`Bins` fails when the grouping function returns non-numeric types

[shy86]

When using groupby with Bins in YAXArrays, the grouping works correctly only if the mapping function returns simple numeric types. If the function returns other types—such as strings produced by format—the bins produce empty groups or throw an error.

For example, I want to groupby year and month:

using Dates
using YAXArrays
using YAXArrays: YAXArrays as YAX

axlist = (
    YAX.time(Date("1961-01-01"):Day(1):Date("1961-04-30")),
    YAX.lon(1:10),
    YAX.lat(1:15),
)

data = rand(120, 10, 15)

a = YAXArray(axlist, data)

# Method 1: works
groupby(a, YAX.time => Bins(x -> year(x) * 100 + month(x), [[196101, 196102], [196103, 196104]]))

# Method 2: failed
groupby(a, YAX.time => Bins(x -> format(x, "yyyymm"), [["196101", "196102"], ["196103", "196104"]]))

# Method 3: failed
groupby(a, YAX.time => Bins(yearmonth, [[(1961, 1), (1961, 2)], [(1961, 3), (1961, 4)]]))

[lazarusA]

these https://rafaqz.github.io/DimensionalData.jl/stable/groupby#Binning are the ways DD handles groupby. It looks like Method 3 should work :/, Method 2 ? does it make sense at all? it needs to be a DateTime type.

[shy86]

these https://rafaqz.github.io/DimensionalData.jl/stable/groupby#Binning are the ways DD handles groupby. It looks like Method 3 should work :/, Method 2 ? does it make sense at all? it needs to be a DateTime type.这些链接：https://rafaqz.github.io/DimensionalData.jl/stable/groupby#Binning，就是 DD 处理 groupby 的方式。看起来方法 3 应该可行 :/，但方法 2 呢？它到底有没有道理啊？毕竟它必须是 DateTime 类型才行啊。

I have read the documentation, and my understanding is that the third argument of Bins should correspond to the values returned by the grouping function. Since my format(x, "yyyymm") function returns strings, I provided string bins accordingly, assuming this matched the documented behaviour.

After testing, both Method 2 and Method 3 throw errors:

julia> groupby(a, YAX.time => Bins(yearmonth, [[(1961, 1), (1961, 2)], [(1961, 3), (1961, 4)]]))
ERROR: ArgumentError: Chunk sizes must be strictly positive