Fix bug #66745: Scientific notation

We document that float to string is problematic for BCMath.

git-svn-id: https://svn.php.net/repository/phpdoc/en/trunk@343020 c90b9560-bf6c-de11-be94-00142212c4b1

reference/bc/book.xml

@@ -13,6 +13,31 @@
     For arbitrary precision mathematics PHP offers the Binary Calculator which
    supports numbers of any size and precision up to 2147483647-1 (or 0x7FFFFFFF-1) decimals, represented as strings.
   </para>
+  <caution>
+   <para>
+    Passing values of type <type>float</type> to a BCMath function which expects
+    a <type>string</type> as operand may not have the desired effect due to the
+    way PHP converts <type>float</type> values to <type>string</type>, namely
+    that the <type>string</type> may be in exponential notation (what is not
+    supported by BCMath), and that the decimal separator is locale dependend
+    (while BCMath always expects a decimal point).
+   </para>
+   <informalexample>
+    <programlisting role="php">
+<![CDATA[
+<?php
+$num1 = 0; // (string) 0 => '0'
+$num2 = -0.000005; // (string) -0.000005 => '-5.05E-6'
+echo bcadd($num1, $num2, 6); // => '0.000000'
+
+setlocale(LC_NUMERIC, 'de_DE'); // uses a decimal comma
+$num2 = 1.2; // (string) 1.2 => '1,2'
+echo bcsub($num1, $num2, 1); // => '0.0'
+?>
+]]>
+    </programlisting>
+   </informalexample>
+  </caution>
  </preface>
  <!-- }}} -->