Open piever opened 6 years ago
Consider how best to optimize @apply
for grouped data. Ideally, the macro should keep track of anything that happens during the grouping and only select the columns that are used.
Regarding the distributed table support, we're missing a method for view(::DTable, i)
and either broadcasting or indexing on Dagger.DArray
EDIT: all row-wise operation should automatically use delayedmap
in the distributed case. Possibly add a chunk
version of macros (@apply_in_chunks
)
[x] Think about
summarize
groupby
etc... Is there need for macros?[x] Pipeline when using functions from
IndexedTables
that don't take data as first argument?[x] Docstrings
[x] support distributed tables?
[x] allow escaping symbols
[x]
_
to refer to table inside@with
[ ] Make all macros compatible with table of tables (better to wait for table of tables proper support in JuliaDB)
[x]
begin end
syntax to make multiple transformations in a row