[REVIEW] Adding decimal32 and decimal64 support to parquet reading

CHANGELOG.md

@@ -29,6 +29,7 @@
 - PR #6652 Add support for struct columns in concatenate
 - PR #6675 Add DecimalDtype to cuDF
 - PR #6739 Add Java bindings for is_timestamp
+- PR #6808 Add support for reading decimal32 and decimal64 from parquet
 - PR #6811 First class support for unbounded window function bounds
 - PR #6768 Add support for scatter() on list columns
 - PR #6796 Add create_metadata_file in dask_cudf

cpp/include/cudf/fixed_point/fixed_point.hpp

@@ -515,6 +515,19 @@ class fixed_point {
     Rep const value = detail::shift<Rep, Rad>(_value, scale_type{scale - _scale});
     return fixed_point<Rep, Rad>{scaled_integer<Rep>{value, scale}};
   }
+
+  /**
+   * @brief Returns a string representation of the fixed_point value.
+   */
+  std::string print() const
+  {
+    int const n          = std::pow(10, -_scale);
+    int const f          = _value % n;
+    auto const num_zeros = std::max(0, (-_scale - static_cast<int32_t>(std::to_string(f).size())));
+    auto const zeros     = num_zeros <= 0 ? std::string("") : std::string(num_zeros, '0');
+    return std::to_string(_value / n) + std::string(".") + zeros +
+           std::to_string(std::abs(_value) % n);
+  }
 };  // namespace numeric
 
 /** @brief Function that converts Rep to `std::string`

cpp/include/cudf/io/parquet.hpp

@@ -64,6 +64,10 @@ class parquet_reader_options {
   // Cast timestamp columns to a specific type
   data_type _timestamp_type{type_id::EMPTY};
 
+  // force decimal reading to error if resorting to
+  // doubles for storage of types unsupported by cudf
+  bool _strict_decimal_types = false;
+
   /**
    * @brief Constructor from source info.
    *
@@ -130,6 +134,12 @@ class parquet_reader_options {
    */
   data_type get_timestamp_type() const { return _timestamp_type; }
 
+  /**
+   * @brief Returns true if strict decimal types is set, which errors if reading
+   * a decimal type that is unsupported.
+   */
+  bool is_enabled_strict_decimal_types() const { return _strict_decimal_types; }
+
   /**
    * @brief Sets names of the columns to be read.
    *
@@ -199,6 +209,14 @@ class parquet_reader_options {
    * @param type The timestamp data_type to which all timestamp columns need to be cast.
    */
   void set_timestamp_type(data_type type) { _timestamp_type = type; }
+
+  /**
+   * @brief Enables/disables strict decimal type checking.
+   *
+   * @param val If true, cudf will error if reading a decimal type that is unsupported. If false,
+   * cudf will convert unsupported types to double.
+   */
+  void set_strict_decimal_types(bool val) { _strict_decimal_types = val; }
 };
 
 class parquet_reader_options_builder {
@@ -303,6 +321,18 @@ class parquet_reader_options_builder {
     return *this;
   }
 
+  /**
+   * @brief Sets to enable/disable error with unsupported decimal types.
+   *
+   * @param val Boolean value whether to error with unsupported decimal types.
+   * @return this for chaining.
+   */
+  parquet_reader_options_builder& use_strict_decimal_types(bool val)
+  {
+    options._strict_decimal_types = val;
+    return *this;
+  }
+
   /**
    * @brief move parquet_reader_options member once it's built.
    */

cpp/include/cudf_test/column_wrapper.hpp

@@ -527,6 +527,53 @@ class fixed_point_column_wrapper : public detail::column_wrapper {
     : fixed_point_column_wrapper(std::cbegin(values), std::cend(values), scale)
   {
   }
+
+  /**
+   * @brief Construct a nullable column of the fixed-point elements from a range.
+   *
+   * Constructs a nullable column of the fixed-point elements in the range `[begin,end)` using the range 
+   * `[v, v + distance(begin,end))` interpreted as Booleans to indicate the validity of each element.
+   *
+   * If `v[i] == true`, element `i` is valid, else it is null.
+   *
+   * Example:
+   * @code{.cpp}
+   * // Creates a nullable column of DECIMAL32 elements with 5 elements: {null, 100, null, 300,
+   * null}
+   * auto elements = make_counting_transform_iterator(0, [](auto i){ return i; });
+   * auto validity = make_counting_transform_iterator(0, [](auto i){ return i%2; });
+   * fixed_point_column_wrapper<int32_t> w(elements, elements + 5, validity, 2);
+   * @endcode
+   *
+   * Note: similar to `std::vector`, this "range" constructor should be used
+   *       with parentheses `()` and not braces `{}`. The latter should only
+   *       be used for the `initializer_list` constructors
+   *
+   * @param begin The beginning of the sequence of elements
+   * @param end The end of the sequence of elements
+   * @param v The beginning of the sequence of validity indicators
+   */
+  template <typename FixedPointRepIterator, typename ValidityIterator>
+  fixed_point_column_wrapper(FixedPointRepIterator begin,
+                             FixedPointRepIterator end,
+                             ValidityIterator v,
+                             numeric::scale_type scale)
+    : column_wrapper{}
+  {
+    CUDF_EXPECTS(numeric::is_supported_representation_type<Rep>(), "not valid representation type");
+
+    auto const size         = cudf::distance(begin, end);
+    auto const elements     = thrust::host_vector<Rep>(begin, end);
+    auto const is_decimal32 = std::is_same<Rep, int32_t>::value;
+    auto const id           = is_decimal32 ? type_id::DECIMAL32 : type_id::DECIMAL64;
+    auto const data_type    = cudf::data_type{id, static_cast<int32_t>(scale)};
+
+    wrapped.reset(new cudf::column{data_type,
+                                   size,
+                                   rmm::device_buffer{elements.data(), size * sizeof(Rep)},
+                                   detail::make_null_mask(v, v + size),
+                                   cudf::UNKNOWN_NULL_COUNT});
+  }
 };
 
 /**

cpp/src/io/parquet/page_data.cu

@@ -745,10 +745,10 @@ static const __device__ __constant__ double kPow10[40] = {
  * @param[in] dst Pointer to row output data
  * @param[in] dtype Stored data type
  */
-inline __device__ void gpuOutputDecimal(volatile page_state_s *s,
-                                        int src_pos,
-                                        double *dst,
-                                        int dtype)
+inline __device__ void gpuOutputDecimalAsFloat(volatile page_state_s *s,
+                                               int src_pos,
+                                               double *dst,
+                                               int dtype)
 {
   const uint8_t *dict;
   uint32_t dict_pos, dict_size = s->dict_size, dtype_len_in;
@@ -983,15 +983,16 @@ static __device__ bool setupLocalPageInfo(page_state_s *const s,
           break;
       }
       // Special check for downconversions
-      s->dtype_len_in = s->dtype_len;
-      if (s->col.converted_type == DECIMAL) {
-        s->dtype_len = 8;  // Convert DECIMAL to 64-bit float
-      } else if ((s->col.data_type & 7) == INT32) {
+      s->dtype_len_in          = s->dtype_len;
+      uint16_t const data_type = s->col.data_type & 7;
+      if (s->col.converted_type == DECIMAL && data_type != INT32 && data_type != INT64) {
+        s->dtype_len = 8;  // FLOAT output
+      } else if (data_type == INT32) {
         if (dtype_len_out == 1) s->dtype_len = 1;  // INT8 output
         if (dtype_len_out == 2) s->dtype_len = 2;  // INT16 output
-      } else if ((s->col.data_type & 7) == BYTE_ARRAY && dtype_len_out == 4) {
+      } else if (data_type == BYTE_ARRAY && dtype_len_out == 4) {
         s->dtype_len = 4;  // HASH32 output
-      } else if ((s->col.data_type & 7) == INT96) {
+      } else if (data_type == INT96) {
         s->dtype_len = 8;  // Convert to 64-bit timestamp
       }
 
@@ -1685,9 +1686,13 @@ extern "C" __global__ void __launch_bounds__(block_size)
           gpuOutputString(s, src_pos, dst);
         else if (dtype == BOOLEAN)
           gpuOutputBoolean(s, src_pos, static_cast<uint8_t *>(dst));
-        else if (s->col.converted_type == DECIMAL)
-          gpuOutputDecimal(s, src_pos, static_cast<double *>(dst), dtype);
-        else if (dtype == INT96)
+        else if (s->col.converted_type == DECIMAL) {
+          switch (dtype) {
+            case INT32: gpuOutputFast(s, src_pos, static_cast<uint32_t *>(dst)); break;
+            case INT64: gpuOutputFast(s, src_pos, static_cast<uint2 *>(dst)); break;
+            default: gpuOutputDecimalAsFloat(s, src_pos, static_cast<double *>(dst), dtype); break;
+          }
+        } else if (dtype == INT96)
           gpuOutputInt96Timestamp(s, src_pos, static_cast<int64_t *>(dst));
         else if (dtype_len == 8) {
           if (s->ts_scale)

cpp/src/io/parquet/reader_impl.cu

@@ -98,7 +98,8 @@ parquet::ConvertedType logical_type_to_converted_type(parquet::LogicalType const
  */
 type_id to_type_id(SchemaElement const &schema,
                    bool strings_to_categorical,
-                   type_id timestamp_type_id)
+                   type_id timestamp_type_id,
+                   bool strict_decimal_types)
 {
   parquet::Type physical                = schema.type;
   parquet::ConvertedType converted_type = schema.converted_type;
@@ -132,7 +133,12 @@ type_id to_type_id(SchemaElement const &schema,
       return (timestamp_type_id != type_id::EMPTY) ? timestamp_type_id
                                                    : type_id::TIMESTAMP_MILLISECONDS;
     case parquet::DECIMAL:
-      if (decimal_scale != 0 || (physical != parquet::INT32 && physical != parquet::INT64)) {
+      if (physical == parquet::INT32)
+        return type_id::DECIMAL32;
+      else if (physical == parquet::INT64)
+        return type_id::DECIMAL64;
+      else {
+        CUDF_EXPECTS(strict_decimal_types == false, "Unsupported decimal type read!");
         return type_id::FLOAT64;
       }
       break;
@@ -215,8 +221,9 @@ std::tuple<int32_t, int32_t, int8_t> conversion_info(type_id column_type_id,
   }
 
   int8_t converted_type = converted;
-  if (converted_type == parquet::DECIMAL && column_type_id != type_id::FLOAT64) {
-    converted_type = parquet::UNKNOWN;  // Not converting to float64
+  if (converted_type == parquet::DECIMAL && column_type_id != type_id::FLOAT64 &&
+      column_type_id != type_id::DECIMAL32 && column_type_id != type_id::DECIMAL64) {
+    converted_type = parquet::UNKNOWN;  // Not converting to float64 or decimal
   }
   return std::make_tuple(type_width, clock_rate, converted_type);
 }
@@ -443,7 +450,7 @@ class aggregate_metadata {
         R"(\])"                           // Match closing square brackets
       };
       std::smatch sm;
-      if (std::regex_search(it->second, sm, index_columns_expr)) { return std::move(sm[1].str()); }
+      if (std::regex_search(it->second, sm, index_columns_expr)) { return sm[1].str(); }
     }
     return "";
   }
@@ -463,7 +470,7 @@ class aggregate_metadata {
         if (sm.size() == 2) {  // 2 = whole match, first item
           if (std::find(names.begin(), names.end(), sm[1].str()) == names.end()) {
             std::regex esc_quote{R"(\\")"};
-            names.emplace_back(std::move(std::regex_replace(sm[1].str(), esc_quote, R"(")")));
+            names.emplace_back(std::regex_replace(sm[1].str(), esc_quote, R"(")"));
           }
         }
         str = sm.suffix();
@@ -549,14 +556,16 @@ class aggregate_metadata {
    * reproduce the linear list of output columns that correspond to an input column.
    * @param[in] strings_to_categorical Type conversion parameter
    * @param[in] timestamp_type_id Type conversion parameter
+   * @param[in] strict_decimal_types True if it is an error to load an unsupported decimal type
    *
    */
   void build_column_info(int &schema_idx,
                          std::vector<input_column_info> &input_columns,
                          std::vector<column_buffer> &output_columns,
                          std::deque<int> &nesting,
                          bool strings_to_categorical,
-                         type_id timestamp_type_id) const
+                         type_id timestamp_type_id,
+                         bool strict_decimal_types) const
   {
     int start_schema_idx = schema_idx;
     auto const &schema   = get_schema(schema_idx);
@@ -571,16 +580,19 @@ class aggregate_metadata {
                         output_columns,
                         nesting,
                         strings_to_categorical,
-                        timestamp_type_id);
+                        timestamp_type_id,
+                        strict_decimal_types);
       return;
     }
 
     // if we're at the root, this is a new output column
-    int index = (int)output_columns.size();
     nesting.push_back(static_cast<int>(output_columns.size()));
-    output_columns.emplace_back(
-      data_type{to_type_id(schema, strings_to_categorical, timestamp_type_id)},
-      schema.repetition_type == OPTIONAL ? true : false);
+    auto const col_type =
+      to_type_id(schema, strings_to_categorical, timestamp_type_id, strict_decimal_types);
+    auto const dtype = col_type == type_id::DECIMAL32 || col_type == type_id::DECIMAL64
+                         ? data_type{col_type, numeric::scale_type{-schema.decimal_scale}}
+                         : data_type{col_type};
+    output_columns.emplace_back(dtype, schema.repetition_type == OPTIONAL ? true : false);
     column_buffer &output_col = output_columns.back();
     output_col.name           = schema.name;
 
@@ -591,7 +603,8 @@ class aggregate_metadata {
                         output_col.children,
                         nesting,
                         strings_to_categorical,
-                        timestamp_type_id);
+                        timestamp_type_id,
+                        strict_decimal_types);
     }
 
     // if I have no children, we're at a leaf and I'm an input column (that is, one with actual
@@ -619,7 +632,8 @@ class aggregate_metadata {
   auto select_columns(std::vector<std::string> const &use_names,
                       bool include_index,
                       bool strings_to_categorical,
-                      type_id timestamp_type_id) const
+                      type_id timestamp_type_id,
+                      bool strict_decimal_types) const
   {
     auto const &pfm = per_file_metadata[0];
 
@@ -670,7 +684,8 @@ class aggregate_metadata {
                         output_columns,
                         nesting,
                         strings_to_categorical,
-                        timestamp_type_id);
+                        timestamp_type_id,
+                        strict_decimal_types);
     }
 
     return std::make_tuple(
@@ -1343,6 +1358,8 @@ reader::impl::impl(std::vector<std::unique_ptr<datasource>> &&sources,
     _timestamp_type = options.get_timestamp_type();
   }
 
+  _strict_decimal_types = options.is_enabled_strict_decimal_types();
+
   // Strings may be returned as either string or categorical columns
   _strings_to_categorical = options.is_enabled_convert_strings_to_categories();
 
@@ -1351,7 +1368,8 @@ reader::impl::impl(std::vector<std::unique_ptr<datasource>> &&sources,
     _metadata->select_columns(options.get_columns(),
                               options.is_enabled_use_pandas_metadata(),
                               _strings_to_categorical,
-                              _timestamp_type.id());
+                              _timestamp_type.id(),
+                              _strict_decimal_types);
 }
 
 table_with_metadata reader::impl::read(size_type skip_rows,
@@ -1418,12 +1436,12 @@ table_with_metadata reader::impl::read(size_type skip_rows,
         int32_t clock_rate;
         int8_t converted_type;
 
-        std::tie(type_width, clock_rate, converted_type) =
-          conversion_info(to_type_id(schema, _strings_to_categorical, _timestamp_type.id()),
-                          _timestamp_type.id(),
-                          schema.type,
-                          schema.converted_type,
-                          schema.type_length);
+        std::tie(type_width, clock_rate, converted_type) = conversion_info(
+          to_type_id(schema, _strings_to_categorical, _timestamp_type.id(), _strict_decimal_types),
+          _timestamp_type.id(),
+          schema.type,
+          schema.converted_type,
+          schema.type_length);
 
         column_chunk_offsets[chunks.size()] =
           (col_meta.dictionary_page_offset != 0)

cpp/src/io/parquet/reader_impl.hpp

@@ -210,6 +210,7 @@ class reader::impl {
 
   bool _strings_to_categorical = false;
   data_type _timestamp_type{type_id::EMPTY};
+  bool _strict_decimal_types = false;
 };
 
 }  // namespace parquet