Description

sensokame

opened on Sep 19, 2023

Issue body actions

Environment

• Pythonnet version: Version: 3.0.1
• Python version: 3.10
• Operating System: Any (Windows/Linux)
• .NET Runtime: 6.0.0

Details

• Trying to substract System Decimal from Python Decimal (the other way works fine)

  TODO

• import both system decimal and python decimal

• create a separate variable from each type

• subtract System Decimal from Python Decimal

    from System import Decimal as SysDecimal
    from decimal import Decimal as PyDecimal
    sys_dec = SysDecimal(1)
    py_dec = PyDecimal(1)
    print(str(sys_dec - py_dec )) # works fine
    print(str(py_dec - sys_dec)) # crashes

• error thrown
  Traceback (most recent call last):
  File "", line 1, in
  RecursionError: maximum recursion depth exceeded while calling a Python object

Activity

1. dariooo512 commented on Sep 19, 2023

   

   dariooo512

   on Sep 19, 2023

   More actions

   I have the same issue:

   same setup, here is my case:

   var offers = new [] { new {Discount = 2.0m}};
   Runtime.PythonDLL = @"C:\Python311\python311.dll";
   PythonEngine.Initialize();
   PythonEngine.BeginAllowThreads();
   using (Py.GIL())
   {
       File.WriteAllText("script.py", @"
   def execute(offers):
       total_discount = 0
       for offer in offers:
           total_discount += offer.Discount
       average_discount = total_discount / len(offers)
       return average_discount
   ");
       var script = Py.Import("script");
       var result = script.InvokeMethod("execute", offers.ToPython());
   }
   

   maximum recursion depth exceeded while calling a Python object
   File [...], line 4, in execute
   total_discount += offer.Discount

   hope it helps
2. filmor commented on Sep 23, 2023

   

   filmor

   on Sep 23, 2023

   Member

   More actions

   I'll have a look into the recursion (which I can reproduce and is a bug), but we don't actually by default provide conversions from/to Python decimals. You must be running a non-upstream version of Python.NET or have more code that you are not showing.
3. filmor

   added

   bug

   on Oct 16, 2023
4. gertdreyer commented on Feb 21, 2024

   

   gertdreyer

   on Feb 21, 2024

   Contributor

   More actions

   I have the same issue.

   Pythonnet version: 3.0.3
   Pythonnet version: 3.11
   Dotnet 6.0: Windows and Linux

   For a codec I am using a naive string based conversion of cs System.Decimal to py decimal.Decimal and the inverse, in our use case the possible loss of precision should not be a concern.

   I have traced it down to the __rsub__ operation on the cs type.

   Operation	Status
   cs - py	Working
   cs.__sub__(py)	Working
   py - cs	RecursionError
   py.__sub__(cs)	NotImplemented
   cs.__rsub__(py)	RecursionError

   The same happens with multiply and addition right-hand methods as well. As per the python rules for operators the forward operators are tried first and if they are NotImplemented for a type the right hand ones are tried where the RecursionError occurs. The codec conversion does not get called for the python decimal foward operators which results in the python decimal conversion code being called which throws the NotImplemented exception

   Codec:

   public class DecimalCodec : IPyObjectDecoder, IPyObjectEncoder
       {
           public bool CanDecode(PyType objectType, Type targetType)
           {
               return objectType.Name == "decimal.Decimal" && targetType == typeof(decimal);
           }
   
           public bool CanEncode(Type type)
           {
               return typeof(decimal) == type;
           }
   
           public bool TryDecode<T>(PyObject pyObj, out T? value)
           {
               if (typeof(T) != typeof(decimal))
               {
                   value = default(T);
                   return false;
               }
               try
               {
                   string decstr = (pyObj as dynamic).__str__();
                   bool parseSuccess = decimal.TryParse(decstr, out decimal csdeci);
                   value = (T)Convert.ChangeType(csdeci, typeof(decimal));
                   return parseSuccess;
               }
               catch (Exception ex)
               {
                   value = default(T);
                   return false;
               }
           }
   
           public PyObject? TryEncode(object value)
           {
               if (typeof(decimal) != value.GetType())
               {
                   return null;
               };
               try
               {
                   dynamic pydeci = PyModule.Import("decimal");
                   dynamic thedecimal = pydeci.Decimal(value.ToString());
                   return thedecimal;
               }
               catch (Exception ex)
               {
                   return null;
               }
           }
       }

   Reacted by Stuart Barclay
5. lostmsu

   closed this as completedin #2327on Feb 26, 2024

Issue actions