LINELIST_LOST_ACTION (@Bisaloo, #126). The
environment variable needs to be set before the R session is started due
to the way R environment variables work. This allow users to set a
stricter default behaviour for tag loss if they want to ensure tag loss
never go unnoticed (#104).meta$Date in CITATION to Date
before attempting to run format() (@Bisaloo, #125)make_linelist() and set_tags() no
longer accept a named list of characters as input. Instead,
make_linelist(), set_tags() and
tags_types() now use rlang’s dynamic
dots, which means that you can splice list arguments. This
implementation is more robust, better tested, and makes it explicit that
users want to splice the list (@Bisaloo, #96).
my_tags <- list(
id = "case_ID",
date_onset = "date_of_prodrome",
age = "age",
gender = "gender"
)
make_linelist(obj, my_tags)
# OR
make_linelist(
obj,
id = "case_ID",
date_onset = "date_of_prodrome",
age = "age",
gender = "gender"
)my_tags <- list(
id = "case_ID",
date_onset = "date_of_prodrome",
age = "age",
gender = "gender"
)
make_linelist(obj, !!!my_tags)
# OR
make_linelist(
obj,
id = "case_ID",
date_onset = "date_of_prodrome",
age = "age",
gender = "gender"
)linelist warnings and errors in the case of a tag loss now have a custom class (@Bisaloo, #109), which means it is easier to silence them specifically, or to catch them programmatically for advanced error handling. One example of a new advanced condition handling that was before not possible is:
warning_counter <- 0
withCallingHandlers({
x <- linelist::make_linelist(cars, date_onset = "dist", age = "speed")
x <- x[, -1]
x <- x[, -1]
warning("This is not a linelist warning", call. = FALSE)
}, linelist_warning = function(w) {
warning_counter <<- warning_counter + 1
})
#> Warning: The following tags have lost their variable:
#> age:speed
#> Warning: The following tags have lost their variable:
#> date_onset:dist
#> Warning: This is not a linelist warning
warning("This pipeline generated ", warning_counter, " linelist warnings.")
#> Warning: This pipeline generated 2 linelist warnings.linelist objects now have a new custom
$<-.linelist() to prevent tag loss when subsetting a
linelist object (@Bisaloo, #86). This completes the
functionality already provided by the [<-.linelist() and
[[<-.linelist() methods.
x$tagged_column <- NULL
#> Warning in prune_tags(out, lost_action): The following tags have lost their variable:
#> tag:tagged_columnValidation failures in validate_types() now
integrate a delayed error mechanism (@Bisaloo, #106). This ensures that the error
message will return all the invalid tag types at once rather than having
to go through multiple trials and runs.
# No warning about age, even though it also has an invalid type
x <- make_linelist(cars, age = "speed", gender = "dist")
validate_types(x, ref_types = tags_types(age = "factor"))
#> Error in validate_types(x, ref_types = tags_types(age = "factor")) :
#> Issue when checking class of tag `gender`:
#> Must inherit from class 'character'/'factor', but has class 'numeric'x <- make_linelist(cars, age = "speed", gender = "dist")
validate_types(x, ref_types = tags_types(age = "factor"))
#> Some tags have the wrong class:
#> - gender: Must inherit from class 'character'/'factor', but has class 'numeric'
#> - age: Must inherit from class 'factor', but has class 'numeric'tag_variable() function has been replace
by a vectorized alternative tag_variable, thus improving
performance in make_linelist() and set_tags()
about twofold. The error message when tags are specified by position
with a number larger than the number of columns in the dataset to tag
has also been clarified (@Bisaloo, #110).Increased compatibility with dplyr is now documented and ensured through tests of all dplyr verbs on linelist objects as part of our testing & continuous integration system, as well as a new vignette: https://epiverse-trace.github.io/linelist/articles/compat-dplyr.html (@Bisaloo, #53)
A new selection helper is provided for tidyverse users, based on
the existing selectors provided by the tidyselect package:
has_tag() (@Bisaloo, #61). By feeding it a character
vector of tags to operate on, you can work with dplyr verbs on specific
tagged columns without having to explicitly use the column names:
x %>%
dplyr::select(has_tag(c("id", "date_of_onset")))It is no longer possible to use lost_tags_action()
within a pipeline. It must now be set as a separate step. This makes the
internal code more robust and clarifies what is part of the pipeline
versus a global option (@Bisaloo, #79).
The select_tags() function is now deprecated to
ensure we provide just one clear way to address a given issue and that
our “happy path” is clearly signposted (@Bisaloo, #61). If you were using this
function, we now recommend using the more explicit two-steps
process:
# Deprecated
x %>%
select_tags("age")
# Instead use
x %>%
tags_df() %>%
select(age)
# Or
x %>%
select(has_tag("age")) %>%
tags_df()The custom select.linelist() method has been
deprecated as providing a custom [.linelist() is sufficient
to ensure compatibility with dplyr::select() default
methods, including triggering lost_tags_action() on tag
removal (@Bisaloo,
#61). A full deletion of this method is not possible at the moment
because we want to provide a smooth transition for users that relied on
the custom tags argument of the
select.linelist() method. It is now recommend instead to
use the new has_tag() selection helper:
x %>%
dplyr::select(has_tag(c("id", "date_of_onset")))
# Instead of
x %>%
select(tags = c("id", "date_of_onset"))The custom rename.linelist() method has been removed
as providing a custom names<-().linelist method is
sufficient to ensure compatibility with dplyr::rename(),
including appropriate modification of the tags. (@Bisaloo, #60)
added a hex logo thanks to @dgmascarina contribution
added short lay description to README thanks to Emma Marty’s contribution
fixed some typos in package documentation (@pitmonticone, #30)
make_linelist(x) now errors if x
inherits from data.table (#55, @Bisaloo, based on discussions with @TimTaylor).[.linelist() now works to subset by column when
including just one argument (#54, @Bisaloo). E.g., x[1]. As an
indirect effect, this also improves compatibility with dplyr verbs that
rely on this method (#51).make_linelist(allow_extra = TRUE)) no longer causes an
error (#65, @Bisaloo;
reported by @TimTaylor in #63)This is the first proof-of-concept release on CRAN.
the linelist class implements a tagging system to
keep track of key epidemiological variables in a
data.frame
validation routines for tagged variables
accessors to retrieve tagged variables
an API for easy extension to additional tagged variables
dedicated S3 methods providing safeguards for operations which could lose tagged variables
full documentation of all functions
a vignette
100% test coverage