GH-47752: [C++][Python][Parquet] Add minimum compression savings for V2 pages

cpp/src/parquet/column_writer.cc

@@ -1057,8 +1057,12 @@ void ColumnWriterImpl::BuildDataPageV2(int64_t definition_levels_rle_size,
   bool page_is_compressed = false;
   if (pager_->has_compressor() && values->size() > 0) {
     pager_->Compress(*values, compressor_temp_buffer_.get());
-    if (compressor_temp_buffer_->size() < values->size()) {
-      page_is_compressed = true;
+    if (const auto minimum = properties_->min_space_savings()) {
+      const double savings =
+          1.0 - static_cast<double>(compressor_temp_buffer_->size()) / values->size();
+      page_is_compressed = savings >= *minimum;
+    } else {
+      page_is_compressed = compressor_temp_buffer_->size() < values->size();
     }
   }
   std::shared_ptr<Buffer> compressed_values =

cpp/src/parquet/column_writer_test.cc

@@ -15,7 +15,9 @@
 // specific language governing permissions and limitations
 // under the License.
 
+#include <cmath>
 #include <memory>
+#include <optional>
 #include <utility>
 #include <vector>
 
@@ -112,7 +114,8 @@ class TestPrimitiveWriter : public PrimitiveTypedTest<TestType> {
       const ParquetVersion::type version = ParquetVersion::PARQUET_1_0,
       const ParquetDataPageVersion data_page_version = ParquetDataPageVersion::V1,
       bool enable_checksum = false, int64_t page_size = kDefaultDataPageSize,
-      int64_t max_rows_per_page = kDefaultMaxRowsPerPage) {
+      int64_t max_rows_per_page = kDefaultMaxRowsPerPage,
+      std::optional<double> min_space_savings = {}) {
     sink_ = CreateOutputStream();
     WriterProperties::Builder wp_builder;
     wp_builder.version(version)->data_page_version(data_page_version);
@@ -130,6 +133,7 @@ class TestPrimitiveWriter : public PrimitiveTypedTest<TestType> {
     wp_builder.max_statistics_size(column_properties.max_statistics_size());
     wp_builder.data_pagesize(page_size);
     wp_builder.max_rows_per_page(max_rows_per_page);
+    wp_builder.min_space_savings(min_space_savings);
     writer_properties_ = wp_builder.build();
 
     metadata_ = ColumnChunkMetaDataBuilder::Make(writer_properties_, this->descr_);
@@ -2084,6 +2088,99 @@ TEST_F(TestValuesWriterInt32Type, AvoidCompressedInDataPageV2) {
     verify_only_one_uncompressed_page(/*total_num_values=*/1);
   }
 }
+
+TEST_F(TestValuesWriterInt32Type, MinSpaceSavingsDataPageV2) {
+  this->SetUpSchema(Repetition::OPTIONAL);
+  this->descr_ = this->schema_.Column(0);
+  this->GenerateData(SMALL_SIZE);
+  std::fill(this->values_.begin(), this->values_.end(), 0);
+
+  // Calculate the actual savings to test the exact threshold.
+  ASSERT_OK_AND_ASSIGN(auto codec, Codec::Create(Compression::ZSTD));
+  const int64_t original_size = SMALL_SIZE * sizeof(int32_t);
+  std::vector<uint8_t> input(original_size, 0);
+  std::vector<uint8_t> compressed(codec->MaxCompressedLen(original_size, input.data()));
+  ASSERT_OK_AND_ASSIGN(auto compressed_size,
+                       codec->Compress(original_size, input.data(), compressed.size(),
+                                       compressed.data()));
+
+  const double savings = 1.0 - static_cast<double>(compressed_size) / original_size;
+  ASSERT_GT(savings, 0.0);
+  ASSERT_LT(savings, 1.0);
+
+  const std::vector<std::pair<std::optional<double>, bool>> cases = {
+      {std::nullopt, true},
+      {0.0, true},
+      {std::nextafter(savings, 0.0), true},
+      {savings, true},
+      {std::nextafter(savings, 1.0), false},
+      {1.0, false},
+  };
+
+  ColumnProperties column_properties;
+  column_properties.set_compression(Compression::ZSTD);
+
+  for (const auto& [minimum, expected_compressed] : cases) {
+    auto writer = this->BuildWriter(
+        SMALL_SIZE, column_properties, ParquetVersion::PARQUET_2_LATEST,
+        ParquetDataPageVersion::V2, false, kDefaultDataPageSize,
+        kDefaultMaxRowsPerPage, minimum);
+    writer->WriteBatch(SMALL_SIZE, this->def_levels_.data(), nullptr,
+                       this->values_ptr_);
+    writer->Close();
+    ASSERT_OK_AND_ASSIGN(auto buffer, this->sink_->Finish());
+
+    auto page_reader = PageReader::Open(
+        std::make_shared<::arrow::io::BufferReader>(buffer), SMALL_SIZE,
+        Compression::ZSTD, default_reader_properties(), *this->descr_);
+    auto page = page_reader->NextPage();
+    ASSERT_NE(page, nullptr);
+    ASSERT_EQ(PageType::DATA_PAGE_V2, page->type());
+    auto data_page = std::static_pointer_cast<DataPageV2>(page);
+    ASSERT_EQ(expected_compressed, data_page->is_compressed());
+    ASSERT_EQ(page_reader->NextPage(), nullptr);
+
+    // Verify that the data still reads correctly.
+    auto read_pages = PageReader::Open(
+        std::make_shared<::arrow::io::BufferReader>(buffer), SMALL_SIZE,
+        Compression::ZSTD, default_reader_properties(), *this->descr_);
+    auto reader = std::static_pointer_cast<Int32Reader>(
+        ColumnReader::Make(this->descr_, std::move(read_pages)));
+    std::vector<int32_t> values(SMALL_SIZE);
+    std::vector<int16_t> definitions(SMALL_SIZE);
+    int64_t values_read = 0;
+    ASSERT_EQ(SMALL_SIZE,
+              reader->ReadBatch(SMALL_SIZE, definitions.data(), nullptr,
+                                values.data(), &values_read));
+    ASSERT_EQ(SMALL_SIZE, values_read);
+    ASSERT_EQ(this->values_, values);
+    ASSERT_EQ(this->def_levels_, definitions);
+  }
+}
+
+TEST_F(TestValuesWriterInt32Type, MinSpaceSavingsIgnoredForDataPageV1) {
+  this->GenerateData(SMALL_SIZE);
+  ColumnProperties column_properties;
+  column_properties.set_compression(Compression::ZSTD);
+  std::shared_ptr<Buffer> default_output;
+
+  for (std::optional<double> minimum :
+       {std::optional<double>{}, std::optional<double>{1.0}}) {
+    auto writer = this->BuildWriter(
+        SMALL_SIZE, column_properties, ParquetVersion::PARQUET_2_LATEST,
+        ParquetDataPageVersion::V1, false, kDefaultDataPageSize,
+        kDefaultMaxRowsPerPage, minimum);
+    writer->WriteBatch(SMALL_SIZE, nullptr, nullptr, this->values_ptr_);
+    writer->Close();
+    ASSERT_OK_AND_ASSIGN(auto buffer, this->sink_->Finish());
+
+    if (!minimum) {
+      default_output = buffer;
+    } else {
+      ASSERT_TRUE(default_output->Equals(*buffer));
+    }
+  }
+}
 #endif
 
 // Test writing and reading geometry columns

cpp/src/parquet/properties.h

@@ -403,6 +403,7 @@ class PARQUET_EXPORT WriterProperties {
           max_rows_per_page_(properties.max_rows_per_page()),
           version_(properties.version()),
           data_page_version_(properties.data_page_version()),
+          min_space_savings_(properties.min_space_savings()),
           created_by_(properties.created_by()),
           store_decimal_as_integer_(properties.store_decimal_as_integer()),
           page_checksum_enabled_(properties.page_checksum_enabled()),
@@ -529,6 +530,18 @@ class PARQUET_EXPORT WriterProperties {
       return this;
     }
 
+    /// Minimum space savings required to compress V2 data page values.
+    ///
+    /// Savings are (1.0 - compressed_size / uncompressed_size), excluding
+    /// repetition and definition levels. For example, 0.1 requires 10% savings.
+    /// If unset, keep compression whenever it makes the values smaller.
+    /// Ignored for V1 pages and when compression is disabled.
+    /// Values outside [0, 1], including NaN, cause build() to throw.
+    Builder* min_space_savings(std::optional<double> min_space_savings) {
+      min_space_savings_ = min_space_savings;
+      return this;
+    }
+
     /// Specify the Parquet file version.
     /// Default PARQUET_2_6.
     Builder* version(ParquetVersion::type version) {
@@ -883,6 +896,10 @@ class PARQUET_EXPORT WriterProperties {
     /// \brief Build the WriterProperties with the builder parameters.
     /// \return The WriterProperties defined by the builder.
     std::shared_ptr<WriterProperties> build() {
+      if (min_space_savings_ &&
+          !(*min_space_savings_ >= 0.0 && *min_space_savings_ <= 1.0)) {
+        throw ParquetException("min_space_savings must be in range [0, 1]");
+      }
       std::unordered_map<std::string, ColumnProperties> column_properties;
       auto get = [&](const std::string& key) -> ColumnProperties& {
         auto it = column_properties.find(key);
@@ -918,7 +935,7 @@ class PARQUET_EXPORT WriterProperties {
           pagesize_, max_rows_per_page_, version_, created_by_, page_checksum_enabled_,
           size_statistics_level_, std::move(file_encryption_properties_),
           default_column_properties_, column_properties, data_page_version_,
-          store_decimal_as_integer_, std::move(sorting_columns_),
+          min_space_savings_, store_decimal_as_integer_, std::move(sorting_columns_),
           content_defined_chunking_enabled_, content_defined_chunking_options_));
     }
 
@@ -933,6 +950,7 @@ class PARQUET_EXPORT WriterProperties {
     int64_t max_rows_per_page_;
     ParquetVersion::type version_;
     ParquetDataPageVersion data_page_version_;
+    std::optional<double> min_space_savings_ = {};
     std::string created_by_;
     bool store_decimal_as_integer_;
     bool page_checksum_enabled_;
@@ -973,6 +991,9 @@ class PARQUET_EXPORT WriterProperties {
     return parquet_data_page_version_;
   }
 
+  /// Minimum space savings required to compress V2 data page values, if set.
+  std::optional<double> min_space_savings() const { return min_space_savings_; }
+
   inline ParquetVersion::type version() const { return parquet_version_; }
 
   inline std::string created_by() const { return parquet_created_by_; }
@@ -1100,7 +1121,8 @@ class PARQUET_EXPORT WriterProperties {
       std::shared_ptr<FileEncryptionProperties> file_encryption_properties,
       const ColumnProperties& default_column_properties,
       const std::unordered_map<std::string, ColumnProperties>& column_properties,
-      ParquetDataPageVersion data_page_version, bool store_short_decimal_as_integer,
+      ParquetDataPageVersion data_page_version, std::optional<double> min_space_savings,
+      bool store_short_decimal_as_integer,
       std::vector<SortingColumn> sorting_columns, bool content_defined_chunking_enabled,
       CdcOptions content_defined_chunking_options)
       : pool_(pool),
@@ -1110,6 +1132,7 @@ class PARQUET_EXPORT WriterProperties {
         pagesize_(pagesize),
         max_rows_per_page_(max_rows_per_page),
         parquet_data_page_version_(data_page_version),
+        min_space_savings_(min_space_savings),
         parquet_version_(version),
         parquet_created_by_(created_by),
         store_decimal_as_integer_(store_short_decimal_as_integer),
@@ -1129,6 +1152,7 @@ class PARQUET_EXPORT WriterProperties {
   int64_t pagesize_;
   int64_t max_rows_per_page_;
   ParquetDataPageVersion parquet_data_page_version_;
+  std::optional<double> min_space_savings_;
   ParquetVersion::type parquet_version_;
   std::string parquet_created_by_;
   bool store_decimal_as_integer_;

cpp/src/parquet/properties_test.cc

@@ -80,6 +80,31 @@ TEST(TestWriterProperties, DefaultCompression) {
             ::arrow::util::kUseDefaultCompressionLevel);
 }
 
+TEST(TestWriterProperties, MinSpaceSavings) {
+  WriterProperties::Builder builder;
+  ASSERT_FALSE(builder.build()->min_space_savings().has_value());
+
+  for (double savings : {0.0, 0.1, 1.0}) {
+    auto properties = builder.min_space_savings(savings)->build();
+    ASSERT_EQ(std::optional<double>(savings), properties->min_space_savings());
+
+    auto copied = WriterProperties::Builder(*properties).build();
+    ASSERT_EQ(properties->min_space_savings(), copied->min_space_savings());
+  }
+
+  auto properties = builder.min_space_savings(std::nullopt)->build();
+  ASSERT_FALSE(properties->min_space_savings().has_value());
+}
+
+TEST(TestWriterProperties, InvalidMinSpaceSavings) {
+  for (double savings : {-0.1, 1.1, std::numeric_limits<double>::infinity(),
+                         -std::numeric_limits<double>::infinity(),
+                         std::numeric_limits<double>::quiet_NaN()}) {
+    EXPECT_THROW(WriterProperties::Builder().min_space_savings(savings)->build(),
+                 ParquetException);
+  }
+}
+
 TEST(TestWriterProperties, AdvancedHandling) {
   WriterProperties::Builder builder;
   builder.compression("gzip", Compression::GZIP);

python/pyarrow/_dataset_parquet.pyx

@@ -655,6 +655,7 @@ cdef class ParquetFileWriteOptions(FileWriteOptions):
             ),
             column_encoding=self._properties["column_encoding"],
             data_page_version=self._properties["data_page_version"],
+            min_space_savings=self._properties["min_space_savings"],
             encryption_properties=self._properties["encryption_properties"],
             write_batch_size=self._properties["write_batch_size"],
             dictionary_pagesize_limit=self._properties["dictionary_pagesize_limit"],
@@ -707,6 +708,7 @@ cdef class ParquetFileWriteOptions(FileWriteOptions):
             use_byte_stream_split=False,
             column_encoding=None,
             data_page_version="1.0",
+            min_space_savings=None,
             use_deprecated_int96_timestamps=False,
             coerce_timestamps=None,
             allow_truncated_timestamps=False,

python/pyarrow/_parquet.pxd

@@ -57,7 +57,8 @@ cdef shared_ptr[WriterProperties] _create_writer_properties(
     sorting_columns=*,
     store_decimal_as_integer=*,
     use_content_defined_chunking=*,
-    bloom_filter_options=*
+    bloom_filter_options=*,
+    min_space_savings=*
 ) except *
 
 

python/pyarrow/_parquet.pyx

@@ -2076,13 +2076,21 @@ cdef shared_ptr[WriterProperties] _create_writer_properties(
         sorting_columns=None,
         store_decimal_as_integer=False,
         use_content_defined_chunking=False,
-        bloom_filter_options=None) except *:
+        bloom_filter_options=None,
+        min_space_savings=None) except *:
 
     """General writer properties"""
     cdef:
         shared_ptr[WriterProperties] properties
         WriterProperties.Builder props
         CdcOptions cdc_options
+        double savings
+
+    if min_space_savings is not None:
+        savings = min_space_savings
+        if not (0.0 <= savings <= 1.0):
+            raise ValueError("min_space_savings must be in range [0, 1]")
+        props.min_space_savings(optional[double](savings))
 
     # data_page_version
 
@@ -2405,7 +2413,8 @@ cdef class ParquetWriter(_Weakrefable):
                   store_decimal_as_integer=False,
                   use_content_defined_chunking=False,
                   write_time_adjusted_to_utc=False,
-                  bloom_filter_options=None):
+                  bloom_filter_options=None,
+                  min_space_savings=None):
         cdef:
             shared_ptr[WriterProperties] properties
             shared_ptr[ArrowWriterProperties] arrow_properties
@@ -2442,7 +2451,8 @@ cdef class ParquetWriter(_Weakrefable):
             sorting_columns=sorting_columns,
             store_decimal_as_integer=store_decimal_as_integer,
             use_content_defined_chunking=use_content_defined_chunking,
-            bloom_filter_options=bloom_filter_options
+            bloom_filter_options=bloom_filter_options,
+            min_space_savings=min_space_savings,
         )
         arrow_properties = _create_arrow_writer_properties(
             use_deprecated_int96_timestamps=use_deprecated_int96_timestamps,

python/pyarrow/includes/libparquet.pxd

@@ -485,6 +485,7 @@ cdef extern from "parquet/api/writer.h" namespace "parquet" nogil:
     cdef cppclass WriterProperties:
         cppclass Builder:
             Builder* data_page_version(ParquetDataPageVersion version)
+            Builder* min_space_savings(optional[double] min_space_savings)
             Builder* version(ParquetVersion version)
             Builder* compression(ParquetCompression codec)
             Builder* compression(const c_string& path,

python/pyarrow/parquet/core.py

@@ -1090,6 +1090,7 @@ def __init__(self, where, schema, filesystem=None,
                  max_rows_per_page=None,
                  bloom_filter_options=None,
                  use_content_defined_chunking=False,
+                 min_space_savings=None,
                  **options):
         if use_deprecated_int96_timestamps is None:
             # Use int96 timestamps for Spark
@@ -1147,6 +1148,7 @@ def __init__(self, where, schema, filesystem=None,
             max_rows_per_page=max_rows_per_page,
             bloom_filter_options=bloom_filter_options,
             use_content_defined_chunking=use_content_defined_chunking,
+            min_space_savings=min_space_savings,
             **options)
         self.is_open = True
 
@@ -2039,6 +2041,7 @@ def write_table(table, where, row_group_size=None, version='2.6',
                 max_rows_per_page=None,
                 bloom_filter_options=None,
                 use_content_defined_chunking=False,
+                min_space_savings=None,
                 **kwargs):
     # Implementor's note: when adding keywords here / updating defaults, also
     # update it in write_to_dataset and _dataset_parquet.pyx ParquetFileWriteOptions
@@ -2074,6 +2077,7 @@ def write_table(table, where, row_group_size=None, version='2.6',
                 max_rows_per_page=max_rows_per_page,
                 bloom_filter_options=bloom_filter_options,
                 use_content_defined_chunking=use_content_defined_chunking,
+                min_space_savings=min_space_savings,
                 **kwargs) as writer:
             writer.write_table(table, row_group_size=row_group_size)
     except Exception: