incorporated read twice parser from Jeremie

1 files changed, 47 insertions, 1 deletions

diff --git a/vowpalwabbit/parse_primitives.h b/vowpalwabbit/parse_primitives.h
index 1200f5fe..0ed374eb 100644
--- a/vowpalwabbit/parse_primitives.h
+++ b/vowpalwabbit/parse_primitives.h
@@ -9,6 +9,7 @@ embodied in the content of this file are licensed under the BSD
 
 #include<iostream>
 #include <stdint.h>
+#include <math.h>
 #include "v_array.h"
 #include "io.h"
 #include "example.h"
@@ -100,10 +101,55 @@ inline void print_substring(substring s)
   std::cout.write(s.begin,s.end - s.begin);
 }
 
+// The following function is a home made strtof. The
+// differences are :
+//  - much faster (around 50% but depends on the string to parse)
+//  - less error control, but utilised inside a very strict parser
+//    in charge of error detection.
+inline float parseFloat(char * p, char **end)
+{
+    if (!*p || *p == '?')
+        return 0;
+    int s = 1;
+    while (*p == ' ') p++;
+
+    if (*p == '-') {
+        s = -1; p++;
+    }
+
+    double acc = 0;
+    while (*p >= '0' && *p <= '9')
+        acc = acc * 10 + *p++ - '0';
+
+    int num_dec = 0;
+    if (*p == '.') {
+        p++;
+        while (*p >= '0' && *p <= '9') {
+            acc = acc *10 + (*p++ - '0') ;
+            num_dec++;
+        }
+    }
+    int exp_acc = 0;
+    if(*p == 'e' || *p == 'E'){
+    	p++;
+    	int exp_s = 1;
+    	if (*p == '-') {
+    	        exp_s = -1; p++;
+    	}
+    	while (*p >= '0' && *p <= '9')
+    	        exp_acc = exp_acc * 10 + *p++ - '0';
+    	exp_acc *= exp_s;
+
+    }
+    acc *= pow(10,exp_acc-num_dec);
+    *end = p;
+    return s * acc;
+}
+
 inline float float_of_substring(substring s)
 {
   char* endptr = s.end;
-  float f = strtof(s.begin,&endptr);
+  float f = parseFloat(s.begin,&endptr);
   if (endptr == s.begin && s.begin != s.end)
     {
       std::cout << "error: " << std::string(s.begin, s.end-s.begin).c_str() << " is not a float" << std::endl;