {
    auto& strings = string_array_vec[i];
    auto& ids = ids_array_vec[i];
    ids.resize(strings.size());
    getOrAddBulk(strings, &ids[0]);
  }
}

Here is the call graph for this function:

getOrAddBulkParallel()

template<class T , class String >

void StringDictionary::getOrAddBulkParallel	(	const std::vector< String > &	string_vec,
		T *	encoded_vec
	)

Definition at line 421 of file StringDictionary.cpp.

References appendToStorageBulk(), CHECK, CHECK_LT, client_no_timeout_, computeBucketFromStorageAndMemory(), fillRateIsHigh(), getOrAddBulk(), getOrAddBulkRemote(), hashStrings(), increaseCapacityFromStorageAndMemory(), INVALID_STR_ID, invalidateInvertedIndex(), materialize_hashes_, MAX_STRCOUNT, MAX_STRLEN, offsets_path_, rk_hashes_, rw_mutex_, str_count_, and string_id_hash_table_.

Referenced by getOrAddBulk().

                                                                  {
  if (client_no_timeout_) {
    getOrAddBulkRemote(input_strings, output_string_ids);
    return;
  }
  // Run rk_hash on the input strings up front, and in parallel,
  // as the string hashing does not need to be behind the subsequent write_lock
  std::vector<uint32_t> input_strings_rk_hashes(input_strings.size());
  hashStrings(input_strings, input_strings_rk_hashes);

  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);
  size_t shadow_str_count =
      str_count_;  // Need to shadow str_count_ now with bulk add methods
  const size_t storage_high_water_mark = shadow_str_count;
  std::vector<size_t> string_memory_ids;
  size_t sum_new_string_lengths = 0;
  string_memory_ids.reserve(input_strings.size());
  size_t input_string_idx{0};
  for (const auto& input_string : input_strings) {
    // Currently we make empty strings null
    if (input_string.empty()) {
      output_string_ids[input_string_idx++] = inline_int_null_value<T>();
      continue;
    }
    // TODO: Recover gracefully if an input string is too long
    CHECK(input_string.size() <= MAX_STRLEN);

    if (fillRateIsHigh(shadow_str_count)) {
      // resize when more than 50% is full
      increaseCapacityFromStorageAndMemory(storage_high_water_mark,
                                           input_strings,
                                           string_memory_ids,
                                           input_strings_rk_hashes);
    }
    // Get the rk_hash for this input_string
    const uint32_t input_string_rk_hash = input_strings_rk_hashes[input_string_idx];

    uint32_t hash_bucket = computeBucketFromStorageAndMemory(input_string_rk_hash,
                                                             input_string,
                                                             string_id_hash_table_,
                                                             storage_high_water_mark,
                                                             input_strings,
                                                             string_memory_ids);

    // If the hash bucket is not empty, that is our string id
    // (computeBucketFromStorageAndMemory) already checked to ensure the input string and
    // bucket string are equal)
    if (string_id_hash_table_[hash_bucket] != INVALID_STR_ID) {
      output_string_ids[input_string_idx++] = string_id_hash_table_[hash_bucket];
      continue;
    }
    // Did not find string, so need to add record to dictionary
    // First check there is room
    if (shadow_str_count == static_cast<size_t>(max_valid_int_value<T>())) {
      log_encoding_error<T>(input_string);
      output_string_ids[input_string_idx++] = inline_int_null_value<T>();
      continue;
    }
    CHECK_LT(shadow_str_count, MAX_STRCOUNT)
        << "Maximum number (" << shadow_str_count
        << ") of Dictionary encoded Strings reached for this column, offset path "
           "for column is  "
        << offsets_path_;

    string_memory_ids.push_back(input_string_idx);
    sum_new_string_lengths += input_string.size();
    string_id_hash_table_[hash_bucket] = static_cast<int32_t>(shadow_str_count);
    if (materialize_hashes_) {
      rk_hashes_[shadow_str_count] = input_string_rk_hash;
    }
    output_string_ids[input_string_idx++] = shadow_str_count++;
  }
  appendToStorageBulk(input_strings, string_memory_ids, sum_new_string_lengths);
  str_count_ = shadow_str_count;

  invalidateInvertedIndex();
}

Here is the call graph for this function:

Here is the caller graph for this function:

getOrAddBulkRemote()

template<class T , class String >

template void StringDictionary::getOrAddBulkRemote ( const std::vector< String > &  string_vec, T *  encoded_vec  )

private

Definition at line 517 of file StringDictionary.cpp.

References CHECK, and client_no_timeout_.

Referenced by getOrAddBulk(), and getOrAddBulkParallel().

                                                          {
  CHECK(client_no_timeout_);
  std::vector<int32_t> string_ids;
  client_no_timeout_->get_or_add_bulk(string_ids, string_vec);
  size_t out_idx{0};
  for (size_t i = 0; i < string_ids.size(); ++i) {
    const auto string_id = string_ids[i];
    const bool invalid = string_id > max_valid_int_value<T>();
    if (invalid || string_id == inline_int_null_value<int32_t>()) {
      if (invalid) {
        log_encoding_error<T>(string_vec[i]);
      }
      encoded_vec[out_idx++] = inline_int_null_value<T>();
      continue;
    }
    encoded_vec[out_idx++] = string_id;
  }
}

Here is the caller graph for this function:

getOrAddImpl()

int32_t StringDictionary::getOrAddImpl ( const std::string &  str )

privatenoexcept

Definition at line 1062 of file StringDictionary.cpp.

References appendToStorage(), CHECK, CHECK_LT, computeBucket(), fillRateIsHigh(), increaseCapacity(), INVALID_STR_ID, invalidateInvertedIndex(), materialize_hashes_, MAX_STRCOUNT, MAX_STRLEN, offsets_path_, anonymous_namespace{StringDictionary.cpp}::rk_hash(), rk_hashes_, rw_mutex_, str_count_, and string_id_hash_table_.

Referenced by getOrAdd().

                                                                    {
  // @TODO(wei) treat empty string as NULL for now
  if (str.size() == 0) {
    return inline_int_null_value<int32_t>();
  }
  CHECK(str.size() <= MAX_STRLEN);
  uint32_t bucket;
  const uint32_t hash = rk_hash(str);
  {
    mapd_shared_lock<mapd_shared_mutex> read_lock(rw_mutex_);
    bucket = computeBucket(hash, str, string_id_hash_table_);
    if (string_id_hash_table_[bucket] != INVALID_STR_ID) {
      return string_id_hash_table_[bucket];
    }
  }
  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);
  // need to recalculate the bucket in case it changed before
  // we got the lock
  bucket = computeBucket(hash, str, string_id_hash_table_);
  if (string_id_hash_table_[bucket] == INVALID_STR_ID) {
    CHECK_LT(str_count_, MAX_STRCOUNT)
        << "Maximum number (" << str_count_
        << ") of Dictionary encoded Strings reached for this column, offset path "
           "for column is  "
        << offsets_path_;
    if (fillRateIsHigh(str_count_)) {
      // resize when more than 50% is full
      increaseCapacity();
      bucket = computeBucket(hash, str, string_id_hash_table_);
    }
    appendToStorage(str);
    string_id_hash_table_[bucket] = static_cast<int32_t>(str_count_);
    if (materialize_hashes_) {
      rk_hashes_[str_count_] = hash;
    }
    ++str_count_;
    invalidateInvertedIndex();
  }
  return string_id_hash_table_[bucket];
}

Here is the call graph for this function:

Here is the caller graph for this function:

getRegexpLike()

std::vector< int32_t > StringDictionary::getRegexpLike	(	const std::string &	pattern,
		const char	escape,
		const size_t	generation
	)		const

Definition at line 903 of file StringDictionary.cpp.

References CHECK, CHECK_GT, CHECK_LE, client_, cpu_threads(), getStringUnlocked(), anonymous_namespace{StringDictionary.cpp}::is_regexp_like(), regex_cache_, run_benchmark_import::result, rw_mutex_, and str_count_.

                                                                                    {
  mapd_lock_guard<mapd_shared_mutex> write_lock(rw_mutex_);
  if (client_) {
    return client_->get_regexp_like(pattern, escape, generation);
  }
  const auto cache_key = std::make_pair(pattern, escape);
  const auto it = regex_cache_.find(cache_key);
  if (it != regex_cache_.end()) {
    return it->second;
  }
  std::vector<int32_t> result;
  std::vector<std::thread> workers;
  int worker_count = cpu_threads();
  CHECK_GT(worker_count, 0);
  std::vector<std::vector<int32_t>> worker_results(worker_count);
  CHECK_LE(generation, str_count_);
  for (int worker_idx = 0; worker_idx < worker_count; ++worker_idx) {
    workers.emplace_back([&worker_results,
                          &pattern,
                          generation,
                          escape,
                          worker_idx,
                          worker_count,
                          this]() {
      for (size_t string_id = worker_idx; string_id < generation;
           string_id += worker_count) {
        const auto str = getStringUnlocked(string_id);
        if (is_regexp_like(str, pattern, escape)) {
          worker_results[worker_idx].push_back(string_id);
        }
      }
    });
  }
  for (auto& worker : workers) {
    worker.join();
  }
  for (const auto& worker_result : worker_results) {
    result.insert(result.end(), worker_result.begin(), worker_result.end());
  }
  const auto it_ok = regex_cache_.insert(std::make_pair(cache_key, result));
  CHECK(it_ok.second);

  return result;
}

Here is the call graph for this function:

getString()

std::string StringDictionary::getString

(

int32_t

string_id

)

const

Definition at line 571 of file StringDictionary.cpp.

References client_, getStringUnlocked(), and rw_mutex_.

Referenced by StringValueConverter::convertToColumnarFormatFromDict(), and populate_string_ids().

                                                             {
  mapd_shared_lock<mapd_shared_mutex> read_lock(rw_mutex_);
  if (client_) {
    std::string ret;
    client_->get_string(ret, string_id);
    return ret;
  }
  return getStringUnlocked(string_id);
}

Here is the call graph for this function:

Here is the caller graph for this function:

getStringBytes()

std::pair< char *, size_t > StringDictionary::getStringBytes ( int32_t  string_id ) const

noexcept

Definition at line 586 of file StringDictionary.cpp.

References CHECK, CHECK_LE, CHECK_LT, client_, getStringBytesChecked(), rw_mutex_, and str_count_.

             {
  mapd_shared_lock<mapd_shared_mutex> read_lock(rw_mutex_);
  CHECK(!client_);
  CHECK_LE(0, string_id);
  CHECK_LT(string_id, static_cast<int32_t>(str_count_));
  return getStringBytesChecked(string_id);
}

Here is the call graph for this function:

getStringBytesChecked()

std::pair< char *, size_t > StringDictionary::getStringBytesChecked ( const int  string_id ) const

privatenoexcept

Definition at line 1109 of file StringDictionary.cpp.

References CHECK, and getStringFromStorage().

Referenced by getStringBytes().

                                        {
  const auto str_canary = getStringFromStorage(string_id);
  CHECK(!str_canary.canary);
  return std::make_pair(str_canary.c_str_ptr, str_canary.size);
}

Here is the call graph for this function:

Here is the caller graph for this function:

getStringChecked()

std::string StringDictionary::getStringChecked ( const int  string_id ) const

privatenoexcept

Definition at line 1103 of file StringDictionary.cpp.

References CHECK, and getStringFromStorage().

Referenced by getStringUnlocked(), increaseCapacity(), and increaseCapacityFromStorageAndMemory().

                                                                               {
  const auto str_canary = getStringFromStorage(string_id);
  CHECK(!str_canary.canary);
  return std::string(str_canary.c_str_ptr, str_canary.size);
}

Here is the call graph for this function:

Here is the caller graph for this function:

getStringFromStorage()

StringDictionary::PayloadString StringDictionary::getStringFromStorage ( const int  string_id ) const

privatenoexcept

Definition at line 1294 of file StringDictionary.cpp.

References CHECK_GE, isTemp_, StringDictionary::StringIdxEntry::off, offset_fd_, offset_map_, payload_fd_, payload_map_, and StringDictionary::StringIdxEntry::size.

Referenced by getCompare(), getNumStringsFromStorage(), getStringBytesChecked(), getStringChecked(), mergeSortedCache(), sortCache(), and StringDictionary().

                                        {
  if (!isTemp_) {
    CHECK_GE(payload_fd_, 0);
    CHECK_GE(offset_fd_, 0);
  }
  CHECK_GE(string_id, 0);
  const StringIdxEntry* str_meta = offset_map_ + string_id;
  if (str_meta->size == 0xffff) {
    // hit the canary
    return {nullptr, 0, true};
  }
  return {payload_map_ + str_meta->off, str_meta->size, false};
}

Here is the caller graph for this function:

getStringFromStorageFast()

std::string_view StringDictionary::getStringFromStorageFast ( const int  string_id ) const

privatenoexcept

Definition at line 1288 of file StringDictionary.cpp.

References StringDictionary::StringIdxEntry::off, offset_map_, payload_map_, and StringDictionary::StringIdxEntry::size.

Referenced by computeBucket(), and computeBucketFromStorageAndMemory().

             {
  const StringIdxEntry* str_meta = offset_map_ + string_id;
  return {payload_map_ + str_meta->off, str_meta->size};
}

Here is the caller graph for this function:

getStringUnlocked()

std::string StringDictionary::getStringUnlocked ( int32_t  string_id ) const

privatenoexcept

Definition at line 581 of file StringDictionary.cpp.

References CHECK_LT, getStringChecked(), and str_count_.

Referenced by copyStrings(), getEquals(), getLike(), getRegexpLike(), and getString().

                                                                              {
  CHECK_LT(string_id, static_cast<int32_t>(str_count_));
  return getStringChecked(string_id);
}

Here is the call graph for this function:

Here is the caller graph for this function:

getUnlocked()

int32_t StringDictionary::getUnlocked ( const std::string &  str ) const

privatenoexcept

Definition at line 565 of file StringDictionary.cpp.

References computeBucket(), anonymous_namespace{StringDictionary.cpp}::rk_hash(), and string_id_hash_table_.

Referenced by getIdOfString().

                                                                         {
  const uint32_t hash = rk_hash(str);
  auto str_id = string_id_hash_table_[computeBucket(hash, str, string_id_hash_table_)];
  return str_id;
}

Here is the call graph for this function:

Here is the caller graph for this function:

hashStrings()

template<class String >

void StringDictionary::hashStrings ( const std::vector< String > &  string_vec, std::vector< uint32_t > &  hashes  ) const

privatenoexcept

Method to rk_hash a vector of strings in parallel.

Parameters

string_vec	input vector of strings to be hashed
hashes	space for the output - should be pre-sized to match string_vec size

Definition at line 346 of file StringDictionary.cpp.

References CHECK_EQ, and anonymous_namespace{StringDictionary.cpp}::rk_hash().

Referenced by getOrAddBulkParallel().

                                                                               {
  CHECK_EQ(string_vec.size(), hashes.size());

  tbb::parallel_for(tbb::blocked_range<size_t>(0, string_vec.size()),
                    [&string_vec, &hashes](const tbb::blocked_range<size_t>& r) {
                      for (size_t curr_id = r.begin(); curr_id != r.end(); ++curr_id) {
                        if (string_vec[curr_id].empty()) {
                          continue;
                        }
                        hashes[curr_id] = rk_hash(string_vec[curr_id]);
                      }
                    });
}

Here is the call graph for this function:

Here is the caller graph for this function:

increaseCapacity()

void StringDictionary::increaseCapacity ( )

privatenoexcept

Definition at line 1006 of file StringDictionary.cpp.

References computeUniqueBucketWithHash(), getStringChecked(), INVALID_STR_ID, materialize_hashes_, anonymous_namespace{StringDictionary.cpp}::rk_hash(), rk_hashes_, str_count_, and string_id_hash_table_.

Referenced by getOrAddBulk(), and getOrAddImpl().

                                                 {
  std::vector<int32_t> new_str_ids(string_id_hash_table_.size() * 2, INVALID_STR_ID);

  if (materialize_hashes_) {
    for (size_t i = 0; i < string_id_hash_table_.size(); ++i) {
      if (string_id_hash_table_[i] != INVALID_STR_ID) {
        const uint32_t hash = rk_hashes_[string_id_hash_table_[i]];
        uint32_t bucket = computeUniqueBucketWithHash(hash, new_str_ids);
        new_str_ids[bucket] = string_id_hash_table_[i];
      }
    }
    rk_hashes_.resize(rk_hashes_.size() * 2);
  } else {
    for (size_t i = 0; i < str_count_; ++i) {
      const auto str = getStringChecked(i);
      const uint32_t hash = rk_hash(str);
      uint32_t bucket = computeUniqueBucketWithHash(hash, new_str_ids);
      new_str_ids[bucket] = i;
    }
  }
  string_id_hash_table_.swap(new_str_ids);
}

Here is the call graph for this function:

Here is the caller graph for this function:

increaseCapacityFromStorageAndMemory()

template<class String >

void StringDictionary::increaseCapacityFromStorageAndMemory ( const size_t  storage_high_water_mark, const std::vector< String > &  input_strings, const std::vector< size_t > &  string_memory_ids, const std::vector< uint32_t > &  input_strings_rk_hashes  )

privatenoexcept

Definition at line 1030 of file StringDictionary.cpp.

References computeUniqueBucketWithHash(), getStringChecked(), INVALID_STR_ID, materialize_hashes_, anonymous_namespace{StringDictionary.cpp}::rk_hash(), rk_hashes_, and string_id_hash_table_.

Referenced by getOrAddBulkParallel().

                                                                 {
  std::vector<int32_t> new_str_ids(string_id_hash_table_.size() * 2, INVALID_STR_ID);
  if (materialize_hashes_) {
    for (size_t i = 0; i < string_id_hash_table_.size(); ++i) {
      if (string_id_hash_table_[i] != INVALID_STR_ID) {
        const uint32_t hash = rk_hashes_[string_id_hash_table_[i]];
        uint32_t bucket = computeUniqueBucketWithHash(hash, new_str_ids);
        new_str_ids[bucket] = string_id_hash_table_[i];
      }
    }
    rk_hashes_.resize(rk_hashes_.size() * 2);
  } else {
    for (size_t storage_idx = 0; storage_idx != storage_high_water_mark; ++storage_idx) {
      const auto storage_string = getStringChecked(storage_idx);
      const uint32_t hash = rk_hash(storage_string);
      uint32_t bucket = computeUniqueBucketWithHash(hash, new_str_ids);
      new_str_ids[bucket] = storage_idx;
    }
    for (size_t memory_idx = 0; memory_idx != string_memory_ids.size(); ++memory_idx) {
      size_t string_memory_id = string_memory_ids[memory_idx];
      uint32_t bucket = computeUniqueBucketWithHash(
          input_strings_rk_hashes[string_memory_id], new_str_ids);
      new_str_ids[bucket] = storage_high_water_mark + memory_idx;
    }
  }
  string_id_hash_table_.swap(new_str_ids);
}

Here is the call graph for this function:

Here is the caller graph for this function:

insertInSortedCache()

void StringDictionary::insertInSortedCache ( std::string  str, int32_t  str_id  )

private

invalidateInvertedIndex()

void StringDictionary::invalidateInvertedIndex ( )

privatenoexcept

Definition at line 1369 of file StringDictionary.cpp.

References compare_cache_, equal_cache_, like_cache_, and regex_cache_.

Referenced by getOrAddBulk(), getOrAddBulkParallel(), and getOrAddImpl().

                                                        {
  if (!like_cache_.empty()) {
    decltype(like_cache_)().swap(like_cache_);
  }
  if (!regex_cache_.empty()) {
    decltype(regex_cache_)().swap(regex_cache_);
  }
  if (!equal_cache_.empty()) {
    decltype(equal_cache_)().swap(equal_cache_);
  }
  compare_cache_.invalidateInvertedIndex();
}

Here is the caller graph for this function:

mergeSortedCache()

void StringDictionary::mergeSortedCache ( std::vector< int32_t > &  temp_sorted_cache )

private

Definition at line 1423 of file StringDictionary.cpp.

References getStringFromStorage(), sorted_cache, and string_lt().

Referenced by buildSortedCache().

                                                                             {
  // this method is not thread safe
  std::vector<int32_t> updated_cache(temp_sorted_cache.size() + sorted_cache.size());
  size_t t_idx = 0, s_idx = 0, idx = 0;
  for (; t_idx < temp_sorted_cache.size() && s_idx < sorted_cache.size(); idx++) {
    auto t_string = getStringFromStorage(temp_sorted_cache[t_idx]);
    auto s_string = getStringFromStorage(sorted_cache[s_idx]);
    const auto insert_from_temp_cache =
        string_lt(t_string.c_str_ptr, t_string.size, s_string.c_str_ptr, s_string.size);
    if (insert_from_temp_cache) {
      updated_cache[idx] = temp_sorted_cache[t_idx++];
    } else {
      updated_cache[idx] = sorted_cache[s_idx++];
    }
  }
  while (t_idx < temp_sorted_cache.size()) {
    updated_cache[idx++] = temp_sorted_cache[t_idx++];
  }
  while (s_idx < sorted_cache.size()) {
    updated_cache[idx++] = sorted_cache[s_idx++];
  }
  sorted_cache.swap(updated_cache);
}

Here is the call graph for this function:

Here is the caller graph for this function:

populate_string_array_ids()

void StringDictionary::populate_string_array_ids ( std::vector< std::vector< int32_t >> &  dest_array_ids, StringDictionary *  dest_dict, const std::vector< std::vector< int32_t >> &  source_array_ids, const StringDictionary *  source_dict  )

static

Definition at line 1474 of file StringDictionary.cpp.

References populate_string_ids(), and logger::thread_id().

Referenced by DictionaryValueConverter< TARGET_TYPE >::processArrayBuffer().

                                         {
  dest_array_ids.resize(source_array_ids.size());

  std::atomic<size_t> row_idx{0};
  auto processor = [&row_idx, &dest_array_ids, dest_dict, &source_array_ids, source_dict](
                       int thread_id) {
    for (;;) {
      auto row = row_idx.fetch_add(1);

      if (row >= dest_array_ids.size()) {
        return;
      }
      const auto& source_ids = source_array_ids[row];
      auto& dest_ids = dest_array_ids[row];
      populate_string_ids(dest_ids, dest_dict, source_ids, source_dict);
    }
  };

  const int num_worker_threads = std::thread::hardware_concurrency();

  if (source_array_ids.size() / num_worker_threads > 10) {
    std::vector<std::future<void>> worker_threads;
    for (int i = 0; i < num_worker_threads; ++i) {
      worker_threads.push_back(std::async(std::launch::async, processor, i));
    }

    for (auto& child : worker_threads) {
      child.wait();
    }
    for (auto& child : worker_threads) {
      child.get();
    }
  } else {
    processor(0);
  }
}

Here is the call graph for this function:

Here is the caller graph for this function:

populate_string_ids()

void StringDictionary::populate_string_ids ( std::vector< int32_t > &  dest_ids, StringDictionary *  dest_dict, const std::vector< int32_t > &  source_ids, const StringDictionary *  source_dict, const std::map< int32_t, std::string >  transient_mapping = {}  )

static

Populates provided dest_ids vector with string ids corresponding to given source strings.

Given a vector of source string ids and corresponding source dictionary, this method populates a vector of destination string ids by either returning the string id of matching strings in the destination dictionary or creating new entries in the dictionary. Source string ids can also be transient if they were created by a function (e.g LOWER/UPPER functions). A map of transient string ids to string values is provided in order to handle this use case.

Parameters

dest_ids	- vector of destination string ids to be populated
dest_dict	- destination dictionary
source_ids	- vector of source string ids for which destination ids are needed
source_dict	- source dictionary
transient_mapping	- map of transient source string ids to string values

Definition at line 1447 of file StringDictionary.cpp.

References getOrAddBulk(), and getString().

Referenced by populate_string_array_ids(), and DictionaryValueConverter< TARGET_TYPE >::processBuffer().

                                                        {
  std::vector<std::string> strings;

  for (const int32_t source_id : source_ids) {
    if (source_id == std::numeric_limits<int32_t>::min()) {
      strings.emplace_back("");
    } else if (source_id < 0) {
      if (auto string_itr = transient_mapping.find(source_id);
          string_itr != transient_mapping.end()) {
        strings.emplace_back(string_itr->second);
      } else {
        throw std::runtime_error("Unexpected negative source ID");
      }
    } else {
      strings.push_back(source_dict->getString(source_id));
    }
  }

  dest_ids.resize(strings.size());
  dest_dict->getOrAddBulk(strings, &dest_ids[0]);
}

Here is the call graph for this function:

Here is the caller graph for this function:

processDictionaryFutures()

void StringDictionary::processDictionaryFutures ( std::vector< std::future< std::vector< std::pair< uint32_t, unsigned int >>>> &  dictionary_futures )

private

Definition at line 226 of file StringDictionary.cpp.

References computeUniqueBucketWithHash(), materialize_hashes_, payload_file_off_, rk_hashes_, str_count_, and string_id_hash_table_.

Referenced by StringDictionary().

                            {
  for (auto& dictionary_future : dictionary_futures) {
    dictionary_future.wait();
    auto hashVec = dictionary_future.get();
    for (auto& hash : hashVec) {
      uint32_t bucket = computeUniqueBucketWithHash(hash.first, string_id_hash_table_);
      payload_file_off_ += hash.second;
      string_id_hash_table_[bucket] = static_cast<int32_t>(str_count_);
      if (materialize_hashes_) {
        rk_hashes_[str_count_] = hash.first;
      }
      ++str_count_;
    }
  }
  dictionary_futures.clear();
}

Here is the call graph for this function:

Here is the caller graph for this function:

sortCache()

void StringDictionary::sortCache ( std::vector< int32_t > &  cache )

private

Definition at line 1410 of file StringDictionary.cpp.

References getStringFromStorage(), and string_lt().

Referenced by buildSortedCache().

                                                          {
  // This method is not thread-safe.

  // this boost sort is creating some problems when we use UTF-8 encoded strings.
  // TODO (vraj): investigate What is wrong with boost sort and try to mitigate it.

  std::sort(cache.begin(), cache.end(), [this](int32_t a, int32_t b) {
    auto a_str = this->getStringFromStorage(a);
    auto b_str = this->getStringFromStorage(b);
    return string_lt(a_str.c_str_ptr, a_str.size, b_str.c_str_ptr, b_str.size);
  });
}

Here is the call graph for this function:

Here is the caller graph for this function:

storageEntryCount()

size_t StringDictionary::storageEntryCount

(

)

const

Definition at line 595 of file StringDictionary.cpp.

References client_, rw_mutex_, and str_count_.

Referenced by ArrowCsvForeignStorage::createDictionaryEncodedColumn().

                                                 {
  mapd_shared_lock<mapd_shared_mutex> read_lock(rw_mutex_);
  if (client_) {
    return client_->storage_entry_count();
  }
  return str_count_;
}

Here is the caller graph for this function:

Member Data Documentation

CANARY_BUFFER

char* StringDictionary::CANARY_BUFFER {nullptr}

private

Definition at line 226 of file StringDictionary.h.

Referenced by addMemoryCapacity(), addStorageCapacity(), and ~StringDictionary().

canary_buffer_size

size_t StringDictionary::canary_buffer_size = 0

private

Definition at line 227 of file StringDictionary.h.

Referenced by addMemoryCapacity(), and addStorageCapacity().

client_

std::unique_ptr<StringDictionaryClient> StringDictionary::client_

private

Definition at line 223 of file StringDictionary.h.

Referenced by checkpoint(), copyStrings(), getCompare(), getIdOfString(), getLike(), getOrAdd(), getRegexpLike(), getString(), getStringBytes(), storageEntryCount(), and ~StringDictionary().

client_no_timeout_

std::unique_ptr<StringDictionaryClient> StringDictionary::client_no_timeout_

private

Definition at line 224 of file StringDictionary.h.

Referenced by getOrAddBulk(), getOrAddBulkParallel(), and getOrAddBulkRemote().

collisions_

size_t StringDictionary::collisions_

private

Definition at line 202 of file StringDictionary.h.

Referenced by computeUniqueBucketWithHash(), and StringDictionary().

compare_cache_

DictionaryCache<std::string, compare_cache_value_t> StringDictionary::compare_cache_

mutableprivate

Definition at line 221 of file StringDictionary.h.

Referenced by getCompare(), and invalidateInvertedIndex().

equal_cache_

std::map<std::string, int32_t> StringDictionary::equal_cache_

mutableprivate

Definition at line 220 of file StringDictionary.h.

Referenced by getEquals(), and invalidateInvertedIndex().

INVALID_STR_ID

constexpr int32_t StringDictionary::INVALID_STR_ID = -1

static

Definition at line 113 of file StringDictionary.h.

Referenced by computeBucket(), computeBucketFromStorageAndMemory(), computeUniqueBucketWithHash(), count_matches_impl(), count_matches_sharded(), CodeGenerator::createInValuesBitmap(), anonymous_namespace{RelAlgTranslator.cpp}::fill_dictionary_encoded_in_vals(), fill_hash_join_buff_impl(), fill_hash_join_buff_sharded_impl(), fill_row_ids_impl(), fill_row_ids_sharded_impl(), StringDictionaryProxy::getIdOfString(), StringDictionaryProxy::getIdOfStringNoGeneration(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), StringDictionaryProxy::getOrAddTransient(), StringDictionaryProxy::getString(), StringDictionaryProxy::getStringBytes(), increaseCapacity(), increaseCapacityFromStorageAndMemory(), GenericKeyHandler::operator()(), StringDictionary(), anonymous_namespace{HashJoinRuntime.cpp}::translate_str_id_to_outer_dict(), and truncate_to_generation().

isTemp_

bool StringDictionary::isTemp_

private

Definition at line 206 of file StringDictionary.h.

Referenced by addOffsetCapacity(), addPayloadCapacity(), checkAndConditionallyIncreaseOffsetCapacity(), checkAndConditionallyIncreasePayloadCapacity(), checkpoint(), getStringFromStorage(), StringDictionary(), and ~StringDictionary().

like_cache_

std::map<std::tuple<std::string, bool, bool, char>, std::vector<int32_t> > StringDictionary::like_cache_

mutableprivate

Definition at line 218 of file StringDictionary.h.

Referenced by getLike(), and invalidateInvertedIndex().

materialize_hashes_

bool StringDictionary::materialize_hashes_

private

Definition at line 207 of file StringDictionary.h.

Referenced by computeBucket(), computeBucketFromStorageAndMemory(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), increaseCapacity(), increaseCapacityFromStorageAndMemory(), processDictionaryFutures(), and StringDictionary().

MAX_STRCOUNT

constexpr size_t StringDictionary::MAX_STRCOUNT = (1U << 31) - 1

static

Definition at line 115 of file StringDictionary.h.

Referenced by getOrAddBulk(), getOrAddBulkParallel(), and getOrAddImpl().

MAX_STRLEN

constexpr size_t StringDictionary::MAX_STRLEN = (1 << 15) - 1

static

Definition at line 114 of file StringDictionary.h.

Referenced by import_export::TypedImportBuffer::add_value(), import_export::TypedImportBuffer::addDictEncodedString(), import_export::TypedImportBuffer::addDictEncodedStringArray(), getEquals(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), and import_export::delimited_parser::parse_string_array().

offset_fd_

int StringDictionary::offset_fd_

private

Definition at line 210 of file StringDictionary.h.

Referenced by addOffsetCapacity(), checkAndConditionallyIncreaseOffsetCapacity(), checkpoint(), getStringFromStorage(), StringDictionary(), and ~StringDictionary().

offset_file_size_

size_t StringDictionary::offset_file_size_

private

Definition at line 213 of file StringDictionary.h.

Referenced by addOffsetCapacity(), checkAndConditionallyIncreaseOffsetCapacity(), checkpoint(), StringDictionary(), and ~StringDictionary().

offset_map_

StringIdxEntry* StringDictionary::offset_map_

private

Definition at line 211 of file StringDictionary.h.

Referenced by addOffsetCapacity(), appendToStorage(), appendToStorageBulk(), checkAndConditionallyIncreaseOffsetCapacity(), checkpoint(), getStringFromStorage(), getStringFromStorageFast(), StringDictionary(), and ~StringDictionary().

offsets_path_

std::string StringDictionary::offsets_path_

private

Definition at line 208 of file StringDictionary.h.

Referenced by getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), and StringDictionary().

payload_fd_

int StringDictionary::payload_fd_

private

Definition at line 209 of file StringDictionary.h.

Referenced by addPayloadCapacity(), checkAndConditionallyIncreasePayloadCapacity(), checkpoint(), getStringFromStorage(), StringDictionary(), and ~StringDictionary().

payload_file_off_

size_t StringDictionary::payload_file_off_

private

Definition at line 215 of file StringDictionary.h.

Referenced by appendToStorage(), appendToStorageBulk(), checkAndConditionallyIncreasePayloadCapacity(), and processDictionaryFutures().

payload_file_size_

size_t StringDictionary::payload_file_size_

private

Definition at line 214 of file StringDictionary.h.

Referenced by addPayloadCapacity(), checkAndConditionallyIncreasePayloadCapacity(), checkpoint(), StringDictionary(), and ~StringDictionary().

payload_map_

char* StringDictionary::payload_map_

private

Definition at line 212 of file StringDictionary.h.

Referenced by addPayloadCapacity(), appendToStorage(), appendToStorageBulk(), checkAndConditionallyIncreasePayloadCapacity(), checkpoint(), getStringFromStorage(), getStringFromStorageFast(), StringDictionary(), and ~StringDictionary().

regex_cache_

std::map<std::pair<std::string, char>, std::vector<int32_t> > StringDictionary::regex_cache_

mutableprivate

Definition at line 219 of file StringDictionary.h.

Referenced by getRegexpLike(), and invalidateInvertedIndex().

rk_hashes_

std::vector<uint32_t> StringDictionary::rk_hashes_

private

Definition at line 204 of file StringDictionary.h.

Referenced by computeBucket(), computeBucketFromStorageAndMemory(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), increaseCapacity(), increaseCapacityFromStorageAndMemory(), processDictionaryFutures(), and StringDictionary().

rw_mutex_

mapd_shared_mutex StringDictionary::rw_mutex_

mutableprivate

Definition at line 216 of file StringDictionary.h.

Referenced by copyStrings(), getCompare(), getIdOfString(), getLike(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), getRegexpLike(), getString(), getStringBytes(), storageEntryCount(), and StringDictionary().

sorted_cache

std::vector<int32_t> StringDictionary::sorted_cache

private

Definition at line 205 of file StringDictionary.h.

Referenced by buildSortedCache(), getCompare(), and mergeSortedCache().

str_count_

size_t StringDictionary::str_count_

private

Definition at line 201 of file StringDictionary.h.

Referenced by appendToStorage(), appendToStorageBulk(), buildSortedCache(), checkAndConditionallyIncreaseOffsetCapacity(), copyStrings(), getCompare(), getEquals(), getLike(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), getRegexpLike(), getStringBytes(), getStringUnlocked(), increaseCapacity(), processDictionaryFutures(), storageEntryCount(), and StringDictionary().

string_id_hash_table_

std::vector<int32_t> StringDictionary::string_id_hash_table_

private

Definition at line 203 of file StringDictionary.h.

Referenced by fillRateIsHigh(), getOrAddBulk(), getOrAddBulkParallel(), getOrAddImpl(), getUnlocked(), increaseCapacity(), increaseCapacityFromStorageAndMemory(), processDictionaryFutures(), and StringDictionary().

strings_cache_

std::shared_ptr<std::vector<std::string> > StringDictionary::strings_cache_

mutableprivate

Definition at line 222 of file StringDictionary.h.

Referenced by copyStrings().

---

The documentation for this class was generated from the following files:

• /home/jenkins-slave/workspace/core-os-doxygen/StringDictionary/StringDictionary.h
• /home/jenkins-slave/workspace/core-os-doxygen/StringDictionary/StringDictionary.cpp