WIP: Use monkeytype to apply typehints

[hmaerki]

Adding typehints

I saw this issue #30 and tried to add type hints.

This is how far I got - now it is time to get feedback.

What I did:

pytest tests --monkeytype-output=monkeytype.sqlite3
monkeytype run example_mbus.py

This way I collected runtime data from the test folder and from example_mbus.py running agains real meters.

Then I applied the typehints monkeytype apply meterbus.zzz.

The result is here WIP: Applies types hints using monkeytype.

The easy typehints

About 60% of the typehints are simple and make sense.

The complex typehints

Then I continued manually starting to resolve the Any type hints.
For example:
def _parse_vifx(self) -> Any:
which I manually resolved to:
def _parse_vifx(self) -> Tuple[Union[None, int, float], Union[None, str, MeasureUnit], Union[None, str, VIFUnit, VIFUnitExt, VIFUnitSecExt], Union[None, VIFUnitEnhExt]]:

Another complex structure is:
Before:

lut = {
        # E000 0nnn    Energy Wh (0.001Wh to 10000Wh)
        0x00: (1.0e-3, MeasureUnit.WH, VIFUnit.ENERGY_WH),

and with the typehints resolved:

lut: Dict[int, Tuple[float, Union[str, MeasureUnit], Union[str, VIFUnit, VIFUnitExt, VIFUnitSecExt]]] = {
        # E000 0nnn    Energy Wh (0.001Wh to 10000Wh)
        0x00: (1.0e-3, MeasureUnit.WH, VIFUnit.ENERGY_WH),

These changes are here: hmaerki@d6984ab

How to continue?

• First of all: Why bother? The code works also without type hints and there are many tests.
• The benefit of adding typehints is
  • Internally: That it will be easier to maintain and understand the code.
  • Externally: The library is easier to use.

If typehints are really wanted, I see several work packages:

• A) Add the easy typehints (60% of what monkeytype found).
• B) Add typehints specially for the API (meterbus/init.py, meterbus/serial.py)
• C) Get rid of decimal.Decimal in favor of float.
• D) Rething the types in core_objects. There are many enumerations and there datatypes spread in the whole package. There might be ways to hide all the different datatypes.
• E) def _parse_vifx() returns a tuple of four values (Tuple[Union[None, int, float], Union[None, str, MeasureUnit], Union[None, str, VIFUnit, VIFUnitExt, VIFUnitSecExt], Union[None, VIFUnitEnhExt]]:). There might be a cleaner way to do that.

My proposal would be to go for A) and B).