--- a/Include/floatobject.h	Tue May 08 14:20:11 2001 +0000
+++ b/Include/floatobject.h	Tue May 08 15:19:57 2001 +0000
@@ -20,12 +20,31 @@
 
 #define PyFloat_Check(op) ((op)->ob_type == &PyFloat_Type)
 
-extern DL_IMPORT(PyObject *) PyFloat_FromString(PyObject*, char**);
+/* Return Python float from string PyObject.  Second argument ignored on
+   input, and, if non-NULL, NULL is stored into *junk (this tried to serve a
+   purpose once but can't be made to work as intended). */
+extern DL_IMPORT(PyObject *) PyFloat_FromString(PyObject*, char** junk);
+
+/* Return Python float from C double. */
 extern DL_IMPORT(PyObject *) PyFloat_FromDouble(double);
+
+/* Extract C double from Python float.  The macro version trades safety for
+   speed. */
 extern DL_IMPORT(double) PyFloat_AsDouble(PyObject *);
+#define PyFloat_AS_DOUBLE(op) (((PyFloatObject *)(op))->ob_fval)
 
-/* Macro, trading safety for speed */
-#define PyFloat_AS_DOUBLE(op) (((PyFloatObject *)(op))->ob_fval)
+/* Write repr(v) into the char buffer argument, followed by null byte.  The
+   buffer must be "big enough"; >= 100 is very safe.
+   PyFloat_AsReprString(buf, x) strives to print enough digits so that
+   PyFloat_FromString(buf) then reproduces x exactly. */
+extern DL_IMPORT(void) PyFloat_AsReprString(char*, PyFloatObject *v);
+
+/* Write str(v) into the char buffer argument, followed by null byte.  The
+   buffer must be "big enough"; >= 100 is very safe.  Note that it's
+   unusual to be able to get back the float you started with from
+   PyFloat_AsString's result -- use PyFloat_AsReprString() if you want to
+   preserve precision across conversions. */
+extern DL_IMPORT(void) PyFloat_AsString(char*, PyFloatObject *v);
 
 #ifdef __cplusplus
 }