Fix underscore float parsing (#596)

* Fix underscore float parsing

* Add tests

* Update float parsing

---------

Co-authored-by: A. R. Shajii <ars@ars.me>

Author: inumanag

codon/parser/ast/expr.cpp

@@ -2,6 +2,8 @@
 
 #include "expr.h"
 
+#include <algorithm>
+#include <iterator>
 #include <memory>
 #include <string>
 #include <vector>
@@ -149,9 +151,14 @@ FloatExpr::FloatExpr(double floatValue)
   this->floatValue = std::make_unique<double>(floatValue);
 }
 FloatExpr::FloatExpr(const std::string &value, std::string suffix)
-    : Expr(), value(value), suffix(std::move(suffix)) {
+    : Expr(), value(), suffix(std::move(suffix)) {
+  this->value.reserve(value.size());
+  std::copy_if(value.begin(), value.end(), std::back_inserter(this->value),
+               [](char c) { return c != '_'; });
+
   double result;
-  auto r = fast_float::from_chars(value.data(), value.data() + value.size(), result);
+  auto r = fast_float::from_chars(this->value.data(),
+                                  this->value.data() + this->value.size(), result);
   if (r.ec == std::errc() || r.ec == std::errc::result_out_of_range)
     floatValue = std::make_unique<double>(result);
   else

test/parser/simplify_expr.codon

@@ -39,6 +39,8 @@ print 1844674407_3709551999 #! integer '18446744073709551999' cannot fit into 64
 print 5.15 #: 5.15
 print 2e2 #: 200
 print 2.e-2 #: 0.02
+print 1_000.0 #: 1000
+print 1_000e9 #: 1e+12
 
 #%% float_suffix,barebones
 @extend