!4597 [Dataset] C++ API Support for build_vocab

mindspore/ccsrc/minddata/dataset/api/CMakeLists.txt

@@ -26,4 +26,5 @@ add_library(cpp-API OBJECT
         iterator.cc
         transforms.cc
         samplers.cc
+        text.cc
         )

mindspore/ccsrc/minddata/dataset/api/datasets.cc

@@ -34,6 +34,7 @@
 #include "minddata/dataset/engine/datasetops/source/voc_op.h"
 // Dataset operator headers (in alphabetical order)
 #include "minddata/dataset/engine/datasetops/batch_op.h"
+#include "minddata/dataset/engine/datasetops/build_vocab_op.h"
 #include "minddata/dataset/engine/datasetops/concat_op.h"
 #include "minddata/dataset/engine/datasetops/map_op/map_op.h"
 #include "minddata/dataset/engine/datasetops/project_op.h"
@@ -263,6 +264,37 @@ std::shared_ptr<BatchDataset> Dataset::Batch(int32_t batch_size, bool drop_remai
   return ds;
 }
 
+// Function to create a Vocab from dataset
+std::shared_ptr<Vocab> Dataset::BuildVocab(const std::vector<std::string> &columns,
+                                           const std::pair<int64_t, int64_t> &freq_range, int64_t top_k,
+                                           const std::vector<std::string> &special_tokens, bool special_first) {
+  auto vocab = std::make_shared<Vocab>();
+  auto ds = std::make_shared<BuildVocabDataset>(vocab, columns, freq_range, top_k, special_tokens, special_first);
+
+  if (!ds->ValidateParams()) {
+    return nullptr;
+  }
+
+  ds->children.push_back(shared_from_this());
+
+  // Run tree here to starting building vocab
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  if (iter == nullptr) {
+    MS_LOG(ERROR) << "Fail to run iterator in BuildVocab.";
+    return nullptr;
+  }
+
+  // Finish building vocab by triggering GetNextRow
+  std::unordered_map<std::string, std::shared_ptr<Tensor>> row;
+  iter->GetNextRow(&row);
+  if (vocab == nullptr) {
+    MS_LOG(ERROR) << "Fail to build vocab.";
+    return nullptr;
+  }
+
+  return vocab;
+}
+
 // Function to create a Concat dataset
 std::shared_ptr<ConcatDataset> Dataset::Concat(const std::vector<std::shared_ptr<Dataset>> &datasets) {
   auto ds = std::make_shared<ConcatDataset>(datasets);
@@ -1450,13 +1482,52 @@ std::vector<std::shared_ptr<DatasetOp>> BatchDataset::Build() {
 
 bool BatchDataset::ValidateParams() {
   if (batch_size_ <= 0) {
-    MS_LOG(ERROR) << "Batch: Batch size cannot be negative";
+    MS_LOG(ERROR) << "Batch: batch_size should be positive integer, but got: " << batch_size_;
     return false;
   }
 
   return true;
 }
 
+BuildVocabDataset::BuildVocabDataset(std::shared_ptr<Vocab> vocab, const std::vector<std::string> &columns,
+                                     const std::pair<int64_t, int64_t> &freq_range, int64_t top_k,
+                                     const std::vector<std::string> &special_tokens, bool special_first)
+    : vocab_(vocab),
+      columns_(columns),
+      freq_range_(freq_range),
+      top_k_(top_k),
+      special_tokens_(special_tokens),
+      special_first_(special_first) {}
+
+// Function to build BuildVocabDataset
+std::vector<std::shared_ptr<DatasetOp>> BuildVocabDataset::Build() {
+  // A vector containing shared pointer to the Dataset Ops that this object will create
+  std::vector<std::shared_ptr<DatasetOp>> node_ops;
+
+  std::shared_ptr<BuildVocabOp> build_vocab_op;
+  build_vocab_op = std::make_shared<BuildVocabOp>(vocab_, columns_, freq_range_, top_k_, special_tokens_,
+                                                  special_first_, num_workers_, connector_que_size_);
+  node_ops.push_back(build_vocab_op);
+  return node_ops;
+}
+
+bool BuildVocabDataset::ValidateParams() {
+  if (vocab_ == nullptr) {
+    MS_LOG(ERROR) << "BuildVocab: vocab is null.";
+    return false;
+  }
+  if (top_k_ < 0) {
+    MS_LOG(ERROR) << "BuildVocab: top_k shoule be positive, but got: " << top_k_;
+    return false;
+  }
+  if (freq_range_.first < 0 || freq_range_.second > kDeMaxFreq || freq_range_.first > freq_range_.second) {
+    MS_LOG(ERROR) << "BuildVocab: requency_range [a,b] should be 0 <= a <= b (a,b are inclusive), "
+                  << "but got [" << freq_range_.first << ", " << freq_range_.second << "]";
+    return false;
+  }
+  return true;
+}
+
 // Function to build ConcatOp
 ConcatDataset::ConcatDataset(const std::vector<std::shared_ptr<Dataset>> &datasets) : datasets_(datasets) {
   this->children = datasets_;

mindspore/ccsrc/minddata/dataset/api/text.cc

@@ -0,0 +1,64 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * 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.
+ */
+
+#include "minddata/dataset/include/text.h"
+#include "minddata/dataset/text/kernels/lookup_op.h"
+
+namespace mindspore {
+namespace dataset {
+namespace api {
+namespace text {
+
+std::shared_ptr<LookupOperation> Lookup(const std::shared_ptr<Vocab> &vocab, const std::string &unknown_token) {
+  auto op = std::make_shared<LookupOperation>(vocab, unknown_token);
+
+  if (!op->ValidateParams()) {
+    return nullptr;
+  }
+  return op;
+}
+
+// LookupOperation
+LookupOperation::LookupOperation(const std::shared_ptr<Vocab> &vocab, const std::string &unknown_token)
+    : vocab_(vocab), unknown_token_(unknown_token), default_id_(Vocab::kNoTokenExists) {}
+
+bool LookupOperation::ValidateParams() {
+  if (vocab_ == nullptr) {
+    LOG(ERROR) << "Lookup: vocab object type is incorrect or null.";
+    return false;
+  }
+  if (unknown_token_.empty()) {
+    LOG(ERROR) << "Lookup: no unknown token is specified.";
+    return false;
+  } else {
+    default_id_ = vocab_->Lookup(unknown_token_);
+    if (default_id_ == Vocab::kNoTokenExists) {
+      LOG(ERROR) << "Lookup: unknown_token: [" + unknown_token_ + "], does not exist in vocab.";
+      return false;
+    }
+  }
+  return true;
+}
+
+std::shared_ptr<TensorOp> LookupOperation::Build() {
+  std::shared_ptr<LookupOp> tensor_op = std::make_shared<LookupOp>(vocab_, default_id_);
+  return tensor_op;
+}
+
+}  // namespace text
+}  // namespace api
+}  // namespace dataset
+}  // namespace mindspore

mindspore/ccsrc/minddata/dataset/core/constants.h

@@ -59,6 +59,8 @@ inline void BitClear(uint32_t *bits, uint32_t bitMask) { *bits &= (~bitMask); }
 
 constexpr int32_t kDeMaxDim = std::numeric_limits<int32_t>::max();  // 2147483647 or 2^32 -1
 constexpr int32_t kDeMaxRank = std::numeric_limits<int32_t>::max();
+constexpr int64_t kDeMaxFreq = std::numeric_limits<int64_t>::max();  // 9223372036854775807 or 2^(64-1)
+constexpr int64_t kDeMaxTopk = std::numeric_limits<int64_t>::max();
 
 constexpr uint32_t kCfgRowsPerBuffer = 1;
 constexpr uint32_t kCfgParallelWorkers = 4;

mindspore/ccsrc/minddata/dataset/include/datasets.h

@@ -30,6 +30,7 @@
 #include "minddata/dataset/include/iterator.h"
 #include "minddata/dataset/include/samplers.h"
 #include "minddata/dataset/include/type_id.h"
+#include "minddata/dataset/text/vocab.h"
 
 namespace mindspore {
 namespace dataset {
@@ -39,6 +40,7 @@ class DatasetOp;
 class DataSchema;
 class Tensor;
 class TensorShape;
+class Vocab;
 
 namespace api {
 
@@ -61,6 +63,7 @@ class TextFileDataset;
 class VOCDataset;
 // Dataset Op classes (in alphabetical order)
 class BatchDataset;
+class BuildVocabDataset;
 class ConcatDataset;
 class MapDataset;
 class ProjectDataset;
@@ -325,6 +328,24 @@ class Dataset : public std::enable_shared_from_this<Dataset> {
   /// \return Shared pointer to the current BatchDataset
   std::shared_ptr<BatchDataset> Batch(int32_t batch_size, bool drop_remainder = false);
 
+  /// \brief Function to create a Vocab from source dataset
+  /// \notes Build a vocab from a dataset. This would collect all the unique words in a dataset and return a vocab
+  ///    which contains top_k most frequent words (if top_k is specified)
+  /// \param[in] columns Column names to get words from. It can be a vector of column names
+  /// \param[in] freq_range A tuple of integers (min_frequency, max_frequency). Words within the frequency
+  ///    range would be kept. 0 <= min_frequency <= max_frequency <= total_words. min_frequency/max_frequency
+  ///    can be set to default, which corresponds to 0/total_words separately
+  /// \param[in] top_k Number of words to be built into vocab. top_k most frequent words are
+  //     taken. The top_k is taken after freq_range. If not enough top_k, all words will be taken
+  /// \param[in] special_tokens A list of strings, each one is a special token
+  /// \param[in] special_first Whether special_tokens will be prepended/appended to vocab, If special_tokens
+  ///    is specified and special_first is set to default, special_tokens will be prepended
+  /// \return Shared pointer to the current Vocab
+  std::shared_ptr<Vocab> BuildVocab(const std::vector<std::string> &columns = {},
+                                    const std::pair<int64_t, int64_t> &freq_range = {0, kDeMaxFreq},
+                                    int64_t top_k = kDeMaxTopk, const std::vector<std::string> &special_tokens = {},
+                                    bool special_first = true);
+
   /// \brief Function to create a ConcatDataset
   /// \notes Concat the datasets in the input
   /// \param[in] datasets List of shared pointers to the dataset that should be concatenated together
@@ -859,6 +880,33 @@ class BatchDataset : public Dataset {
   std::map<std::string, std::pair<TensorShape, std::shared_ptr<Tensor>>> pad_map_;
 };
 
+class BuildVocabDataset : public Dataset {
+ public:
+  /// \brief Constructor
+  BuildVocabDataset(std::shared_ptr<Vocab> vocab, const std::vector<std::string> &columns,
+                    const std::pair<int64_t, int64_t> &freq_range, int64_t top_k,
+                    const std::vector<std::string> &special_tokens, bool special_first);
+
+  /// \brief Destructor
+  ~BuildVocabDataset() = default;
+
+  /// \brief a base class override function to create the required runtime dataset op objects for this class
+  /// \return The list of shared pointers to the newly created DatasetOps
+  std::vector<std::shared_ptr<DatasetOp>> Build() override;
+
+  /// \brief Parameters validation
+  /// \return bool true if all the params are valid
+  bool ValidateParams() override;
+
+ private:
+  std::shared_ptr<Vocab> vocab_;
+  std::vector<std::string> columns_;
+  std::pair<int64_t, int64_t> freq_range_;
+  int64_t top_k_;
+  std::vector<std::string> special_tokens_;
+  bool special_first_;
+};
+
 class ConcatDataset : public Dataset {
  public:
   /// \brief Constructor

mindspore/ccsrc/minddata/dataset/include/text.h

@@ -0,0 +1,65 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * 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.
+ */
+
+#ifndef MINDSPORE_CCSRC_MINDDATA_DATASET_API_TEXT_H_
+#define MINDSPORE_CCSRC_MINDDATA_DATASET_API_TEXT_H_
+
+#include <vector>
+#include <memory>
+#include <string>
+#include "minddata/dataset/core/constants.h"
+#include "minddata/dataset/include/transforms.h"
+#include "minddata/dataset/text/vocab.h"
+
+namespace mindspore {
+namespace dataset {
+namespace api {
+
+// Transform operations for text
+namespace text {
+
+// Text Op classes (in alphabetical order)
+class LookupOperation;
+
+/// \brief Lookup operator that looks up a word to an id.
+/// \param[in] vocab a Vocab object.
+/// \param[in] unknown_token word to use for lookup if the word being looked up is out of Vocabulary (oov).
+///   If unknown_token is oov, runtime error will be thrown
+/// \return Shared pointer to the current TensorOperation.
+std::shared_ptr<LookupOperation> Lookup(const std::shared_ptr<Vocab> &vocab, const std::string &unknown_token);
+
+/* ####################################### Derived TensorOperation classes ################################# */
+
+class LookupOperation : public TensorOperation {
+ public:
+  explicit LookupOperation(const std::shared_ptr<Vocab> &vocab, const std::string &unknown_token);
+
+  ~LookupOperation() = default;
+
+  std::shared_ptr<TensorOp> Build() override;
+
+  bool ValidateParams() override;
+
+ private:
+  std::shared_ptr<Vocab> vocab_;
+  std::string unknown_token_;
+  int32_t default_id_;
+};
+}  // namespace text
+}  // namespace api
+}  // namespace dataset
+}  // namespace mindspore
+#endif  // MINDSPORE_CCSRC_MINDDATA_DATASET_API_TEXT_H_

mindspore/ccsrc/minddata/dataset/text/vocab.cc

@@ -17,8 +17,10 @@
 #include <unordered_set>
 #include <unordered_map>
 #include <utility>
+#include <algorithm>
 
 #include "minddata/dataset/text/vocab.h"
+#include "utils/log_adapter.h"
 
 namespace mindspore {
 namespace dataset {
@@ -51,6 +53,147 @@ Status Vocab::BuildFromPyList(const py::list &words, const py::list &special_tok
   return Status::OK();
 }
 
+Status Vocab::BuildFromPyDict(const py::dict &words, std::shared_ptr<Vocab> *vocab) {
+  std::unordered_map<WordType, WordIdType> word2id;
+  for (auto p : words) {
+    word2id[py::str(p.first)] = py::reinterpret_borrow<py::int_>(p.second);
+  }
+  *vocab = std::make_shared<Vocab>(std::move(word2id));
+  return Status::OK();
+}
+
+void Vocab::append_word(const std::string &word) {
+  if (word2id_.find(word) == word2id_.end()) {
+    word2id_[word] = word2id_.size();
+  }
+}
+
+Status Vocab::BuildFromUnorderedMap(const std::unordered_map<WordType, WordIdType> &words,
+                                    std::shared_ptr<Vocab> *vocab) {
+  // Validate parameters and build map
+  std::unordered_map<WordType, WordIdType> word2id;
+  for (auto p : words) {
+    if (p.second < 0) {
+      MS_LOG(ERROR) << "index can not be negetive, but got " << p.second;
+      RETURN_STATUS_UNEXPECTED("index can not be negetive, but got " + std::to_string(p.second));
+    }
+    word2id[p.first] = p.second;
+  }
+  *vocab = std::make_shared<Vocab>(std::move(word2id));
+  return Status::OK();
+}
+
+Status Vocab::BuildFromVector(const std::vector<WordType> &words, const std::vector<WordType> &special_tokens,
+                              bool prepend_special, std::shared_ptr<Vocab> *vocab) {
+  // Validate parameters
+  std::string duplicate_word;
+  for (const WordType &word : words) {
+    if (std::count(words.begin(), words.end(), word) > 1) {
+      if (duplicate_word.find(word) == std::string::npos) {
+        duplicate_word = duplicate_word + ", " + word;
+      }
+    }
+  }
+  if (!duplicate_word.empty()) {
+    MS_LOG(ERROR) << "words contains duplicate word: " << duplicate_word;
+    RETURN_STATUS_UNEXPECTED("words contains duplicate word: " + duplicate_word);
+  }
+
+  std::string duplicate_sp;
+  for (const WordType &sp : special_tokens) {
+    if (std::count(special_tokens.begin(), special_tokens.end(), sp) > 1) {
+      if (duplicate_sp.find(sp) == std::string::npos) {
+        duplicate_sp = duplicate_sp + ", " + sp;
+      }
+    }
+  }
+  if (!duplicate_sp.empty()) {
+    MS_LOG(ERROR) << "special_tokens contains duplicate word: " << duplicate_sp;
+    RETURN_STATUS_UNEXPECTED("special_tokens contains duplicate word: " + duplicate_sp);
+  }
+
+  std::unordered_map<WordType, WordIdType> word2id;
+
+  // if special is added in front, normal words id will start from number of special tokens
+  WordIdType word_id = prepend_special ? static_cast<WordIdType>(special_tokens.size()) : 0;
+  for (auto word : words) {
+    word2id[word] = word_id++;
+  }
+
+  word_id = prepend_special ? 0 : word2id.size();
+
+  for (auto special_token : special_tokens) {
+    word2id[special_token] = word_id++;
+  }
+
+  *vocab = std::make_shared<Vocab>(std::move(word2id));
+  return Status::OK();
+}
+
+Status Vocab::BuildFromFileCpp(const std::string &path, const std::string &delimiter, int32_t vocab_size,
+                               const std::vector<WordType> &special_tokens, bool prepend_special,
+                               std::shared_ptr<Vocab> *vocab) {
+  // Validate parameters
+  if (vocab_size < 0 && vocab_size != -1) {
+    MS_LOG(ERROR) << "vocab_size shoule be either -1 or positive integer, but got " << vocab_size;
+    RETURN_STATUS_UNEXPECTED("vocab_size shoule be either -1 or positive integer, but got " +
+                             std::to_string(vocab_size));
+  }
+
+  std::string duplicate_sp;
+  for (const WordType &sp : special_tokens) {
+    if (std::count(special_tokens.begin(), special_tokens.end(), sp) > 1) {
+      if (duplicate_sp.find(sp) == std::string::npos) {
+        duplicate_sp = duplicate_sp + ", " + sp;
+      }
+    }
+  }
+  if (!duplicate_sp.empty()) {
+    MS_LOG(ERROR) << "special_tokens contains duplicate word: " << duplicate_sp;
+    RETURN_STATUS_UNEXPECTED("special_tokens contains duplicate word: " + duplicate_sp);
+  }
+
+  std::unordered_set<std::string> specials;
+  // used to check that words in file don't contain any special token that already exists
+  for (auto word : special_tokens) {
+    specials.insert(word);
+  }
+  WordIdType word_id = prepend_special ? static_cast<WordIdType>(special_tokens.size()) : 0;
+  std::unordered_map<WordType, WordIdType> word2id;
+  std::fstream handle(path, std::ios::in);
+  if (!handle.good() || !handle.is_open()) {
+    MS_LOG(ERROR) << "fail to open:" + path;
+    RETURN_STATUS_UNEXPECTED("fail to open:" + path);
+  }
+  std::string word;
+  while (std::getline(handle, word)) {
+    if (!delimiter.empty()) {
+      // if delimiter is not found, find_first_of would return std::string::npos which is -1
+      word = word.substr(0, word.find_first_of(delimiter));
+    }
+    if (word2id.find(word) != word2id.end()) {
+      MS_LOG(ERROR) << "duplicate word:" + word + ".";
+      RETURN_STATUS_UNEXPECTED("duplicate word:" + word + ".");
+    }
+    if (specials.find(word) != specials.end()) {
+      MS_LOG(ERROR) << word + " is already in special_tokens.";
+      RETURN_STATUS_UNEXPECTED(word + " is already in special_tokens.");
+    }
+    word2id[word] = word_id++;
+    // break if enough row is read, if vocab_size is smaller than 0
+    if (word2id.size() == vocab_size) break;
+  }
+
+  word_id = prepend_special ? 0 : word2id.size();
+
+  for (auto special_token : special_tokens) {
+    word2id[special_token] = word_id++;
+  }
+
+  *vocab = std::make_shared<Vocab>(std::move(word2id));
+  return Status::OK();
+}
+
 Status Vocab::BuildFromFile(const std::string &path, const std::string &delimiter, int32_t vocab_size,
                             const py::list &special_tokens, bool prepend_special, std::shared_ptr<Vocab> *vocab) {
   // python validator checks special_tokens doesn't contain any duplicate words
@@ -86,21 +229,6 @@ Status Vocab::BuildFromFile(const std::string &path, const std::string &delimite
   return Status::OK();
 }
 
-Status Vocab::BuildFromPyDict(const py::dict &words, std::shared_ptr<Vocab> *vocab) {
-  std::unordered_map<WordType, WordIdType> word2id;
-  for (auto p : words) {
-    word2id[py::str(p.first)] = py::reinterpret_borrow<py::int_>(p.second);
-  }
-  *vocab = std::make_shared<Vocab>(std::move(word2id));
-  return Status::OK();
-}
-
-void Vocab::append_word(const std::string &word) {
-  if (word2id_.find(word) == word2id_.end()) {
-    word2id_[word] = word2id_.size();
-  }
-}
-
 const WordIdType Vocab::kNoTokenExists = -1;
 
 }  // namespace dataset

mindspore/ccsrc/minddata/dataset/text/vocab.h

@@ -57,6 +57,34 @@ class Vocab {
   static Status BuildFromFile(const std::string &path, const std::string &delimiter, int32_t vocab_size,
                               const py::list &special_tokens, bool prepend_special, std::shared_ptr<Vocab> *vocab);
 
+  /// \brief Build a vocab from a c++ map. id needs to start from 2, no duplicate and continuous
+  /// \param[in] words An unordered_map containing word, word id pair.
+  /// \param[out] vocab A vocab object
+  /// \return Error code
+  static Status BuildFromUnorderedMap(const std::unordered_map<WordType, WordIdType> &words,
+                                      std::shared_ptr<Vocab> *vocab);
+
+  /// \brief Build a vocab from a c++ vector. id needs to start from 2, no duplicate and continuous
+  /// \param[in] words A vector of string, used to build vocab, id starts from 2
+  /// \param[in] special_tokens A vector of string contain special tokens
+  /// \param[in] prepend_special Whether special_tokens will be prepended/appended to vocab
+  /// \param[out] vocab A vocab object
+  /// \return Error code
+  static Status BuildFromVector(const std::vector<WordType> &words, const std::vector<WordType> &special_tokens,
+                                bool prepend_special, std::shared_ptr<Vocab> *vocab);
+
+  /// \brief Build a vocab from reading a vocab file, id are automatically assigned, start from 2
+  /// \param[in] path Path to vocab file , each line is assumed to contain 1 word
+  /// \param[in] delimiter Delimiter to break each line with
+  /// \param[in] vocab_size Number of words to read from file
+  /// \param[in] special_tokens A vector of string contain special tokens
+  /// \param[in] prepend_special Whether special_tokens will be prepended/appended to vocab
+  /// \param[out] vocab A vocab object
+  /// \return Error code
+  static Status BuildFromFileCpp(const std::string &path, const std::string &delimiter, int32_t vocab_size,
+                                 const std::vector<WordType> &special_tokens, bool prepend_special,
+                                 std::shared_ptr<Vocab> *vocab);
+
   // Lookup the id of a word, if word doesn't exist in vocab, return default_id
   // @param const WordType word - word to look up
   // @param WordIdType default_id - word id to return to user when its not in the vocab

tests/ut/cpp/dataset/CMakeLists.txt

@@ -97,6 +97,7 @@ SET(DE_UT_SRCS
         concatenate_op_test.cc
         cyclic_array_test.cc
         perf_data_test.cc
+        build_vocab_test.cc
         c_api_samplers_test.cc
         c_api_transforms_test.cc
         c_api_dataset_ops_test.cc
@@ -104,12 +105,13 @@ SET(DE_UT_SRCS
         c_api_dataset_clue_test.cc
         c_api_dataset_coco_test.cc
         c_api_dataset_csv_test.cc
-        c_api_dataset_filetext_test.cc
+        c_api_dataset_textfile_test.cc
         c_api_dataset_manifest_test.cc
         c_api_dataset_randomdata_test.cc
         c_api_dataset_voc_test.cc
         c_api_datasets_test.cc
         c_api_dataset_iterator_test.cc
+        c_api_dataset_vocab.cc
         tensor_op_fusion_pass_test.cc
         sliding_window_op_test.cc
         epoch_ctrl_op_test.cc

tests/ut/cpp/dataset/build_vocab_test.cc

@@ -0,0 +1,229 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * 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.
+ */
+#include <fstream>
+#include <iostream>
+#include <memory>
+#include <vector>
+#include <string>
+
+#include "common/common.h"
+#include "minddata/dataset/include/datasets.h"
+#include "minddata/dataset/include/status.h"
+
+using mindspore::dataset::Tensor;
+using mindspore::dataset::Status;
+using mindspore::dataset::Vocab;
+
+class MindDataTestVocab : public UT::DatasetOpTesting {
+ protected:
+};
+
+TEST_F(MindDataTestVocab, TestVocabFromUnorderedMap) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromUnorderedMap.";
+  // Build a map
+  std::unordered_map<std::string, int32_t> dict;
+  dict["banana"] = 0;
+  dict["apple"] = 1;
+  dict["cat"] = 2;
+  dict["dog"] = 3;
+
+  // Build vocab from map
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromUnorderedMap(dict, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  std::vector<std::string> words = {"apple", "dog", "egg"};
+  std::vector<int32_t> expected = {1, 3, -1};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromEmptyMap) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromEmptyMap.";
+  // Build vocab from empty map
+  std::unordered_map<std::string, int32_t> dict;
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromUnorderedMap(dict, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  // Expect that we will return -1 when word is not in vocab
+  std::vector<std::string> words = {"apple", "dog", "egg"};
+  std::vector<int32_t> expected = {-1, -1, -1};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromMapFail) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromMapFail.";
+  // Build a map
+  std::unordered_map<std::string, int32_t> dict;
+  dict["banana"] = 0;
+  dict["apple"] = -1;
+
+  // Expected failure: index of word can not be negative
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromUnorderedMap(dict, &vocab);
+  EXPECT_NE(s, Status::OK());
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromVectorPrependSpTokens) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromVectorPrependSpTokens.";
+  // Build vocab from a vector of words, special tokens are prepended to vocab
+  std::vector<std::string> list = {"apple", "banana", "cat", "dog", "egg"};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromVector(list, {"<unk>"}, true, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  // Expect that we will return -1 when word is not in vocab
+  std::vector<std::string> words = {"apple", "banana", "fox"};
+  std::vector<int32_t> expected = {1, 2, -1};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromVectorAppendSpTokens) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromVectorAppendSpTokens.";
+  // Build vocab from a vector of words, special tokens are appended to vocab
+  std::vector<std::string> list = {"apple", "banana", "cat", "dog", "egg"};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromVector(list, {"<unk>"}, false, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  std::vector<std::string> words = {"apple", "<unk>", "fox"};
+  std::vector<int32_t> expected = {0, 5, -1};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromVectorWithNoSpTokens) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromVectorWithNoSpTokens.";
+  // Build vocab from a vector of words with no special tokens
+  std::vector<std::string> list = {"apple", "banana", "cat", "dog", "egg"};
+  std::vector<std::string> sp_tokens = {};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromVector(list, sp_tokens, true, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  std::vector<std::string> words = {"apple", "banana", "fox", "<pad>"};
+  std::vector<int32_t> expected = {0, 1, -1, -1};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromEmptyVector) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromEmptyVector.";
+  // Build vocab from empty vector
+  std::vector<std::string> list = {};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromVector(list, {}, false, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  // Expect that we will return -1 when word is not in vocab
+  std::vector<std::string> words = {"apple", "banana", "fox"};
+  std::vector<int32_t> expected = {-1, -1, -1};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromVectorFail1) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromVectorFail1.";
+  // Build vocab from a vector of words with no special tokens
+  std::vector<std::string> list = {"apple", "apple", "cat", "cat", "egg"};
+  std::vector<std::string> sp_tokens = {};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+
+  // Expected failure: duplicate word apple
+  Status s = Vocab::BuildFromVector(list, sp_tokens, true, &vocab);
+  EXPECT_NE(s, Status::OK());
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromVectorFail2) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromVectorFail2.";
+  // Build vocab from a vector of words with no special tokens
+  std::vector<std::string> list = {"apple", "dog", "egg"};
+  std::vector<std::string> sp_tokens = {"<pad>", "<unk>", "<pad>", "<unk>", "<none>"};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+
+  // Expected failure: duplicate special token <pad> <unk>
+  Status s = Vocab::BuildFromVector(list, sp_tokens, true, &vocab);
+  EXPECT_NE(s, Status::OK());
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromFile) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromFile.";
+  // Build vocab from local file
+  std::string vocab_dir = datasets_root_path_ + "/testVocab/vocab_list.txt";
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromFileCpp(vocab_dir, ",", -1, {"<pad>", "<unk>"}, true, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Look up specified words
+  std::vector<std::string> words = {"not", "all"};
+  std::vector<int32_t> expected = {2, 3};
+  for (uint32_t i = 0; i < words.size(); ++i) {
+    int32_t x = vocab->Lookup(words[i]);
+    EXPECT_EQ(x, expected[i]);
+  }
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromFileFail1) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromFileFail1.";
+  // Build vocab from local file which is not exist
+  std::string vocab_dir = datasets_root_path_ + "/testVocab/not_exist.txt";
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromFileCpp(vocab_dir, ",", -1, {}, true, &vocab);
+  EXPECT_NE(s, Status::OK());
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromFileFail2) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromFileFail2.";
+  // Build vocab from local file
+  std::string vocab_dir = datasets_root_path_ + "/testVocab/vocab_list.txt";
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+
+  // Expected failure: vocab_size shoule be either -1 or positive integer
+  Status s = Vocab::BuildFromFileCpp(vocab_dir, ",", -2, {}, true, &vocab);
+  EXPECT_NE(s, Status::OK());
+}
+
+TEST_F(MindDataTestVocab, TestVocabFromFileFail3) {
+  MS_LOG(INFO) << "Doing MindDataTestVocab-TestVocabFromFileFail2.";
+  // Build vocab from local file which is not exist
+  std::string vocab_dir = datasets_root_path_ + "/testVocab/vocab_list.txt";
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+
+  // Expected failure: duplicate special token <unk>
+  Status s = Vocab::BuildFromFileCpp(vocab_dir, ",", -1, {"<unk>", "<unk>"}, true, &vocab);
+  EXPECT_NE(s, Status::OK());
+}

tests/ut/cpp/dataset/c_api_dataset_voc_test.cc

@@ -14,7 +14,6 @@
  * limitations under the License.
  */
 #include "common/common.h"
-#include "minddata/dataset/engine/datasetops/source/voc_op.h"
 #include "minddata/dataset/include/datasets.h"
 
 using namespace mindspore::dataset::api;

tests/ut/cpp/dataset/c_api_dataset_vocab.cc

@@ -0,0 +1,254 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * 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.
+ */
+#include <fstream>
+#include <iostream>
+#include <memory>
+#include <vector>
+#include <string>
+
+#include "common/common.h"
+#include "minddata/dataset/include/datasets.h"
+#include "minddata/dataset/include/status.h"
+#include "minddata/dataset/include/transforms.h"
+#include "minddata/dataset/include/text.h"
+
+using namespace mindspore::dataset::api;
+using mindspore::dataset::ShuffleMode;
+using mindspore::dataset::Tensor;
+using mindspore::dataset::Status;
+using mindspore::dataset::Vocab;
+
+class MindDataTestPipeline : public UT::DatasetOpTesting {
+ protected:
+};
+
+TEST_F(MindDataTestPipeline, TestVocabLookupOp) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabLookupOp.";
+
+  // Create a TextFile dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Create a vocab from vector
+  std::vector<std::string> list = {"home", "IS", "behind", "the", "world", "ahead", "!"};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromVector(list, {"<pad>", "<unk>"}, true, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Create Lookup operation on ds
+  std::shared_ptr<TensorOperation> lookup = text::Lookup(vocab, "<unk>");
+  EXPECT_NE(lookup, nullptr);
+
+  // Create Map operation on ds
+  ds = ds->Map({lookup}, {"text"});
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, std::shared_ptr<Tensor>> row;
+  iter->GetNextRow(&row);
+
+  uint64_t i = 0;
+  std::vector<int32_t> expected = {2, 1, 4, 5, 6, 7};
+  while (row.size() != 0) {
+    auto ind = row["text"];
+    MS_LOG(INFO) << ind->shape() << " " << *ind;
+    std::shared_ptr<Tensor> expected_item;
+    Tensor::CreateScalar(expected[i], &expected_item);
+    EXPECT_EQ(*ind, *expected_item);
+    iter->GetNextRow(&row);
+    i++;
+  }
+}
+
+TEST_F(MindDataTestPipeline, TestVocabLookupOpFail1) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabLookupOpFail1.";
+  // Create a TextFile Dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Build vocab from vector
+  std::vector<std::string> list = {"home", "IS", "behind", "the", "world", "ahead", "!"};
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromVector(list, {}, true, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Create lookup op for ds
+  // Expected failure: "<unk>" is not a word of vocab
+  std::shared_ptr<TensorOperation> lookup = text::Lookup(vocab, "<unk>");
+  EXPECT_EQ(lookup, nullptr);
+}
+
+TEST_F(MindDataTestPipeline, TestVocabLookupOpFail2) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabLookupOpFail2.";
+  // Vocab has nothing
+  std::shared_ptr<Vocab> vocab;
+
+  // Create lookup op
+  // Expected failure: vocab is null
+  std::shared_ptr<TensorOperation> lookup = text::Lookup(vocab, "");
+  EXPECT_EQ(lookup, nullptr);
+}
+
+TEST_F(MindDataTestPipeline, TestVocabLookupOpWithEmptyUnknownToken) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabLookupOpWithEmptyUnknownToken.";
+
+  // Create a TextFile dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Create a vocab from map
+  std::unordered_map<std::string, int32_t> dict;
+  dict["Home"] = 3;
+  std::shared_ptr<Vocab> vocab = std::make_shared<Vocab>();
+  Status s = Vocab::BuildFromUnorderedMap(dict, &vocab);
+  EXPECT_EQ(s, Status::OK());
+
+  // Create Lookup operation on ds
+  // Expected failure: "" is not a word of vocab
+  std::shared_ptr<TensorOperation> lookup = text::Lookup(vocab, "");
+  EXPECT_EQ(lookup, nullptr);
+}
+
+TEST_F(MindDataTestPipeline, TestVocabFromDataset) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabFromDataset.";
+
+  // Create a TextFile dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Create vocab from dataset
+  std::shared_ptr<Vocab> vocab = ds->BuildVocab({"text"}, {0, std::numeric_limits<int64_t>::max()},
+                                                std::numeric_limits<int64_t>::max(), {"<pad>", "<unk>"}, true);
+  EXPECT_NE(vocab, nullptr);
+
+  // Check if vocab has words or not
+  int32_t home_index = vocab->Lookup("home");
+  EXPECT_EQ(home_index, 4);
+
+  // Create Lookup operation on ds
+  std::shared_ptr<TensorOperation> lookup = text::Lookup(vocab, "<unk>");
+  EXPECT_NE(lookup, nullptr);
+
+  // Create Map operation on ds
+  ds = ds->Map({lookup}, {"text"});
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, std::shared_ptr<Tensor>> row;
+  iter->GetNextRow(&row);
+
+  uint64_t i = 0;
+  std::vector<int32_t> expected = {4, 5, 3, 6, 7, 2};
+  while (row.size() != 0) {
+    auto ind = row["text"];
+    MS_LOG(INFO) << ind->shape() << " " << *ind;
+    std::shared_ptr<Tensor> expected_item;
+    Tensor::CreateScalar(expected[i], &expected_item);
+    EXPECT_EQ(*ind, *expected_item);
+    iter->GetNextRow(&row);
+    i++;
+  }
+}
+
+TEST_F(MindDataTestPipeline, TestVocabFromDatasetDefault) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabFromDatasetDefault.";
+
+  // Create a TextFile dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Create vocab from dataset
+  std::shared_ptr<Vocab> vocab = ds->BuildVocab();
+  EXPECT_NE(vocab, nullptr);
+
+  // Check if vocab has words or not
+  int32_t home_index = vocab->Lookup("home");
+  EXPECT_EQ(home_index, 2);
+
+  // Create Lookup operation on ds
+  std::shared_ptr<TensorOperation> lookup = text::Lookup(vocab, "home");
+  EXPECT_NE(lookup, nullptr);
+
+  // Create Map operation on ds
+  ds = ds->Map({lookup});
+  EXPECT_NE(ds, nullptr);
+
+  // Create an iterator over the result of the above dataset
+  // This will trigger the creation of the Execution Tree and launch it.
+  std::shared_ptr<Iterator> iter = ds->CreateIterator();
+  EXPECT_NE(iter, nullptr);
+
+  // Iterate the dataset and get each row
+  std::unordered_map<std::string, std::shared_ptr<Tensor>> row;
+  iter->GetNextRow(&row);
+
+  uint64_t i = 0;
+  std::vector<int32_t> expected = {2, 3, 1, 4, 5, 0};
+  while (row.size() != 0) {
+    auto ind = row["text"];
+    MS_LOG(INFO) << ind->shape() << " " << *ind;
+    std::shared_ptr<Tensor> expected_item;
+    Tensor::CreateScalar(expected[i], &expected_item);
+    EXPECT_EQ(*ind, *expected_item);
+    iter->GetNextRow(&row);
+    i++;
+  }
+}
+
+TEST_F(MindDataTestPipeline, TestVocabFromDatasetFail1) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabFromDatasetFail1.";
+
+  // Create a TextFile dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Create vocab from dataset
+  // Expected failure: top_k can not be negative
+  std::shared_ptr<Vocab> vocab = ds->BuildVocab({"text"}, {0, std::numeric_limits<int64_t>::max()},
+                                                -2, {"<pad>", "<unk>"}, true);
+  EXPECT_EQ(vocab, nullptr);
+}
+
+TEST_F(MindDataTestPipeline, TestVocabFromDatasetFail2) {
+  MS_LOG(INFO) << "Doing MindDataTestPipeline-TestVocabFromDatasetFail2.";
+
+  // Create a TextFile dataset
+  std::string data_file = datasets_root_path_ + "/testVocab/words.txt";
+  std::shared_ptr<Dataset> ds = TextFile({data_file}, 0, ShuffleMode::kFalse);
+  EXPECT_NE(ds, nullptr);
+
+  // Create vocab from dataset
+  // Expected failure: requency_range [a,b] should be 0 <= a <= b
+  std::shared_ptr<Vocab> vocab = ds->BuildVocab({"text"}, {4, 1},
+                                                std::numeric_limits<int64_t>::max(), {"<pad>", "<unk>"}, true);
+  EXPECT_EQ(vocab, nullptr);
+}