How to interpret xticks when plotting time series in Julia with Plots.jl and TimeSeries.jl

Question

I have the following code:

using Plots
using TimeSeries

apple_ta = readtimearray("apple.csv")

p = plot(apple_ta[:Open])

Plots.xticks(p)

1-element Vector{Tuple{Vector{Float64}, Vector{String}}}:
 ([735599.0, 736330.0, 737060.0, 737791.0], ["2015-01-01", "2017-01-01", "2019-01-01", "2021-01-01"])

When I print the xticks, I get the above output but have no idea where values like 735599 are from. First, I thought they must be timestamps of the dates from the XAxis but they aren't. Where do they come from?

My final goal is to be able to set the xticks with xticks! function to whichever dates I want. Does anybody know how to?

PS. Here is the top 5 lines of apple.csv:

Date,Open,Close,Volume
2015-01-02,27.84749984741211,27.332500457763672,212818400
2015-01-05,27.072500228881836,26.5625,257142000
2015-01-06,26.635000228881836,26.565000534057617,263188400
2015-01-07,26.799999237060547,26.9375,160423600
2015-01-08,27.3075008392334,27.97249984741211,237458000

Answer 1

To answer your question directly, Dates.value(::Date) returns the number of days since December 31st, year 0:

julia> Dates.value(Date(0, 12, 31))
0

julia> Dates.value(Date(1, 1, 1))
1

However it seems to me that something is off with your TimeArray - there's a plot recipe in TimeSeries for plotting data with correctly formatted xticks, here's an MWE using the data you provided:

julia> using DataFrames, TimeSeries, Plots, DelimitedFiles

julia> df = readdlm(IOBuffer("""Date,Open,Close,Volume
       2015-01-02,27.84749984741211,27.332500457763672,212818400
       2015-01-05,27.072500228881836,26.5625,257142000
       2015-01-06,26.635000228881836,26.565000534057617,263188400
       2015-01-07,26.799999237060547,26.9375,160423600
       2015-01-08,27.3075008392334,27.97249984741211,237458000
       """), ',')
6×4 Matrix{Any}:
 "Date"          "Open"    "Close"           "Volume"
 "2015-01-02"  27.8475   27.3325    212818400
 "2015-01-05"  27.0725   26.5625    257142000
 "2015-01-06"  26.635    26.565     263188400
 "2015-01-07"  26.8      26.9375    160423600
 "2015-01-08"  27.3075   27.9725    237458000

julia> ts = TimeArray(DataFrame(date = Date.(df[2:end, 1]), value = df[2:end, 3]); timestamp = :date)
5×1 TimeArray{Any, 1, Date, Vector{Any}} 2015-01-02 to 2015-01-08
│            │ value   │
├────────────┼─────────┤
│ 2015-01-02 │ 27.3325 │
│ 2015-01-05 │ 26.5625 │
│ 2015-01-06 │ 26.565  │
│ 2015-01-07 │ 26.9375 │
│ 2015-01-08 │ 27.9725 │

julia> plot(ts)

This produces:

Slightly less than ideal cropping on the right hand side, but this can be fixed by either adding right_margin = 5Plots.mm or rotating the xticks via xrot = 30 (or some other value).

Are you maybe on some outdated version of TimeSeries or Plots? Here are the versions used to produce the above:

(jl_ypFE6F) pkg> st
Status `/tmp/jl_ypFE6F/Project.toml`
  [a93c6f00] DataFrames v1.4.4
  [91a5bcdd] Plots v1.38.0
  [9e3dc215] TimeSeries v0.23.1

Answer 2

The numbers returned by xticks for each date are the result of calling Dates.value function:

using Dates

dt = Date(2015, 01, 01)
value = Dates.value(dt)

>>> println(value)
735599

So, if you want to set custom xticks when plotting dates, you can do the following:

dates = [Date(2015, 01, 01), Date(2016, 01, 01), Date(2017, 01, 01)]
ticks = Dates.value.(dates)
labels = string.(dates)

xticks!(ticks, labels, rotation=20)