Date vs DateTime when the time is 0:00:00

[hurak]

This is clearly a duplicate of #237, which stalled, and that is why I open it as a new issue. Perhaps the two can then merge or just be closed together or whatever.

I created a very minimal Excel file (download here), see the screenshot

When importing this as a table using the code

julia> using XLSX

julia> xt = XLSX.readtable("bug_report_mwe.xlsx", "List1")
XLSX.DataTable(Any[Any[Date("2020-10-22"), DateTime("2020-10-22T00:10:00")], Any[10, 20]], [:DateTime, :Value], Dict(:Value => 2, :DateTime => 1))

the type of the first date-and-time cell entry is just Date, while that of the second is DateTime. The following code shows this explicitly

julia> xt.data
2-element Vector{Any}:
 Any[Date("2020-10-22"), DateTime("2020-10-22T00:10:00")]
 Any[10, 20]

This creates problems when working with date-and-timed data sets (typically with DataFrames.jl or TSFrames.jl) as the type of the elements in the first (the index) column is not a subset of the abstract TimeType. Clearly, even 0:00:00 should be recognised as a valid time.

[hurak]

Well, I have just learnt of the option "infer_eltypes". With it the import does not suffer the above behaviour:

julia> xt = XLSX.readtable("bug_report_mwe.xlsx", "List1", infer_eltypes=true)
XLSX.DataTable(Any[[DateTime("2020-10-22T00:00:00"), DateTime("2020-10-22T00:10:00")], [10, 20]], [:DateTime, :Value], Dict(:Value => 2, :DateTime => 1))

julia> xt.data
2-element Vector{Any}:
 [DateTime("2020-10-22T00:00:00"), DateTime("2020-10-22T00:10:00")]
 [10, 20]

Now I am even hesitating if the originally described behaviour can be regarded as a bug. The fact is that the behaviour with the infer_eltype option is the one that I would expect and that I am perfectly happy with.

[TimG1964]

Issue #155 proposes making the default infer_eltypes=true and PR #313 implements this - if it is accepted.