Merge branch 'branch-0.11' of https://github.com/rapidsai/cudf into 2…

…948_adding_copy_if

CHANGELOG.md

@@ -36,6 +36,7 @@
 - PR #3298 add null replacement iterator for column_device_view
 - PR #3297 Define and implement new groupby API.
 - PR #3396 Update device_atomics with new bool8 and timestamp specializations
+- PR #3411 Java host memory management API
 - PR #3393 Implement df.cov and enable covariance/correlation in dask_cudf
 - PR #3401 Add dask_cudf ORC writer (to_orc)
 - PR #3331 Add copy_if_else
@@ -137,6 +138,7 @@
 - PR #3387 Strings column gather function
 - PR #3389 Move quantiles.hpp + group_quantiles.hpp files to legacy
 - PR #3398 Move reshape.hpp files to legacy
+- PR #3422 Move utilities to legacy
 
 ## Bug Fixes
 
@@ -189,9 +191,11 @@
 - PR #3388 Support getitem with bools when DataFrame has a MultiIndex
 - PR #3408 Fix String and Column (De-)Serialization
 - PR #3372 Fix dask-distributed scatter_by_map bug
+- PR #3419 Fix a bug in parse_into_parts (incomplete input causing walking past the end of string).
 - PR #3413 Fix dask_cudf read_csv file-list bug
-
-
+- PR #3416 Fix memory leak in ColumnVector when pulling strings off the GPU
+- PR #3424 Fix benchmark build by adding libcudacxx to benchmark's CMakeLists.txt
+- PR #3435 Fix diff and shift for empty series
 
 # cuDF 0.10.0 (16 Oct 2019)
 

cpp/CMakeLists.txt

@@ -461,8 +461,8 @@ add_library(cudf
             src/io/utilities/datasource.cpp
             src/io/utilities/legacy/parsing_utils.cu
             src/utilities/legacy/cuda_utils.cu
-            src/utilities/column_utils.cpp
-	    src/copying/gather.cu
+            src/utilities/legacy/column_utils.cpp
+	        src/copying/gather.cu
             src/utilities/legacy/error_utils.cpp
             src/utilities/nvtx/nvtx_utils.cpp
             src/utilities/nvtx/legacy/nvtx_utils.cpp

cpp/benchmarks/CMakeLists.txt

@@ -26,8 +26,9 @@ include_directories("${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES}"
                     "${CMAKE_SOURCE_DIR}"
                     "${CMAKE_SOURCE_DIR}/src"
                     "${CMAKE_SOURCE_DIR}/thirdparty/cub"
-                    "${CMAKE_SOURCE_DIR}/thirdparty/jitify"
                     "${CMAKE_SOURCE_DIR}/thirdparty/dlpack/include"
+                    "${CMAKE_SOURCE_DIR}/thirdparty/jitify"
+                    "${CMAKE_SOURCE_DIR}/thirdparty/libcudacxx/include"
                     "${GTEST_INCLUDE_DIR}"
                     "${GBENCH_INCLUDE_DIR}"
                     "${ARROW_INCLUDE_DIR}"

cpp/benchmarks/join/generate_input_tables.cuh

@@ -28,7 +28,7 @@
 #include <cudf/cudf.h>
 #include <rmm/rmm.h>
 #include <cudf/utilities/error.hpp>
-#include <utilities/device_atomics.cuh>
+#include <cudf/detail/utilities/device_atomics.cuh>
 
 
 __global__ static void init_curand(curandState * state, const int nstates)

cpp/benchmarks/type_dispatcher/type_dispatcher_benchmark.cu

@@ -29,7 +29,7 @@
 #include <table/legacy/device_table.cuh>
 
 #include <random>
-#include <utilities/cuda_utils.hpp>
+#include <utilities/legacy/cuda_utils.hpp>
 #include "../synchronization/synchronization.hpp"
 #include "../fixture/benchmark_fixture.hpp"
 

cpp/custrings/strings/datetime.cu

@@ -215,6 +215,9 @@ struct parse_datetime
             DTFormatItem item = items[idx];
             int slen = (int)item.length;
             //printf("%d:%c=%d\n",(int)fmt.ftype,ch,(int)slen);
+            if( length < slen ){
+                return 1;
+            }
             if(item.item_type==false)
             {
                 // consume fmt.len bytes from datetime
@@ -223,8 +226,6 @@ struct parse_datetime
                 length -= slen;
                 continue;
             }
-            if( length < slen )
-                return 1;
 
             // special logic for each specifier
             switch(item.specifier)

cpp/custrings/tests/test_datetime.cu

@@ -13,22 +13,22 @@ TEST_F(TestTimestamp, ToTimestamp)
 {
     {    
         std::vector<const char*> hstrs{"1974-02-28T01:23:45Z", "2019-07-17T21:34:37Z",
-                                       nullptr, "" };
+                                       nullptr, "", "1974" };
         NVStrings* strs = NVStrings::create_from_array(hstrs.data(),hstrs.size());
         thrust::device_vector<unsigned long> results(hstrs.size(),0);
         strs->timestamp2long("%Y-%m-%dT%H:%M:%SZ", NVStrings::seconds, results.data().get());
-        int expected[] = { 131246625, 1563399277, 0,0 };
+        int expected[] = { 131246625, 1563399277, 0, 0, 0 };
         for( int idx = 0; idx < (int) hstrs.size(); ++idx )
             EXPECT_EQ((int)results[idx],expected[idx]);
         NVStrings::destroy(strs);
     }
 
     {    
-        std::vector<const char*> hstrs{"12.28.1982", "07.17.2019" };
+        std::vector<const char*> hstrs{"12.28.1982", "07.17.2019", "06" };
         NVStrings* strs = NVStrings::create_from_array(hstrs.data(),hstrs.size());
         thrust::device_vector<unsigned long> results(hstrs.size(),0);
         strs->timestamp2long("%m-%d-%Y", NVStrings::days, results.data().get());
-        int expected[] = { 4744, 18094 };
+        int expected[] = { 4744, 18094, 0 };
         for( int idx = 0; idx < (int) hstrs.size(); ++idx )
             EXPECT_EQ((int)results[idx],expected[idx]);
         NVStrings::destroy(strs);

cpp/include/cudf/detail/copy_if_else.cuh

@@ -23,8 +23,6 @@
 #include <cudf/utilities/type_dispatcher.hpp>
 #include <cudf/detail/utilities/cuda.cuh>
 #include <cudf/detail/utilities/integer_utils.hpp>
-#include <utilities/bit_util.cuh>
-#include <utilities/cuda_utils.hpp>
 
 #include <rmm/device_scalar.hpp>
 #include <cub/cub.cuh>