Native Cached Results

[dmadisetti]

Description

I find myself using this pattern a lot

if not os.path.isfile("sample_selection_test_all.npy"):
    np.save(
        "sample_selection_test_all", data_sample_export(test_data, test_sample)
    )
    np.save(
        "sample_selection_train_all",
        data_sample_export(train_data, train_sample),
    )

exported_train = np.load("sample_selection_train_all")
exported_test = np.load("sample_selection_test_all")

I wonder if saving computations + saving outputs is on the roadmap.
(docs recommend functools memo, but I like to restart notebooks sometime)

Suggested solution

result = mo.saved(x)

execution unwraps result, saves it to file, or loads it.
or after cell execution could unwrap result

for stateless, could hash ref dag as file name. (hash concated cell contents)

---

mo.save(plt.gca()) # Last line

allows cell to have a default output. State ensured by the ref dag hash

Alternative

Tons of ways doing this, just provided a suggestion

Additional context

No response

[akshayka]

It's something Myles and I have talked about, yes! We even considered using the same API name. So yes, we'll likely support something like this.