ENG-3152: (#141) DocDB filtering on minor compactions

Summary:
Allow DocDB filtering on minor compactions, and add tests for minor compactions specifically.

Also, instead of treating expired values as tombstones at expiration hybrid time, with this change we'll be treating them as tombstones at the original hybrid time. This removes the need for special handling of collection init markers, because expiration of a collection init marker will no longer hide its sub-keys that have not expired or don't have a TTL.

Test Plan: Jenkins

Reviewers: sergei, pritam.damania, rahuldesirazu

Reviewed By: pritam.damania

Subscribers: kannan, ybase

Differential Revision: https://phabricator.dev.yugabyte.com/D4550

src/yb/common/hybrid_time.h

@@ -164,6 +164,8 @@ class HybridTime {
       default:
         return false;
     }
+    LOG(FATAL) << "Should never happen";
+    return false;  // Never reached.
   }
 
   bool operator <(const HybridTime& other) const {
@@ -242,6 +244,16 @@ inline std::ostream &operator <<(std::ostream &o, const HybridTime &hybridTime)
   return o << hybridTime.ToString();
 }
 
+namespace hybrid_time_literals {
+
+inline HybridTime operator "" _usec_ht(unsigned long long microseconds) { // NOLINT
+  return HybridTime::FromMicros(microseconds);
+}
+
+} // namespace hybrid_time_literals
+
+using hybrid_time_literals::operator"" _usec_ht;
+
 }  // namespace yb
 
 #endif  // YB_COMMON_HYBRID_TIME_H_

src/yb/docdb/doc_operation-test.cc

@@ -548,7 +548,7 @@ void DocOperationRangeFilterTest::TestWithSortingType(ColumnSchema::SortingType
                 { row.k, row.r, row.v },
                 1000,
                 t);
-    ASSERT_OK(FlushRocksDB());
+    ASSERT_OK(FlushRocksDbAndWait());
   }
   std::vector<rocksdb::LiveFileMetaData> live_files;
   rocksdb()->GetLiveFilesMetaData(&live_files);
@@ -651,7 +651,7 @@ SubDocKey(DocKey(0x0000, [1], []), [ColumnId(2); HT{ physical: 1000 w: 2 }]) ->
 SubDocKey(DocKey(0x0000, [1], []), [ColumnId(3); HT{ physical: 1000 w: 3 }]) -> 3; ttl: 1.000s
       )#");
 
-  CompactHistoryBefore(t1);
+  FullyCompactHistoryBefore(t1);
 
   // Verify all entries are purged.
   AssertDocDbDebugDumpStrEq(R"#(
@@ -697,7 +697,7 @@ SubDocKey(DocKey(0x0000, [1], []), [ColumnId(3); HT{ physical: 1000 logical: 1 w
 SubDocKey(DocKey(0x0000, [1], []), [ColumnId(3); HT{ physical: 1000 w: 3 }]) -> 3; ttl: 1.000s
       )#");
 
-  CompactHistoryBefore(t1);
+  FullyCompactHistoryBefore(t1);
 
   // Verify the rest of the columns still live.
   AssertDocDbDebugDumpStrEq(R"#(
@@ -746,7 +746,7 @@ size_t GenerateFiles(int total_batches, DocOperationTest* test) {
       test->WriteQLRow(
           QLWriteRequestPB_QLStmtType_QL_STMT_INSERT, schema, {j, j, j, j}, 1000000, t0);
     }
-    EXPECT_OK(test->FlushRocksDB());
+    EXPECT_OK(test->FlushRocksDbAndWait());
   }
   return expected_files;
 }