Adds retime #529
Adds retime #529
Conversation
|
It works off of #528 |
|
@iblislin What do you think about the interface? I have chosen singleton types instead of Symbols so that a dispatch system can be implemented. |
well, I personally prefer the Symbol, and build with literal type dispatching. e.g. julia> f(::Val{:abc}) = 42
f (generic function with 1 method)
julia> f(x::Symbol) = f(Val(x))
f (generic function with 2 methods)
julia> f(:abc)
42but I'm not sure about the performance penalty. Maybe you could give that a try. |
|
@iblislin I added a documentation page as well. |
|
Could you sync (rebase or just merge) this branch with master? |
minor return iterator and adds test starts to add retime changes _split changes LTS overload Base.split changes interface of _split to return timestamps
adds missing deps another missing dep support symbols as well handle missing data adds docs on retime minor
ValentinKaisermayer
force-pushed
the
vk-retime
branch
from
095736e to
48d72ab
Compare
Done |
|
@iblislin any thoughts? |
|
Any idea about the Documenter build failures? Tons of |
Not really. |
|
Could you check that is |
src/retime.jl
Outdated
| _toExtrapolationMethod(x::ExtrapolationMethod) = x | ||
|
|
||
| function retime(ta, new_dt::Dates.Period; kwargs...) | ||
| new_timestamps = timestamp(ta)[1]:new_dt:timestamp(ta)[end] |
There was a problem hiding this comment.
What is the expected result in Matlab for this case?
julia> ta = TimeArray([DateTime(2025, 1, 1, 8, 0), DateTime(2025, 1, 2, 2, 0), DateTime(2025, 1, 3, 9, 0)], [1,1, 42])
3×1 TimeArray{Int64, 1, DateTime, Vector{Int64}} 2025-01-01T08:00:00 to 2025-01-03T09:00:00
┌─────────────────────┬────┐
│ │ A │
├─────────────────────┼────┤
│ 2025-01-01T08:00:00 │ 1 │
│ 2025-01-02T02:00:00 │ 1 │
│ 2025-01-03T09:00:00 │ 42 │
└─────────────────────┴────┘
julia> retime(ta, Day(1); downsample=:sum)
...There was a problem hiding this comment.
>> ta = timetable([datetime(2025, 1, 1, 8, 0, 0); datetime(2025, 1, 2, 2, 0, 0); datetime(2025, 1, 3, 9, 0, 0)], ...
[1; 1; 42], ...
'VariableNames', {'Values'})
ta =
Time Values
____________________ ______
01-Jan-2025 08:00:00 1
02-Jan-2025 02:00:00 1
03-Jan-2025 09:00:00 42
>> ta_daily = retime(ta, 'daily', 'sum')
ta_daily =
Time Values
____________________ ______
01-Jan-2025 00:00:00 1
02-Jan-2025 00:00:00 1
03-Jan-2025 00:00:00 42 There was a problem hiding this comment.
julia> ta = TimeArray([DateTime(2025, 1, 1, 8, 0), DateTime(2025, 1, 2, 2, 0), DateTime(2025, 1, 3, 9, 0)], [1,1, 42])
3×1 TimeArray{Int64, 1, DateTime, Vector{Int64}} 2025-01-01T08:00:00 to 2025-01-03T09:00:00
┌─────────────────────┬────┐
│ │ A │
├─────────────────────┼────┤
│ 2025-01-01T08:00:00 │ 1 │
│ 2025-01-02T02:00:00 │ 1 │
│ 2025-01-03T09:00:00 │ 42 │
└─────────────────────┴────┘
julia> retime(ta, Day(1); downsample=:sum)
3×1 TimeArray{Int64, 2, DateTime, Matrix{Int64}} 2025-01-01T00:00:00 to 2025-01-03T00:00:00
┌─────────────────────┬───┐
│ │ A │
├─────────────────────┼───┤
│ 2025-01-01T00:00:00 │ 1 │
│ 2025-01-02T00:00:00 │ 1 │
│ 2025-01-03T00:00:00 │ 1 │
└─────────────────────┴───┘Our result is different, since upsampling and downsampling are used depending on the number of samples found in the interval. Here, only one sample is found, but not at the asked position; hence interpolation is used.
There was a problem hiding this comment.
If no sample is found, it is interpolation.
If there are multiple samples, it is aggregation.
The question is what to do when exactly one sample is found?
There was a problem hiding this comment.
well, for the case of exactly one sample is found, could it be identity?
since I'm thinking about this:
- What happened if we have a custom aggregation method like
(x, y) -> x + 2yin Matlab?
Closes #439