PARQUET-1402: [C++] Parquet files with dictionary page offset as 0 is not readable

…adable

pyarrow needs to handle dictionary page offset = 0 as a special case to be compatible with java parquet reader.

Author: shyam <shyam@dremio.com>

Closes #4359 from shyambits2004/5322 and squashes the following commits:

f47762a <shyam> Parquet files with dictionary page offset as 0 is not readable

Author: shyam

cpp/src/parquet/arrow/arrow-reader-writer-test.cc

@@ -2354,6 +2354,12 @@ TEST(TestArrowReaderAdHoc, CorruptedSchema) {
   TryReadDataFile(path, ::arrow::StatusCode::IOError);
 }
 
+TEST(TestArrowReaderAdHoc, HandleDictPageOffsetZero) {
+  // PARQUET-1402: parquet-mr writes files this way which tripped up
+  // some business logic
+  TryReadDataFile(test::get_data_file("dict-page-offset-zero.parquet"));
+}
+
 class TestArrowReaderAdHocSparkAndHvr
     : public ::testing::TestWithParam<
           std::tuple<std::string, std::shared_ptr<::DataType>>> {};

cpp/src/parquet/file_reader.cc

@@ -94,7 +94,8 @@ class SerializedRowGroup : public RowGroupReader::Contents {
     auto col = row_group_metadata_->ColumnChunk(i);
 
     int64_t col_start = col->data_page_offset();
-    if (col->has_dictionary_page() && col_start > col->dictionary_page_offset()) {
+    if (col->has_dictionary_page() && col->dictionary_page_offset() > 0 &&
+        col_start > col->dictionary_page_offset()) {
       col_start = col->dictionary_page_offset();
     }