Correct definition of __slots__ on structs (must be defined in class body)

[mara004]

Currently what struct output looks like with ctypesgen is this

class struct__FPDF_SYSTEMTIME(Structure):
    pass
struct__FPDF_SYSTEMTIME.__slots__ = [
    'wYear',
    'wMonth',
    'wDayOfWeek',
    'wDay',
    'wHour',
    'wMinute',
    'wSecond',
    'wMilliseconds',
]
struct__FPDF_SYSTEMTIME._fields_ = [
    ('wYear', c_ushort),
    ('wMonth', c_ushort),
    ('wDayOfWeek', c_ushort),
    ('wDay', c_ushort),
    ('wHour', c_ushort),
    ('wMinute', c_ushort),
    ('wSecond', c_ushort),
    ('wMilliseconds', c_ushort),
]
FPDF_SYSTEMTIME = struct__FPDF_SYSTEMTIME

We can see that field names are effectively listed twice, which increases file length unnecessarily.
__slots__ can be easily derived from _fields_ using a list comprehension, like this

structname.__slots__ = [n for n, _ in structname._fields_]

Apart from that, what are __slots__ supposed to achieve here? AFAICS, they're never mentioned in the ctypes docs.
And according to my testing, they don't prevent setting non-existent fields.

[nilason]

We can see that field names are effectively listed twice, which increases file length unnecessarily. __slots__ can be easily derived from _fields_ using a list comprehension, like this

structname.__slots__ = [n for n, _ in structname._fields_]

Looks like a good idea to me.

Apart from that, what are __slots__ supposed to achieve here? AFAICS, they're never mentioned in the ctypes docs. And according to my testing, they don't prevent setting non-existent fields, although this myth exists.

It is a Python feat: https://wiki.python.org/moin/UsingSlots

[mara004]

https://wiki.python.org/moin/UsingSlots indeed claims

with the default __dict__, a misspelled variable name results in the creation of a new variable, but with __slots__ it raises in an AttributeError.

This does not seem to hold true with structs, though...

[nilason]

class struct__FPDF_SYSTEMTIME(Structure):

It is a class, isn’t it?

[mara004]

It is a class, isn’t it?

I suppose it is?

>>> import pypdfium2.raw
>>> import inspect
>>> inspect.isclass(pypdfium2.raw.FPDF_SYSTEMTIME)
True

However, structs do seem a bit odd, according to https://stackoverflow.com/q/76887146/15547292 overriding __setattr__ also does not work as expected.

[mara004]

Aah, the slots definition needs to be in the class body, like so

>>> import ctypes
>>> class FPDF_SYSTEMTIME (ctypes.Structure):
...     _fields_ = [("wYear", ctypes.c_ushort)]
...     __slots__ = ["wYear"]
... 
>>> st = FPDF_SYSTEMTIME()
>>> st.w_year = 1
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'FPDF_SYSTEMTIME' object has no attribute 'w_year'
>>> st.wYear = 1

Then it actually prevents bad fields as expected.

[mara004]

See pypdfium2-team@5148408 for corrected fix.