Dave Kleinschmidt

I'm planning to have a go at this this week, since #127 has been merged.

There's a lot of whitespace changes here; can you revert those?

Agreed about this not blocking a release. For syntax, like we discussed on slack maybe you'd do `sin($(lag(x)))` to get the "unprotected" `LagTerm` interpretation and `:(x + y)` to get...

I want to add that this choice was intentional, so it's not quite right to say that this happens _incorrectly_. It's just the current implementation of arbitrary functions blocks all...

How about storing the contrasts for the response along with the contrasts for predictor variables in the `ModelFrame`?

The way you've defined it here `modelcols` is just returning the modelcols for the wrapped term. You need something like ``` StatsModels.modelcols(t::FDTerm, d::NamedTuple) = Δ(t.term, d) ``` Plus something like...

But that doesn't generate the actual numerical columns. I guess you could do ``` Δ(obj::CategoricalTerm, d::ColumnTable) = obj.contrasts[getproperty(d, obj.name)[2:end]] ``` which is basically what `modelcols` does for a categorical term,...

Without concrete code to look at I'm not sure what to say; can you share somehow? Maybe DM on slack if it's private code still?

Just as a way of organizing thoughts about how to approach this. There are (at least) two approaches to dealing with rows. 1. `modlecols` generates each column independently, possibly generating...

I'd say this is right on the edge of package development level of complexity :) The underlying issue why this isn't easily doable with a function is that the "non-special"...