Filtering missing

I encountered unexpected behaviour when attempting to filter values of type Missing. I found a solution at https://discourse.julialang.org/t/query-jl-filtering-on-missing-data/14898. I suppose this issue is a feature request for documentation that clarifies missing values in Query.jl.

Anyhow, the case is as follows.

julia> using DataFrames, Query

julia> df = DataFrame(a=[1,2,3], b=[1,2,missing])
3×2 DataFrame
│ Row │ a     │ b       │
│     │ Int64 │ Int64⍰  │
├─────┼───────┼─────────┤
│ 1   │ 1     │ 1       │
│ 2   │ 2     │ 2       │
│ 3   │ 3     │ missing │

Attempting to filter for rows without values missing.

julia> df |> @filter(_.b !== missing) |> DataFrame
3×2 DataFrame
│ Row │ a     │ b       │
│     │ Int64 │ Int64⍰  │
├─────┼───────┼─────────┤
│ 1   │ 1     │ 1       │
│ 2   │ 2     │ 2       │
│ 3   │ 3     │ missing │

# Expected behaviour.
julia> df[df.b .!== missing, :]
2×2 DataFrame
│ Row │ a     │ b      │
│     │ Int64 │ Int64⍰ │
├─────┼───────┼────────┤
│ 1   │ 1     │ 1      │
│ 2   │ 2     │ 2      │

Attempting to filter for rows with values missing.

julia> df |> @filter(_.b === missing) |> DataFrame
0×2 DataFrame

# Expected behaviour.
julia> df[df.b .=== missing, :]
1×2 DataFrame
│ Row │ a     │ b       │
│     │ Int64 │ Int64⍰  │
├─────┼───────┼─────────┤
│ 1   │ 3     │ missing │

Using DataValues.jl's isna function solution provides the expected result.

df |> @filter(!Query.isna(_.b)) |> DataFrame
2×2 DataFrame
│ Row │ a     │ b      │
│     │ Int64 │ Int64⍰ │
├─────┼───────┼────────┤
│ 1   │ 1     │ 1      │
│ 2   │ 2     │ 2      │

julia> df |> @filter(Query.isna(_.b)) |> DataFrame
1×2 DataFrame
│ Row │ a     │ b       │
│     │ Int64 │ Int64⍰  │
├─────┼───────┼─────────┤
│ 1   │ 3     │ missing │

Start with the issue's Julia examples and the documented Query.isna usage. Clarify how Query.jl filtering handles missing values, including the expected patterns for matching and excluding missing entries. Done means the documentation explains the behavior and provides both filtering examples.