Move to Mir Algorithm and new features

[9il]

Would you mind to move mir-pybuffer to mir-algorithm?

Feature request:

• scalar D types (numbers, chars, strings) to PyObject wrap (so, no need use ctypes at python side for numeric args). The idea that wrapper accept PyObject on D side and decide if it can be converted to a scalar type.
• exceptions instead of errors on python side (if possible), plus no writeln for errors.

[ShigekiKarita]

OK. I will try them. I do not think that exception handling is difficult if it is only for arg type conversion error. It can be done as follows 1. return structure with invalid arg 1-origin index and pybuffer error code from @pybuffer generated D function 2. raise TypeError with useful info and message like "invalid argment at 5th: {errcode}" instead of writeln from pybuffer.CDLL class I think there has been much better exception handling approach between python and C++. Maybe I will look inside pytorch, numpy and some famous C++ based python lib. 2018年9月11日(火) 0:00 Ilya <notifications@github.com>:

…

Would you mind to move mir-pybuffer to mir-algorithm? Feature request: - scalar D types (numbers, chars, strings) to PyObject wrap (so, no need use ctypes at python side for numeric args). The idea that wrapper accept PyObject on D side and decide if it can be converted to a scalar type. - exceptions instead of errors on python side (if possible), plus no writeln for errors. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGbs7pEReFd0A2w3r1yx7AaLkeBuyMFBks5uZn6rgaJpZM4WhlZ7> .

[ShigekiKarita]

Scalar support and moving mir-algorithm are no problem. I appreciate it

[ShigekiKarita]

@9il I'm reworking this by using Python C API and BP in new repository https://github.com/ShigekiKarita/mir-pybind

[9il]

Nice, ping me when it ready for review