0.32.1 introduces a breaking change

[gdkrmr]

The constructor for ZScoreTransform changed between v0.32.0 and v0.32.1 and now requires an extra argument.

[Nosferican]

Could we add a deprecation method and release a 0.32.2 to fix this?

[tbeason]

I think my issue is related to this so I'll piggyback here.

x = randn(100)
standardize(ZScoreTransform,x)

fails on v0.32.1.

As a larger and only semi-related point, can we please just make ZScoreTransform be the default? I don't see why it isn't.

[kleinschmidt]

Thanks for reporting, and sorry for releasing a breaking change in a patch release. @Nosferican, AFAICT there ARE deprecation warnings in the fit(ZScoreTransform, ...) which is how most users interact with it. We could also introduce a deprecation in the constructor itself.

It also looks like the changes introduced in #490 use dims=nothing as a default but don't provide a constructor for that value, which is the source of @tbeason error.

Using z score as the default is unrelated, please open another issue to discuss that if there isn't one already.

[kleinschmidt]

This should be fixed once 0.32.2 is registered: JuliaRegistries/General#10500

@tbeason, I was surprised to find that before #490, standardize and friends didn't work with a vector. It now at least provides a more explicit error message (you have to specify dims=1)