How transformations work
A reviewable list of steps, applied in order, that never modifies your source.
Cleaning in ChartBase is a list of steps, not a one-off edit. Each step is recorded, applied in order, and can be reordered or removed later.
Why steps rather than edits
Because you will get it wrong at least once, and because somebody will eventually ask what you did.
- The source is never modified — steps are applied to a copy.
- Re-importing re-applies the same steps, so a refresh does not undo your cleaning.
- The list is a readable record of every change made to the data.
Applying and reverting
Steps are staged as you build them and applied together. Applying regenerates the queryable copy, which is why charts show new values immediately afterwards. Reverting takes the dataset back to its unmodified import.
Order matters
Steps run top to bottom, and each sees the result of the one before. Filtering out rows before a calculation is not the same as filtering after — if a result looks wrong, the order of the list is the first place to look.
What this is not
It is not a place to express business logic. A step changes the data; a measure expresses a question about it. Cleaning belongs in steps, definitions belong in the model — putting a business rule in a transformation freezes it into the rows and hides it from everyone reading the dashboard.