Description:
------------
When a DBF is created with dbase_create() all float fields ('F') have
their precision set to 0, so are in fact (inexact) integers[1][2].
Contrary to numeric fields ('N') there is no way to specify the field
length and precision.

Contrary to what one might assume, float fields are not supposed to
store real floating point numbers, but rather they are identical to
numeric fields with regard to the storage format. The only difference is
that as of dBASE IV (where float fields have been introduced) floats are
internally converted to IEEE floats (probably doubles) and numerics are
internally handled as BCD to offer exact calculations[3].

[1] <http://www.promotic.eu/en/pmdoc/Subsystems/Db/dBase/DataTypes.htm>
[2] <http://www.okstate.edu/sas/v7/sashtml/books/acespc/z0214453.htm>
[3] <http://www.oocities.org/geoff_wass/dBASE/GaryWhite/dBASE/FAQ/qfloat.htm>


Test script:
---------------
<?php

$filename = __DIR__ . DIRECTORY_SEPARATOR . 'test.dbf';

$db = dbase_create($filename, [['FLOAT', 'F', 15, 5]]);
dbase_close($db);

$stream = fopen($filename, 'r');
fseek($stream, 48);
$bytes = fread($stream, 2);
var_dump(unpack('Clength/Cprecision', $bytes));


Expected result:
----------------
array(2) {
  ["length"]=>
  int(15)
  ["precision"]=>
  int(5)
}

Actual result:
--------------
array(2) {
  ["length"]=>
  int(20)
  ["precision"]=>
  int(0)
}