We didn't put columns in ReshapePhase because they're not consistent, from beginning to end of a reshape phase they can be quite different e.g. in a pivot.
However, parsing types is necessary! E.g. in the boston pipeline, the pivot phase uses a DateColumn to parse date values.
We should add incoming column types that are checked and casted at the beginning of a ReshapePhase.
I have a whole new idea instead of ReshapePhase... since it is becoming more and more like phase. Maybe we make it a Phase but with a flag that says "renumber rows" ?
We didn't put columns in ReshapePhase because they're not consistent, from beginning to end of a reshape phase they can be quite different e.g. in a pivot.
However, parsing types is necessary! E.g. in the boston pipeline, the pivot phase uses a DateColumn to parse date values.
We should add incoming column types that are checked and casted at the beginning of a ReshapePhase.