ARROW-10328: [C++] Vendor fast_float number parsing library

This library is 2x to 3x faster for parsing strings to binary floating-point numbers. Closes #8494 from pitrou/ARROW-10328-fast-float Authored-by: Antoine Pitrou <[email protected]> Signed-off-by: Antoine Pitrou <[email protected]>
apache · Oct 21, 2020 · 16412a1 · 16412a1
1 parent bb4f2a0
commit 16412a1
Show file tree

Hide file tree

Showing 10 changed files with 1,996 additions and 56 deletions.
diff --git a/LICENSE.txt b/LICENSE.txt
@@ -2223,3 +2223,11 @@ exception of some code pulled in from other repositories (such as
 public domain, released using the CC0 1.0 Universal dedication (*).
 
 (*) https://creativecommons.org/publicdomain/zero/1.0/legalcode
+
+--------------------------------------------------------------------------------
+
+The files in cpp/src/arrow/vendored/fast_float/ contain code from
+
+https://github.com/lemire/fast_float
+
+which is made available under the Apache License 2.0.
diff --git a/cpp/src/arrow/util/value_parsing.cc b/cpp/src/arrow/util/value_parsing.cc
@@ -20,70 +20,19 @@
 #include <string>
 #include <utility>
 
-#include "arrow/util/double_conversion.h"
+#include "arrow/vendored/fast_float/fast_float.h"
 
 namespace arrow {
 namespace internal {
 
-namespace {
-
-struct StringToFloatConverterImpl {
-  StringToFloatConverterImpl()
-      : main_converter_(flags_, main_junk_value_, main_junk_value_, "inf", "nan"),
-        fallback_converter_(flags_, fallback_junk_value_, fallback_junk_value_, "inf",
-                            "nan") {}
-
-  // NOTE: This is only supported in double-conversion 3.1+
-  static constexpr int flags_ =
-      util::double_conversion::StringToDoubleConverter::ALLOW_CASE_INSENSIBILITY;
-
-  // Two unlikely values to signal a parsing error
-  static constexpr double main_junk_value_ = 0.7066424364107089;
-  static constexpr double fallback_junk_value_ = 0.40088499148279166;
-
-  util::double_conversion::StringToDoubleConverter main_converter_;
-  util::double_conversion::StringToDoubleConverter fallback_converter_;
-};
-
-static const StringToFloatConverterImpl g_string_to_float;
-
-// Older clang versions need an explicit implementation definition.
-constexpr double StringToFloatConverterImpl::main_junk_value_;
-constexpr double StringToFloatConverterImpl::fallback_junk_value_;
-
-}  // namespace
-
 bool StringToFloat(const char* s, size_t length, float* out) {
-  int processed_length;
-  float v;
-  v = g_string_to_float.main_converter_.StringToFloat(s, static_cast<int>(length),
-                                                      &processed_length);
-  if (ARROW_PREDICT_FALSE(v == static_cast<float>(g_string_to_float.main_junk_value_))) {
-    v = g_string_to_float.fallback_converter_.StringToFloat(s, static_cast<int>(length),
-                                                            &processed_length);
-    if (ARROW_PREDICT_FALSE(v ==
-                            static_cast<float>(g_string_to_float.fallback_junk_value_))) {
-      return false;
-    }
-  }
-  *out = v;
-  return true;
+  const auto res = ::arrow_vendored::fast_float::from_chars(s, s + length, *out);
+  return res.ec == std::errc() && res.ptr == s + length;
 }
 
 bool StringToFloat(const char* s, size_t length, double* out) {
-  int processed_length;
-  double v;
-  v = g_string_to_float.main_converter_.StringToDouble(s, static_cast<int>(length),
-                                                       &processed_length);
-  if (ARROW_PREDICT_FALSE(v == g_string_to_float.main_junk_value_)) {
-    v = g_string_to_float.fallback_converter_.StringToDouble(s, static_cast<int>(length),
-                                                             &processed_length);
-    if (ARROW_PREDICT_FALSE(v == g_string_to_float.fallback_junk_value_)) {
-      return false;
-    }
-  }
-  *out = v;
-  return true;
+  const auto res = ::arrow_vendored::fast_float::from_chars(s, s + length, *out);
+  return res.ec == std::errc() && res.ptr == s + length;
 }
 
 // ----------------------------------------------------------------------

diff --git a/cpp/src/arrow/vendored/fast_float/README.md b/cpp/src/arrow/vendored/fast_float/README.md
@@ -0,0 +1,9 @@
+The files in this directory are vendored from fast_float
+git changeset `dc46ad4c606dc35cb63c947496a18ef8ab1e0f44`.
+
+See https://github.com/lemire/fast_float
+
+Changes:
+- fixed include paths
+- disabled unused `print()` function
+- enclosed in `arrow_vendored` namespace.