Release notes
Backward compatibility breaking changes
* Most of time series transformations (i.e. operations that take a time series and transform its values into another single time series) are now introduced in a dual form: as a method and as a function.
The method syntax is the following
x.pct()
and works in-situ (in-place), meaning it transforms the values inside the time series object `x` and __does not return anything back__.
The function syntax is the following
y = pct(x)
and produces a new time series object, here `y`, leaving the original `x` unchanged. Note that the function name (here, `pct`) needs to be prepended the `irispie` package alias, such as `ir.pct`.
The list of time series operations having a dual form is the following
diff
diff_log
pct
roc
cum_diff
cum_diff_log
cum_pct
cum_roc
mov_avg
mov_sum
mov_prod
fill_missing
aggregate
disaggregate
x13
hpf_trend
hpf_gap
(more to come).
NB: The `hpf` filter is now implemented as a function only (not as a method) because it returns __two__ outputs (the trend and the gap).
Other changes and new features
* The `Databank` class is now named `Databox` to remove confusion when dealing with actual databases and databanks. The name `Databank` is still preserved though s an alias for backward compatibility.