diff --git a/include/ss/extract.hpp b/include/ss/extract.hpp
index b10759d..60d3fdb 100644
--- a/include/ss/extract.hpp
+++ b/include/ss/extract.hpp
@@ -2,6 +2,7 @@
 
 #include "type_traits.hpp"
 #include <cstring>
+#include <fast_float/fast_float.h>
 #include <functional>
 #include <limits>
 #include <optional>
@@ -14,104 +15,17 @@ namespace ss {
 ////////////////
 // number converters
 ////////////////
-template <typename T>
-std::enable_if_t<std::is_floating_point_v<T>, T> pow10(int n) {
-    T ret = 1.0;
-    T r = 10.0;
-    if (n < 0) {
-        n = -n;
-        r = 0.1;
-    }
 
-    while (n) {
-        if (n & 1) {
-            ret *= r;
-        }
-        r *= r;
-        n >>= 1;
-    }
-    return ret;
-}
-
-// TODO not working with large number of digits
 template <typename T>
 std::enable_if_t<std::is_floating_point_v<T>, std::optional<T>> to_num(
     const char* begin, const char* const end) {
-    if (begin == end) {
+    T ret;
+    auto answer = fast_float::from_chars(begin, end, ret);
+
+    if (answer.ec != std::errc() || answer.ptr != end) {
         return std::nullopt;
     }
-    int sign = 1;
-    T int_part = 0.0;
-    T frac_part = 0.0;
-    bool has_frac = false;
-    bool has_exp = false;
-
-    // +/- sign
-    if (*begin == '-') {
-        ++begin;
-        sign = -1;
-    }
-
-    while (begin != end) {
-        if (*begin >= '0' && *begin <= '9') {
-            int_part = int_part * 10 + (*begin - '0');
-        } else if (*begin == '.') {
-            has_frac = true;
-            ++begin;
-            break;
-        } else if (*begin == 'e') {
-            has_exp = true;
-            ++begin;
-            break;
-        } else {
-            return std::nullopt;
-        }
-        ++begin;
-    }
-
-    if (has_frac) {
-        T frac_exp = 0.1;
-
-        while (begin != end) {
-            if (*begin >= '0' && *begin <= '9') {
-                frac_part += frac_exp * (*begin - '0');
-                frac_exp *= 0.1;
-            } else if (*begin == 'e') {
-                has_exp = true;
-                ++begin;
-                break;
-            } else {
-                return std::nullopt;
-            }
-            ++begin;
-        }
-    }
-
-    // parsing exponent part
-    T exp_part = 1.0;
-    if (begin != end && has_exp) {
-        int exp_sign = 1;
-        if (*begin == '-') {
-            exp_sign = -1;
-            ++begin;
-        } else if (*begin == '+') {
-            ++begin;
-        }
-
-        int e = 0;
-        while (begin != end && *begin >= '0' && *begin <= '9') {
-            e = e * 10 + *begin - '0';
-            ++begin;
-        }
-
-        exp_part = pow10<T>(exp_sign * e);
-    }
-
-    if (begin != end) {
-        return std::nullopt;
-    }
-
-    return sign * (int_part + frac_part) * exp_part;
+    return ret;
 }
 
 inline std::optional<short> from_char(char c) {
@@ -249,7 +163,7 @@ bool shift_and_add_overflow(T& value, T digit, F add_last_digit_owerflow) {
 }
 #else
 
-#warning "use clang or gcc!!!"
+#warning "Use clang or gcc if possible."
 template <typename T, typename U>
 bool shift_and_add_overflow(T& value, T digit, U is_negative) {
     digit = (is_negative) ? -digit : digit;
diff --git a/meson.build b/meson.build
index 16fff8f..4ff7542 100644
--- a/meson.build
+++ b/meson.build
@@ -4,5 +4,12 @@ project('ssp', 'cpp',
     'cpp_std=c++17', 
     'buildtype=debugoptimized'])
 
-includes = include_directories('include')
+fast_float_sub = subproject('fast_float')
+fast_float_dep = fast_float_sub.get_variable('fast_float_dep')
+
+ssp_dep = declare_dependency(
+  include_directories: include_directories('include'),
+  dependencies: fast_float_dep
+  )
+
 subdir('test')
diff --git a/subprojects/fast_float.wrap b/subprojects/fast_float.wrap
new file mode 100644
index 0000000..db07118
--- /dev/null
+++ b/subprojects/fast_float.wrap
@@ -0,0 +1,3 @@
+[wrap-git]
+url = https://github.com/red0124/fast_float
+revision = meson
diff --git a/test/meson.build b/test/meson.build
index 9936129..a8596eb 100644
--- a/test/meson.build
+++ b/test/meson.build
@@ -9,11 +9,11 @@ test_sources = files([
 doctest_proj = subproject('doctest')
 doctest_dep = doctest_proj.get_variable('doctest_dep')
 
-test_exe = executable('test_ssp', 
-  sources: test_sources, 
-  dependencies: doctest_dep, 
-  include_directories: includes,
+test_exe = executable(
+  'test_ssp',
+  sources: test_sources,
+  dependencies: [doctest_dep, ssp_dep],
   cpp_args: '-lstdc++fs'
   )
 
-test('tests_ssp', test_exe)
+test('test_ssp', test_exe)
diff --git a/test/test_extractions.cpp b/test/test_extractions.cpp
index c0866e9..244b368 100644
--- a/test/test_extractions.cpp
+++ b/test/test_extractions.cpp
@@ -2,11 +2,9 @@
 #include <algorithm>
 #include <ss/extract.hpp>
 
-constexpr auto eps = 0.000001;
-using ld = long double;
-
 #define CHECK_FLOATING_CONVERSION(input, type)                                 \
     {                                                                          \
+        auto eps = std::numeric_limits<type>::min();                           \
         std::string s = #input;                                                \
         auto t = ss::to_num<type>(s.c_str(), s.c_str() + s.size());            \
         REQUIRE(t.has_value());                                                \
@@ -14,6 +12,7 @@ using ld = long double;
     }                                                                          \
     {                                                                          \
         /* check negative too */                                               \
+        auto eps = std::numeric_limits<type>::min();                           \
         auto s = std::string("-") + #input;                                    \
         auto t = ss::to_num<type>(s.c_str(), s.c_str() + s.size());            \
         REQUIRE(t.has_value());                                                \
@@ -23,23 +22,18 @@ using ld = long double;
 TEST_CASE("testing extract functions for floating point values") {
     CHECK_FLOATING_CONVERSION(123.456, float);
     CHECK_FLOATING_CONVERSION(123.456, double);
-    CHECK_FLOATING_CONVERSION(123.456, ld);
 
     CHECK_FLOATING_CONVERSION(69, float);
     CHECK_FLOATING_CONVERSION(69, double);
-    CHECK_FLOATING_CONVERSION(69, ld);
 
     CHECK_FLOATING_CONVERSION(420., float);
     CHECK_FLOATING_CONVERSION(420., double);
-    CHECK_FLOATING_CONVERSION(420., ld);
 
     CHECK_FLOATING_CONVERSION(0.123, float);
     CHECK_FLOATING_CONVERSION(0.123, double);
-    CHECK_FLOATING_CONVERSION(0.123, ld);
 
     CHECK_FLOATING_CONVERSION(123e4, float);
     CHECK_FLOATING_CONVERSION(123e4, double);
-    CHECK_FLOATING_CONVERSION(123e4, ld);
 }
 
 #define CHECK_DECIMAL_CONVERSION(input, type)                                  \