jariji

FWIW, `Base` (unfortunately, imho) uses the word `unique`, but in general I think the right word for this concept is not `unique` but `distinct`.

The join is on `[:a,:b]`, and only `:b` has missings, so the tables can be large since the match-all policy works only within rows for which `:a` already matches. The...

The use case is to refine a join by adding the additional column `:b` to the `;on=` list, but `:b` is sometimes missing, in which case it should not restrict...

To be more precise, I want to say `matchmissing for :left_b => :right_b = :matchall` but `matchmising for :a = :error`.

The existing way to do this is to join only on `:a` and then in a second step `subset` based on `:b`.

```jl innerjoin(l,r; on=[:a,:b], matchmissing=[:a=>:error, :b => :matchall]) ``` This is close, though it's still ambiguous whether the `missing`s are expected to be on the left or the right or both.

Inserting implicitly missing values in a sequence is more general than just time. Tidyr has a nice function `full_seq` for this. https://tidyr.tidyverse.org/reference/full_seq.html ```R full_seq(c(1, 2, 4, 5, 10), 1) #>...

Collecting a rowtable from an iterator can be a costly operation and I'd want to know when I'm doing it, so I'd rather get an error.

> Indexing on the left is messier, as it's not entirely clear whether you want points with the same ixs[j] to overwrite or accumulate If I understand correctly, overwriting is...

Another one. I think I'm calling `copy(::DataFrame)`. ```jl signal (11): Segmentation fault in expression starting at /home/... __memmove_sse2_unaligned_erms at /nix/store/s9qbqh7gzacs7h68b2jfmn9l6q4jwfjz-glibc-2.33-59/lib/libc.so.6 (unknown line) jl_pchar_to_string at /buildworker/worker/package_linux64/build/src/array.c:507 unsafe_string at ./strings/string.jl:67 [inlined] fromarrow...