for the cache_key_fn parameter, what is the difference between using

always

vs

lambda: "always"

@task(cache_key_fn=lambda: "always", cache_expiration=timedelta(seconds=20))
def my_task():
      return "something"

This is more of a python question. But I hope this helps.

x="always"
y=lambda : "always"
print(x)
print(y)
print(y())

Running this gives the output.

always
<function <lambda> at 0x7f324e6c7d90>
always

thank you for taking the time to respond. Curious as to why input lambda, a callable function as a parameter rather than a string.

You can write your own functions that generates name for the task run. something like https://github.com/PrefectHQ/prefect/blob/d75c2fb1d57a594195d054f109d3c0d8dd44e81d/src/prefect/tasks.py#L59

Ah — So, task cache key is just a string and in this case, it is set to “always” — is this correct?

I guess so. Setting it to always is bad 😄

Here’s what I found on

cache_key_fn

and why I agree with you:

An optional callable that, given the task run context and call parameters, generates a string key; if the key matches a previous completed state, that state result will be restored instead of running the task again.

The link https://github.com/PrefectHQ/prefect/blob/d75c2fb1d57a594195d054f109d3c0d8dd44e81d/src/prefect/tasks.py#L59 is the correct one

Based on what I have seen I have not came across a state named always — it is kind of funny though

its not a state

right. it is a string

Nope. If the cache key is always "always", when ever ur function is called , it will always return the result returned by the function exec when it was first called.

even if I call the function subsequently, it will continue to return the result when it was first called?

yes

sorry if you already covered this — is there an advantage of using a lambda function over just a string?

Cached tasks when completed registers the cache key to the server along with the storage path. Since ur task always has the same cache key "always" , it will always be a cache hit.

I am still searching… It is in the Prefect documentation

Ur answer to the lamda over string is here

An optional callable that, given the task run context and call parameters, generates a string key; if the key matches a previous completed state, that state result will be restored instead of running the task again.

thank you @Deceivious 👍