From 2f6440a2e93f45c020dae9dffc65653244147a65 Mon Sep 17 00:00:00 2001
From: Wes McKinney <wesm+git@apache.org>
Date: Wed, 12 Jun 2019 15:21:01 -0500
Subject: [PATCH] Remove not-currently-needed enum, add comment about an
 example partition structure

---
 cpp/src/arrow/dataset/dataset.h   |  9 ++-------
 cpp/src/arrow/dataset/partition.h | 30 ++++++++++++++++++++++++++++--
 2 files changed, 30 insertions(+), 9 deletions(-)

diff --git a/cpp/src/arrow/dataset/dataset.h b/cpp/src/arrow/dataset/dataset.h
index ba49b28466dcb..4aba8945b270e 100644
--- a/cpp/src/arrow/dataset/dataset.h
+++ b/cpp/src/arrow/dataset/dataset.h
@@ -18,6 +18,7 @@
 #pragma once
 
 #include <memory>
+#include <string>
 #include <vector>
 
 #include "arrow/dataset/type_fwd.h"
@@ -54,15 +55,9 @@ struct DataSelector {
 /// DataFragments
 class ARROW_DS_EXPORT DataSource {
  public:
-  enum Type {
-    SIMPLE,       // Flat collection
-    PARTITIONED,  // Partitioned collection
-    GENERIC       // All others
-  };
-
   virtual ~DataSource() = default;
 
-  virtual Type type() const = 0;
+  virtual std::string type() const = 0;
 
   virtual std::unique_ptr<DataFragmentIterator> GetFragments(
       const DataSelector& selector) = 0;
diff --git a/cpp/src/arrow/dataset/partition.h b/cpp/src/arrow/dataset/partition.h
index 628480f4457a3..28c55adcc108b 100644
--- a/cpp/src/arrow/dataset/partition.h
+++ b/cpp/src/arrow/dataset/partition.h
@@ -117,9 +117,35 @@ class ARROW_DS_EXPORT HivePartitionScheme : public PartitionScheme {
 // ----------------------------------------------------------------------
 //
 
+// Partitioned datasets come in different forms. Here is an example of
+// a Hive-style partitioned dataset:
+//
+// dataset_root/
+//   key1=$k1_v1/
+//     key2=$k2_v1/
+//       0.parquet
+//       1.parquet
+//       2.parquet
+//       3.parquet
+//     key2=$k2_v2/
+//       0.parquet
+//       1.parquet
+//   key1=$k1_v2/
+//     key2=$k2_v1/
+//       0.parquet
+//       1.parquet
+//     key2=$k2_v2/
+//       0.parquet
+//       1.parquet
+//       2.parquet
+//
+// In this case, the dataset has 11 fragments (11 files) to be
+// scanned, or potentially more if it is configured to split Parquet
+// files at the row group level
+
 class ARROW_DS_EXPORT Partition : public DataSource {
  public:
-  DataSource::Type type() const override;
+  std::string type() const override;
 
   /// \brief The key for this partition source, may be nullptr,
   /// e.g. for the top-level partitioned source container
@@ -129,7 +155,7 @@ class ARROW_DS_EXPORT Partition : public DataSource {
       const Selector& selector) = 0;
 };
 
-/// \brief Container for a dataset partition, which consists of a
+/// \brief Simple implementation of Partition, which consists of a
 /// partition identifier, subpartitions, and some data fragments
 class ARROW_DS_EXPORT SimplePartition : public Partition {
  public: