Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.

Prefect Community

hi prefect-community! is there any way to force a(n over)write of a checkpointed/cached task result? say i have an S3Result, can i call write on it to overwrite the previously checkpointed result (something like `s3result.write(overwrite=True)` )?

Hey <@U02C9KUNQ4W>, are you using targets? Shouldn’t S3 overwrite by default with <https://stackoverflow.com/a/58489223/11163214|the same name> ?

hey, i'm using an S3Result <https://docs.prefect.io/api/latest/engine/results.html#s3result>

or rather, my task is returning an S3Result

Could you give me a code example so I can understand better?

```def long_running_task(df, params):
    # do some work with df and params


with Flow("test_flow", result=S3Result(bucket="bucket_name")) as flow:

    task = task(
        target='./cache/result.csv',
        checkpoint=True,
        max_retries=1,
        retry_delay=pd.Timedelta(seconds=3),
    )

    s3_result = task(long_running_task)(df=df, params=params)

    # is this possible?
    s3_result.write(overwrite=True)```

<@U01QEJ9PP53> something like that - the idea being that i want to overwrite something that was previously checkpointed

I understand. The issue here is that you set the `target` to a file. `targets`  are a form of caching in Prefect where if the file already exists, the task won’t run. What you want to do is instead use
`task(…,result=S3Result(..,location=…))`
the location will achieve the same thing but it not a caching mechanism so the task will still re-run

If you use `target`, the result will just be loaded from there instead of running the task

And then of course, just change your task to return the df instead of the result

```def long_running_task(df, params):
    # do some work with df and params
    return modified_df.  # instead of the S3Result```

Also, check out the <https://docs.prefect.io/api/latest/engine/results.html|result docs> because we have a pandas serializer you can use if that df is pandas

hm, my task already returns a df - however, when i inspect the type of the returned object (`s3_result` above) in the `with` block, it's an S3Result

Ah ok I think that is fine because Prefect will resolve it for you if you do:
```with Flow(...) as flow:
    df = first_task(df)
    df = second_task(df)```
Prefect handles resolving those results and passing them to the next task even if the type might be `S3Result`

yep, that's exactly what i'm doing :sweat_smile:

Ok I misunderstood the code snippet. I thought you were creating S3Result inside the task and returning it. You’re all good. My advice should still stand to remove target. Just let me know if you still have issues :slightly_smiling_face: