holoviz
Make it easier to use pn.state.as_cached by supporting unnamed arguments
When trying to cache (via as_cached) a function that you normally use with unnamed arguments its friction you have to spend time on looking up the name of the argument.
Example:
Normally I do pd.read_csv("my_file.csv") and similar. Thus i would expect to be able to do
pn.state.as_cached("data", pd.read_csv, "my_file.csv")
But this raises
TypeError: unsupported operand type(s) for +: 'float' and 'str'
Traceback (most recent call last):
File "C:\repos\private\panel\.venv\lib\site-packages\bokeh\application\handlers\code_runner.py", line 231, in run
exec(self._code, module.__dict__)
File "C:\repos\private\panel\script.py", line 6, in <module>
data = pn.state.as_cached("data", pd.read_csv, "tabulator.csv")
File "c:\repos\private\panel\panel\io\state.py", line 453, in as_cached
new_expiry = time.monotonic() + ttl if ttl else None
TypeError: unsupported operand type(s) for +: 'float' and 'str'
Thus i have to lookup the name of the argument filepath_or_buffer which I never use otherwise.
Please change to the function signature
from
def as_cached(self, key: str, fn: Callable[[], T], ttl: int = None, **kwargs) -> T:
def as_cached(self, key: str, fn: Callable[[], T], *args, ttl: int = None, **kwargs) -> T:
This will probably need a transition period. What you suggest breaks backward compatibility.
Function signature during the transition period could require ttl to be a keyword:
def as_cached(self, key: str, fn: Callable[[], T], *, ttl: int = None, **kwargs) -> T:
I would suggest making the change in connecting with 1.0. There we can fix anything that should have been different.
We really should review the API and use keyword only args where possible.
Yes. This is such a time waster. Everytime I want to use pd.read_csv and as_cached I have to look up that the argument is called filepath_or_buffer. No body writes that. No body remembers that. It just makes the code harder to read.
I was also confused by this, because according to the docs:
"If provided, the args and kwargs will also be hashed making it easy to cache (or memoize) on the arguments to the function:"
def load_data(*args, **kwargs):
return ... # Load some data
data = pn.state.as_cached('data', load_data, *args, **kwargs)
Which seems to imply that you can pass positional arguments or *args and **kwargs directly after the function, but in reality, everything has to be passed as keyword arguments (named parameters).
Also, currently fn requires a Callable[[], T] which is a function that accepts no arg/kwargs. I believe it should be updated to Callable[..., T] (ellipsis literal) to allow any arbitrary args/kwargs to be passed to the given function.
Didn't think this was worth creating a new issue, but let me know if I should.
