commit 7ff7bebcba2c95f51403052547922b50828c340e
author Alec Posney <aposney@atlassian.com> Fri Dec 29 17:43:06 2017 +0100
committer Uwe L. Korn <uwe@apache.org> Fri Dec 29 17:43:06 2017 +0100
tree 68be8a039668beee9465241740b6250841626c97
parent e2814ccb224a01289ad3a736ce89a956cb7d52fb

PARQUET-1180: Fix behaviour of num_children element of primitive nodes

Per the parquet.thift spec, for primitive nodes the num_children schema
attibute should remain unset. This is implemeted correctly in parquet-mr
see [1]. However currently parquet-cpp does set the num_children
attribute to 0 if it is a primitive node. This pull requests fixes this
issue and the tests that were relying on this behavior

[1] parquet-mr/parquet-hadoop/src/main/java/org/apache/parquet/format/converter/ParquetMetadataConverter.java

Author: Alec Posney <aposney@atlassian.com>

Closes #427 from Posnet/fix-primite-node-num-children and squashes the following commits:

e5af5c6 [Alec Posney] Fix behavior of num_children element of primitive nodes

src/parquet/schema-test.cc

diff --git a/src/parquet/schema-test.cc b/src/parquet/schema-test.cc
index f806c12..c8cce9f 100644
--- a/src/parquet/schema-test.cc
+++ b/src/parquet/schema-test.cc
@@ -47,7 +47,6 @@
   result.__set_name(name);
   result.__set_repetition_type(repetition);
   result.__set_type(type);
-  result.__set_num_children(0);
 
   return result;
 }

src/parquet/schema.cc

diff --git a/src/parquet/schema.cc b/src/parquet/schema.cc
index fa6116f..6075ab6 100644
--- a/src/parquet/schema.cc
+++ b/src/parquet/schema.cc
@@ -350,7 +350,6 @@
   format::SchemaElement* element = static_cast<format::SchemaElement*>(opaque_element);
 
   element->__set_name(name_);
-  element->__set_num_children(0);
   element->__set_repetition_type(ToThrift(repetition_));
   if (logical_type_ != LogicalType::NONE) {
     element->__set_converted_type(ToThrift(logical_type_));