Add a two-argument variant of ParallelFor() that uses the global CPU thread pool

cpp/src/arrow/python/arrow_to_pandas.cc

@@ -40,7 +40,7 @@
 #include "arrow/util/decimal.h"
 #include "arrow/util/logging.h"
 #include "arrow/util/macros.h"
-#include "arrow/util/thread-pool.h"
+#include "arrow/util/parallel.h"
 #include "arrow/visitor_inline.h"
 
 #include "arrow/compute/api.h"
@@ -1458,16 +1458,7 @@ class DataFrameBlockCreator {
     };
 
     if (use_threads) {
-      arrow::internal::ThreadPool* pool = arrow::internal::GetCpuThreadPool();
-      std::vector<std::future<Status>> futures;
-      for (int i = 0; i < table_->num_columns(); ++i) {
-        futures.push_back(pool->Submit(WriteColumn, i));
-      }
-      auto final_status = Status::OK();
-      for (auto& fut : futures) {
-        final_status &= fut.get();
-      }
-      return final_status;
+      return ParallelFor(table_->num_columns(), WriteColumn);
     } else {
       for (int i = 0; i < table_->num_columns(); ++i) {
         RETURN_NOT_OK(WriteColumn(i));

cpp/src/arrow/util/parallel.h

@@ -24,14 +24,31 @@
 #include <vector>
 
 #include "arrow/status.h"
+#include "arrow/util/thread-pool.h"
 
 namespace arrow {
 
 // A parallelizer that takes a `Status(int)` function and calls it with
-// arguments between 0 and `num_tasks - 1`, using the given number of
-// concurrent threads.
-// In most cases it's more appropriate to use the global CPU thread pool
-// (see arrow/util/thread-pool.h)
+// arguments between 0 and `num_tasks - 1`, on an arbitrary number of threads.
+
+template <class FUNCTION>
+Status ParallelFor(int num_tasks, FUNCTION&& func) {
+  auto pool = internal::GetCpuThreadPool();
+  std::vector<std::future<Status>> futures(num_tasks);
+
+  for (int i = 0; i < num_tasks; ++i) {
+    futures[i] = pool->Submit(func, i);
+  }
+  auto st = Status::OK();
+  for (auto& fut : futures) {
+    st &= fut.get();
+  }
+  return st;
+}
+
+// A variant of ParallelFor() with an explicit number of dedicated threads.
+// In most cases it's more appropriate to use the 2-argument ParallelFor (above),
+// or directly the global CPU thread pool (arrow/util/thread-pool.h).
 
 template <class FUNCTION>
 Status ParallelFor(int nthreads, int num_tasks, FUNCTION&& func) {