rok commented on code in PR #14341:
URL: https://github.com/apache/arrow/pull/14341#discussion_r1146208807
##########
cpp/src/parquet/encoding.cc:
##########
@@ -3033,11 +3062,226 @@ class RleBooleanDecoder : public DecoderImpl, virtual
public BooleanDecoder {
// ----------------------------------------------------------------------
// DELTA_BYTE_ARRAY
-class DeltaByteArrayDecoder : public DecoderImpl,
- virtual public TypedDecoder<ByteArrayType> {
+// This is also known as incremental encoding or front compression: for each
element in a
+// sequence of strings, store the prefix length of the previous entry plus the
suffix.
+//
+// This is stored as a sequence of delta-encoded prefix lengths
(DELTA_BINARY_PACKED),
+// followed by the suffixes encoded as delta length byte arrays
(DELTA_LENGTH_BYTE_ARRAY).
+
+// ----------------------------------------------------------------------
+// DeltaByteArrayEncoder
+
+template <typename DType>
+class DeltaByteArrayEncoder : public EncoderImpl, virtual public
TypedEncoder<DType> {
public:
- explicit DeltaByteArrayDecoder(const ColumnDescriptor* descr,
+ using T = typename DType::c_type;
+
+ explicit DeltaByteArrayEncoder(const ColumnDescriptor* descr,
MemoryPool* pool =
::arrow::default_memory_pool())
+ : EncoderImpl(descr, Encoding::DELTA_BYTE_ARRAY, pool),
+ sink_(pool),
+ prefix_length_encoder_(nullptr, pool),
+ suffix_encoder_(nullptr, pool) {}
+
+ std::shared_ptr<Buffer> FlushValues() override;
+
+ int64_t EstimatedDataEncodedSize() override {
+ return prefix_length_encoder_.EstimatedDataEncodedSize() +
+ suffix_encoder_.EstimatedDataEncodedSize();
+ }
+
+ using TypedEncoder<DType>::Put;
+
+ void Put(const ::arrow::Array& values) override;
+
+ void Put(const T* buffer, int num_values) override;
+
+ void PutSpaced(const T* src, int num_values, const uint8_t* valid_bits,
+ int64_t valid_bits_offset) override {
+ if (valid_bits != NULLPTR) {
+ PARQUET_ASSIGN_OR_THROW(auto buffer, ::arrow::AllocateBuffer(num_values
* sizeof(T),
+
this->memory_pool()));
+ T* data = reinterpret_cast<T*>(buffer->mutable_data());
+ int num_valid_values = ::arrow::util::internal::SpacedCompress<T>(
+ src, num_values, valid_bits, valid_bits_offset, data);
+ Put(data, num_valid_values);
+ } else {
+ Put(src, num_values);
+ }
+ }
+
+ protected:
+ void PutFixedLenByteArray(const ::arrow::FixedSizeBinaryArray& array) {
+ const uint32_t byte_width = array.byte_width();
+ uint32_t previous_len = byte_width;
+
+
PARQUET_THROW_NOT_OK(::arrow::VisitArraySpanInline<::arrow::FixedSizeBinaryType>(
+ *array.data(),
+ [&](::std::string_view view) {
+ const ByteArray src{view};
+ if (ARROW_PREDICT_TRUE(last_value_.empty())) {
+ last_value_ = view;
+ suffix_encoder_.Put(&src, 1);
+ prefix_length_encoder_.Put({static_cast<int32_t>(0)}, 1);
+ previous_len = byte_width;
+ } else {
+ uint32_t j = 0;
+ while (j < previous_len) {
+ if (last_value_[j] != view[j]) {
+ break;
+ }
+ j++;
+ }
+ previous_len = j;
+ prefix_length_encoder_.Put({static_cast<int32_t>(j)}, 1);
+
+ const uint8_t* suffix_ptr = src.ptr + j;
+ const uint32_t suffix_length = static_cast<uint32_t>(byte_width -
j);
+ last_value_ =
+ string_view{reinterpret_cast<const char*>(suffix_ptr),
suffix_length};
+ const ByteArray suffix(suffix_length, suffix_ptr);
+ suffix_encoder_.Put(&suffix, 1);
+ }
+ return Status::OK();
+ },
+ []() { return Status::OK(); }));
+ }
+
+ template <typename ArrayType>
+ void PutBinaryArray(const ArrayType& array) {
+ uint32_t previous_len = 0;
+
+ PARQUET_THROW_NOT_OK(::arrow::VisitArraySpanInline<typename
ArrayType::TypeClass>(
+ *array.data(),
+ [&](::std::string_view view) {
+ if (ARROW_PREDICT_TRUE(view.size() > kMaxByteArraySize)) {
Review Comment:
Switched to `>=`. Can FLBA actually be too big?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
