oss port

CMakeLists.txt

@@ -247,8 +247,12 @@ if(ENABLE_CUDA)
   else()
     if(${CMAKE_VERSION} VERSION_GREATER_EQUAL "3.18.0")
       message(STATUS "CMake 3.18+, Setting CUDA_ARCHITECTURES.")
-      set(CMAKE_CUDA_ARCHITECTURES
-          35-virtual
+      if(CMAKE_CUDA_COMPILER_VERSION VERSION_LESS 12.0.0)
+        set(CMAKE_CUDA_ARCHITECTURES 35-virtual)
+      else()
+        set(CMAKE_CUDA_ARCHITECTURES "")
+      endif()
+      list(APPEND CMAKE_CUDA_ARCHITECTURES
           50-virtual
           60-virtual
           70-virtual
@@ -258,8 +262,12 @@ if(ENABLE_CUDA)
       message(STATUS "CUDA_ARCHITECTURES: ${CMAKE_CUDA_ARCHITECTURES}")
     else()
       message(STATUS "CMake 3.17 or under, setting CUDA architecture flags manually.")
-      set(CUDA_COMPILATION_ARCH
-        -gencode=arch=compute_35,code=compute_35;
+      if(CMAKE_CUDA_COMPILER_VERSION VERSION_LESS 12.0.0)
+        set(CUDA_COMPILATION_ARCH -gencode=arch=compute_35,code=compute_35;)
+      else()
+        set(CUDA_COMPILATION_ARCH "")
+      endif()
+      list(APPEND CUDA_COMPILATION_ARCH
         -gencode=arch=compute_50,code=compute_50;
         -gencode=arch=compute_60,code=compute_60;
         -gencode=arch=compute_70,code=compute_70;

DataMgr/BufferMgr/BufferMgr.cpp

@@ -620,17 +620,11 @@ void BufferMgr::deleteBuffersWithPrefix(const ChunkKey& key_prefix, const bool)
                            // reserveBuffer which needs segs_mutex_ and then
                            // chunk_index_mutex_
   std::lock_guard<std::mutex> chunk_index_lock(chunk_index_mutex_);
-  auto startChunkIt = chunk_index_.lower_bound(key_prefix);
-  if (startChunkIt == chunk_index_.end()) {
-    return;
-  }
-
-  auto buffer_it = startChunkIt;
-  while (buffer_it != chunk_index_.end() &&
-         std::search(buffer_it->first.begin(),
-                     buffer_it->first.begin() + key_prefix.size(),
-                     key_prefix.begin(),
-                     key_prefix.end()) != buffer_it->first.begin() + key_prefix.size()) {
+  auto prefix_upper_bound = key_prefix;
+  prefix_upper_bound.emplace_back(std::numeric_limits<ChunkKey::value_type>::max());
+  for (auto buffer_it = chunk_index_.lower_bound(key_prefix),
+            end_chunk_it = chunk_index_.upper_bound(prefix_upper_bound);
+       buffer_it != end_chunk_it;) {
     auto seg_it = buffer_it->second;
     if (seg_it->buffer) {
       if (seg_it->buffer->getPinCount() != 0) {

DataMgr/ForeignStorage/CsvFileBufferParser.cpp

@@ -168,6 +168,8 @@ ParseBufferResult CsvFileBufferParser::parseBuffer(ParseBufferRequest& request,
     std::vector<std::unique_ptr<char[]>>
         tmp_buffers;  // holds string w/ removed escape chars, etc
     const char* line_start = p;
+    row_index_plus_one++;
+    bool incorrect_column_count = false;
     p = import_export::delimited_parser::get_row(p,
                                                  thread_buf_end,
                                                  buf_end,
@@ -177,10 +179,6 @@ ParseBufferResult CsvFileBufferParser::parseBuffer(ParseBufferRequest& request,
                                                  tmp_buffers,
                                                  try_single_thread,
                                                  !columns_are_pre_filtered);
-
-    row_index_plus_one++;
-
-    bool incorrect_column_count = false;
     try {
       validate_expected_column_count(row, num_cols, point_cols, file_path);
     } catch (const ForeignStorageException& e) {

ImportExport/DelimitedParserUtils.cpp

@@ -46,13 +46,18 @@ inline void trim_space(const char*& field_begin, const char*& field_end) {
 inline void trim_quotes(const char*& field_begin,
                         const char*& field_end,
                         const import_export::CopyParams& copy_params) {
-  if (copy_params.quoted && field_end - field_begin > 0 &&
-      *field_begin == copy_params.quote) {
-    ++field_begin;
+  auto quote_begin = field_begin, quote_end = field_end;
+  if (copy_params.quoted) {
+    trim_space(quote_begin, quote_end);
   }
-  if (copy_params.quoted && field_end - field_begin > 0 &&
-      *(field_end - 1) == copy_params.quote) {
-    --field_end;
+  if (copy_params.quoted && quote_end - quote_begin > 0) {
+    if (*quote_begin == copy_params.quote && *(quote_end - 1) == copy_params.quote) {
+      field_begin = ++quote_begin;
+      field_end = (quote_begin == quote_end) ? quote_end : --quote_end;
+    } else {
+      throw import_export::delimited_parser::DelimitedParserException(
+          "Unable to trim quotes.");
+    }
   }
 }
 }  // namespace
@@ -123,13 +128,26 @@ size_t find_end(const char* buffer,
   if (last_line_delim_pos <= 0) {
     size_t excerpt_length = std::min<size_t>(50, size);
     std::string buffer_excerpt{buffer, buffer + excerpt_length};
-    std::string error_message =
-        "Unable to find an end of line character after reading " + std::to_string(size) +
-        " characters. Please ensure that the correct \"line_delimiter\" option is "
-        "specified or update the \"buffer_size\" option appropriately. Row number: " +
-        std::to_string(buffer_first_row_index + 1) +
-        ". First few characters in row: " + buffer_excerpt;
-    throw InsufficientBufferSizeException{error_message};
+    if (in_quote) {
+      std::string quote(1, copy_params.quote);
+      std::string error_message =
+          "Unable to find a matching end quote for the quote character '" + quote +
+          "' after reading " + std::to_string(size) +
+          " characters. Please ensure that all data fields are correctly formatted "
+          "or update the \"buffer_size\" option appropriately. Row number: " +
+          std::to_string(buffer_first_row_index + 1) +
+          ". First few characters in row: " + buffer_excerpt;
+      throw InsufficientBufferSizeException{error_message};
+    } else {
+      std::string error_message =
+          "Unable to find an end of line character after reading " +
+          std::to_string(size) +
+          " characters. Please ensure that the correct \"line_delimiter\" option is "
+          "specified or update the \"buffer_size\" option appropriately. Row number: " +
+          std::to_string(buffer_first_row_index + 1) +
+          ". First few characters in row: " + buffer_excerpt;
+      throw InsufficientBufferSizeException{error_message};
+    }
   }
 
   return last_line_delim_pos + 1;
@@ -244,10 +262,10 @@ const char* get_row(const char* buf,
           }
           const char* field_begin = field_buf;
           const char* field_end = field_buf + j;
+          trim_quotes(field_begin, field_end, copy_params);
           if (copy_params.trim_spaces) {
             trim_space(field_begin, field_end);
           }
-          trim_quotes(field_begin, field_end, copy_params);
           row.emplace_back(field_begin, field_end - field_begin);
         }
         field = p + 1;

ImportExport/DelimitedParserUtils.h

@@ -37,6 +37,11 @@ class InsufficientBufferSizeException : public std::runtime_error {
       : std::runtime_error(message) {}
 };
 
+class DelimitedParserException : public std::runtime_error {
+ public:
+  DelimitedParserException(const std::string& message) : std::runtime_error(message) {}
+};
+
 /**
  * @brief Finds the closest possible row beginning in the given buffer.
  *

ImportExport/Importer.cpp

@@ -2020,6 +2020,7 @@ static ImportStatus import_thread_delimited(
       row.clear();
       std::vector<std::unique_ptr<char[]>>
           tmp_buffers;  // holds string w/ removed escape chars, etc
+      row_index_plus_one++;
       if (DEBUG_TIMING) {
         us = measure<std::chrono::microseconds>::execution([&]() {
           p = import_export::delimited_parser::get_row(p,
@@ -2044,7 +2045,6 @@ static ImportStatus import_thread_delimited(
                                                      try_single_thread,
                                                      true);
       }
-      row_index_plus_one++;
       // Each POINT could consume two separate coords instead of a single WKT
       if (row.size() < num_cols || (num_cols + point_cols) < row.size()) {
         thread_import_status.rows_rejected++;