TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap Struct Reference

A struct that creates a bytemap to map each key to its corresponding one-hot column index. More...

Public Member Functions
	KeyToOneHotColBytemap (const std::vector< int32_t > &top_k_keys, const int32_t min_key, const int32_t max_key, const bool has_other_key)
int16_t	get_col_idx_for_key (const int32_t key) const

Static Public Member Functions
static std::vector< int16_t >	init_bytemap (const std::vector< int32_t > &top_k_keys, const int32_t min_key, const int32_t max_key, const bool has_other_key)

Public Attributes
const int32_t	min_key_
const int32_t	max_key_
const bool	has_other_key_
const int32_t	other_key_
const std::vector< int16_t >	bytemap_

Detailed Description

A struct that creates a bytemap to map each key to its corresponding one-hot column index.

Definition at line 189 of file OneHotEncoder.cpp.

Constructor & Destructor Documentation

TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::KeyToOneHotColBytemap ( const std::vector< int32_t > &  top_k_keys, const int32_t  min_key, const int32_t  max_key, const bool  has_other_key  )

inline

Definition at line 190 of file OneHotEncoder.cpp.

      : min_key_(min_key)
      , max_key_(max_key)
      , has_other_key_(has_other_key)
      , other_key_(top_k_keys.size())
      , bytemap_(init_bytemap(top_k_keys, min_key, max_key, has_other_key)) {}

Member Function Documentation

int16_t TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::get_col_idx_for_key ( const int32_t  key ) const

inline

Definition at line 223 of file OneHotEncoder.cpp.

References bytemap_, has_other_key_, TableFunctions_Namespace::OneHotEncoder_Namespace::INVALID_COL_IDX, max_key_, min_key_, and other_key_.

Referenced by TableFunctions_Namespace::OneHotEncoder_Namespace::one_hot_encode().

                                                              {
    if (key < min_key_ || key > max_key_) {
      return has_other_key_ ? other_key_ : INVALID_COL_IDX;
    }
    return bytemap_[key - min_key_];
  }

Here is the caller graph for this function:

static std::vector<int16_t> TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::init_bytemap ( const std::vector< int32_t > &  top_k_keys, const int32_t  min_key, const int32_t  max_key, const bool  has_other_key  )

inlinestatic

Definition at line 200 of file OneHotEncoder.cpp.

References TableFunctions_Namespace::OneHotEncoder_Namespace::INVALID_COL_IDX.

                                                                     {
    // The bytemap can be quite large if the dictionary-encoded key range is large, so for
    // efficiency we store the offsets as int16_t Since we use `top_k_keys.size()` as the
    // sentinel for the OTHER key, we check to see if the top_k_keys.size() is smaller
    // than the maximum allowable value for int16_t
    if (static_cast<int64_t>(top_k_keys.size()) >= std::numeric_limits<int16_t>::max()) {
      std::ostringstream error_oss;
      error_oss << "Error: More than " << std::numeric_limits<int16_t>::max() - 1
                << " top k categorical keys not allowed.";
      throw std::runtime_error(error_oss.str());
    }
    std::vector<int16_t> bytemap(max_key - min_key + 1,
                                 has_other_key ? top_k_keys.size() : INVALID_COL_IDX);
    int16_t offset = 0;
    for (const auto& key : top_k_keys) {
      bytemap[key - min_key] = offset++;
    }
    return bytemap;
  }

Member Data Documentation

const std::vector<int16_t> TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::bytemap_

Definition at line 234 of file OneHotEncoder.cpp.

Referenced by get_col_idx_for_key().

const bool TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::has_other_key_

Definition at line 232 of file OneHotEncoder.cpp.

Referenced by get_col_idx_for_key().

const int32_t TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::max_key_

Definition at line 231 of file OneHotEncoder.cpp.

Referenced by get_col_idx_for_key().

const int32_t TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::min_key_

Definition at line 230 of file OneHotEncoder.cpp.

Referenced by get_col_idx_for_key().

const int32_t TableFunctions_Namespace::OneHotEncoder_Namespace::KeyToOneHotColBytemap::other_key_

Definition at line 233 of file OneHotEncoder.cpp.

Referenced by get_col_idx_for_key().

---

The documentation for this struct was generated from the following file:

• /home/jenkins-slave/workspace/core-os-doxygen/QueryEngine/TableFunctions/SystemFunctions/os/ML/OneHotEncoder.cpp