Replace pythonic type for representing a pointer with _weakref.ref or something better named.

[arizvisa]

The pythonic types component could be reworked a bit to clean up the atoms (or types) that are used. Instead of using type for a reference, or chr for a character we can allow for more types or use things that are more clear.

We currently have [chr, length] for an array of characters and essentially str is just an alias for that. However, we can use str with the form (str, widthtype, lengthtype) to specify a particular type of string.

As prior noted, we use type to note a reference (pointer) but it would be more clear if we dual-use the weakref type that you find in _weakref.ref. This could also be used like (_weakref.ref, 4) to specify a uint32_t-sized pointer. We'd also need to expose _weakref.ref by importing it into the global namespace so that users could access it as ref or reference when assigning their types.

It'd be nice to figure out a type that lets you specify an enum used for a particular type. This would be vastly different because the way we're specifying structures is by wrapping them with a python object. There's a list of builtin modules that can be scraped for relelvant types in the sys.builtin_module_names tuple.

[arizvisa]

Support for idaapi.tinfo_t should also be added to the pythonic types. This will be an ongoing thing as I think I'm the only person in the world that's using them at this level at the moment.

[arizvisa]

Issue #156 and PR #157 is semi-related to this. None of this is relevant, but I'm truly considering deprecating pythonic-types to exchange idaapi.tinfo_t with it...but that would fucking suck because tinfo_t is hardly even linked to IDA's database.

[arizvisa]

Commit 6b91323 adds the ability to destructively assign a tinfo_t to a member. I'm not sure how much farther I should go. Maybe I should add support for pythonic-types to the database.type namespace to allow assigning a type non-destructively to an address too.

[arizvisa]

Commit e2d094f adds support for variable-length structures (specified as a tuple of the type and size), and also adds the various string types of different widths and length-prefixes (specified as a tuple of (chr, width, prefix)). This is sort of related to commit 3c2ee19 which enables support for arbitrary string-encodings.

[arizvisa]

Maybe it'd be a better idea to use object for pointers? Just throwing this out there..

[arizvisa]

I've settled on type. Adding support for tinfo_t to Pythonic types doesn't work universally without converting tinfo_t into a Pythonic type. This isn't easy to do since tinfo_t can be empty, unsized, etc. To still allow support for tinfo_t, however, in #158 you're able to assign it to any function that would take a Pythonic-type. This results in different semantics in that assigning a pythonic-type is more of a "suggestion", and assigning a tinfo_t will force IDA to apply it (thus destroying any types after it if the type's size has changed).

Nobody cares about this anyways, so this issue is being closed.