add init, del, repr, doc for pyType

[YesDrX]

Hi @yglukhov , please pull the following merge request.

implement init, del, doc for exported types

Example

fix the CI/CD pipeline

Logs

made some minor changes on the readme file

[yglukhov]

Thanks for the PR!

Destroy hooks should already work with regular nim machinery, i.e. =destroy() hooks. If they don't it should be fixed, but no additional conventions should be invented.

Also please use 2 spaces indentation wherever you're using 4.

[YesDrX]

Thanks for the PR!

Destroy hooks should already work with regular nim machinery, i.e. =destroy() hooks. If they don't it should be fixed, but no additional conventions should be invented.

Also please use 2 spaces indentation wherever you're using 4.

I saw that you bind tp_free, tp_dealloc here

nimpy/nimpy.nim

Lines 154 to 160 in 0645af5

	ty.tp_basicsize = td.origSize.cint - sizeof(pointer).cint
	ty.tp_flags = defaultTPFLAGS()
	ty.tp_doc = td.doc
	ty.tp_new = td.newFunc
	ty.tp_free = freeNimObj
	ty.tp_dealloc = destructNimObj
	ty.tp_descr_get = typDescrGet

However, tp_finalize is a separate function (called before tp_dealloc) (del in python). If someone has some extra memory management need be done before destroy the object. tp_finalize is the function to call.

[yglukhov]

I don't understand what's so valuable in the finalize callback vs nim destructors that justifies it, despite the fact that
a. it's not guaranteed to work in all python environments (only newer python versions)
b. it will never be called from the nim side if the object never leaves nim boundary.
c. object can leave python runtime, triggering finalize, but then continue living perfectly "valid" on the nim side, but with finalizer already called
All of the scenarios above are just an awful mess. What's the advantage? What's the "extra memory management" you can only do in finalize but not in destructor?

[YesDrX]

I don't understand what's so valuable in the finalize callback vs nim destructors that justifies it, despite the fact that a. it's not guaranteed to work in all python environments (only newer python versions) b. it will never be called from the nim side if the object never leaves nim boundary. c. object can leave python runtime, triggering finalize, but then continue living perfectly "valid" on the nim side, but with finalizer already called All of the scenarios above are just an awful mess. What's the advantage? What's the "extra memory management" you can only do in finalize but not in destructor?

Ok. __del__ is now removed.

[yglukhov]

Thank you!