I have a basic dataclass
defined with a few fields that looks like this:
from dataclasses import dataclass
from datetime import datetime
@dataclass
class Session:
start_time: datetime
project_name: str
tags: list[str]
end_time: datetime | None = None
# ...
new_session = Session(start_time, "my-project", [])This syntax lets me define the fields that make up arguments to the underlying
__init__ function. This only gets me so far though. What if I want tags to
have a default value of []? What if I want to force those optional fields
(tags and end_time) to be keyword-only arguments?
To take the field definitions of my dataclass a bit further, I need the
field function
which lets me configure more aspects of each field.
from dataclasses import dataclass, field
from datetime import datetime
@dataclass
class Session:
start_time: datetime
project_name: str
tags: list[str] = field(default_factory=list, kw_only=True)
end_time: datetime | None = field(default=None, kw_only=True)
# ...
sesh1 = Session(start1, "my-project", tags=["pytorch", "numpy"])
sesh2 = Session(start2, "other-project", end_time=datetime.now())If I were to say the default value of tags is simply a [], then the
dataclass definition gets evaluated with a reference to that specific instance
of an empty list. Every new Session I create would share and mutate that same
list. For this reason, dataclass detects and warns you away from that.
Instead, I specify list as the default_factory for tags. Each new
Session where tags is not provided will be factory'd a new empty list.
For scalar values, like None, I can use the default parameter instead like
I've done for end_time.
As for marking certain fields as keyword-only, the kw_only parameter can be
overridden with True.
See the docs for field
for all the other configurable options.