Releases: asardaes/table.express
Releases · asardaes/table.express
Version 0.4.2
Version 0.4.1
- Avoid potential deep copies in
left_join
withdata.table
as input (#22). - Fix HTML documentation.
Version 0.4.0
- The logic of
mutate
with.sequential = TRUE
has been changed to support scenarios where
group_by
orkey_by
were used as well. transmute
gained the same.sequential
argument with the same underlying logic.summarize
and aliases gained parameter.assume_optimized
in casetable.express
doesn't know
about a function thatdata.table
can optimize.summarize
and aliases can support the sequential logic mentioned before if the summarizing
functions are not assumed to be optimized.
Version 0.3.3
- More test adjustments to fix problems on CRAN for real.
Version 0.3.2
- Adjusted tests to avoid problems on CRAN.
Version 0.3.1
- Improved compatibility with packages that use
dplyr
without importingdata.table
(#21). - Added
summarize
/summarise
verbs. right_join
gained parameterwhich
.
Version 0.3.0
- Verbs can now be used without
start_expr
/chain
/end_expr
with some considerations, check the
updated vignettes (#3). Importantly, this makestable.express
conflict withdtplyr
. - Fixed a bug that prevented automatically built chains to work after a
left_join
. - The
*_sd
verbs now support lambdas as formulas (#19). - Added
max_by
andmin_by
verbs. mutate_sd
gained parameter.pairwise
.distinct
gained parameters.keep
and.n
.transmute
gained parameter.enlist
(#20).filter_on
now allows empty names for data tables that already have keys (#17).- All filtering verbs now have a
which
parameter (#17). - Added
nest_expr
for certain scenarios that need the captureddata.table
; check the vignette
(#16). - Simplified
semi_join
.
Version 0.2.0
- Added joining verbs and a corresponding vignette describing them (#1).
- Added the
distinct
verb. transmute
is no longer an alias forselect
. This is to make the latter more flexible in some
bare selection cases (like combining severaltidyselect
calls), and leave the former more simple
and preferable for actual transmutation cases (#4).- Both
mutate_sd
andtransmute_sd
now support a list of functions as input, but with slightly
different semantics; check the vignette (#8 and #11). - Now both
mutate_sd
andfilter_sd
support the.COL
pronoun for.SDcols
predicates, but they
remain eager in this regard. select
andfilter_on
gained a.negate
parameter (#9 and #6).mutate
gained a.sequential
parameter to enable usage of newly created columns.- Added some debugging statements when option
table.express.verbose
isTRUE
, or
.verbose = TRUE
instart_expr
/chain
. - Updates to main vignette.
- Semi-breaking change:
*_sd
verbs now have.SDcols
before their ellipsis. This should only
require code adjustment fortransmute_sd
when.SDcols
was left missing.
Version 0.1.1
- Let
transmute_sd
build expressions withlapply
when.how
is a function and.SDcols
is
character or numeric. - Support numeric input in
select
. - Removed
tracemem
example from vignette to avoid CRAN problems.
Version 0.1.0
- Initial version with single-table verbs.