Correct construct nan

Author: dflook

src/python_minifier/__init__.py

@@ -223,7 +223,7 @@ def unparse(module):
     printer = ModulePrinter()
     printer(module)
 
-    print(printer.code)
+    #print(printer.code)
 
     try:
         minified_module = ast.parse(printer.code, 'python_minifier.unparse output')

src/python_minifier/transforms/constant_folding.py

@@ -1,4 +1,5 @@
 import ast
+import math
 
 from python_minifier.expression_printer import ExpressionPrinter
 from python_minifier.transforms.suite_transformer import SuiteTransformer
@@ -32,7 +33,10 @@ def visit_BinOp(self, node):
         except Exception as e:
             return node
 
-        if isinstance(value, str):
+        if math.isnan(value):
+            # There is no nan literal.
+            new_node = ast.Call(func=ast.Name('float', ctx=ast.Load()), args=[ast.Str('nan')], keywords=[])
+        elif isinstance(value, str):
             new_node = ast.Str(s=value)
         elif isinstance(value, bytes):
             new_node = ast.Bytes(s=value)