This is a small utility package for resampling Vectors and Tables to new indices, e.g. a new time resolution.
julia> ]
pkg> add https://github.com/daschw/Resample.jlusing Resample
resample(vec::AbstractVector, org_inds, new_inds, method = Mean())
resample(vec::AbstractVector, org_inds, step, method = Mean())
resample(table, index_col, new_inds, methods = Mean())
resample(table, index_col, step, methods = Mean())Currently implemented methods are Mean(), Sum(), First() and None() (to ignore columns in a Table).
Suppose we have some dummy data for energy and power profiles in a 5 minute resolution.
using Dates
times_long = DateTime(2019):Minute(5):DateTime(2019, 1, 31, 23, 55)
times_short = DateTime(2019):Hour(1):DateTime(2019, 1, 31, 23)
power_long = rand(length(times_long))
energy_long = power_long / 12We can resample the data to the times_short resolution.
using Resample
power_short = resample(power_long, times_long, times_short)
energy_short = resample(energy_long, times_long, times_short, Sum())With Plots we can illustrate the results.
using Plots
plot(times_long, power_long, label = "5 min", st = :step)
plot!(times_short, power_short, label = "1 h", st = :step)
plot(times_long, cumsum(energy_long), label = "5 min", st = :step)
plot!(times_short, cumsum(energy_short), label = "1 h", st = :step)
Instead of a vector of new indices a step length can be provided as well.
power_short_step = resample(power_long, org_inds, Hour(1))
power_short_step == power_shortSuppose we have our data in a DataFrame.
using DataFrames
df = DataFrame(Time = times_long, Power = power_long, Energy = energy_long)We can resample df by selecting the index column and specifying the methods for the remaining columns.
df1 = resample(df, :Time, Hour(1), [Mean(), Sum()])If we want to ignore columns of the DataFrame for resampling we can pass the None() method.
df.Label = rand('a':'c', length(times_long))
df2 = resample(df, :Time, Hour(1); Energy = Sum(), Label = None())
df1 == df2