Motivation
The original motivation for the development of this package actually was that the author found himself writing countless checks and helper code over and over again when managing parameter lists. It became apparent that something similar to python’s dictionary would make life easier and so the idea of the container package was born.
The package has undergone some changes since it’s initial version, but the dict as a use-case for parameter lists remains very valid. So without further ado, let’s see how this works out in practice.
Add or Replace
With a dict the problem of accidentally overriding an
existing parameter value is solved out of the box using the
add
function.
params = add(params, a = 0)
# Error: name 'a' exists already
add(params, x = 0) # ok
# {a = (1L 2L 3L 4L ...), b = "foo", x = 0}
Of course, it’s also possible to indeed override a parameter.
replace_at(params, a = 0)
# {a = 0, b = "foo"}
What if you intend to replace something, but there is nothing to replace?
replace_at(params, x = 0)
# Error: names(s) not found: 'x'
Now you might wonder, what if ‘I don’t care if it is replaced or added’. That’s easy.
replace_at(params, a = 0, .add=TRUE)
# {a = 0, b = "foo"}
replace_at(params, x = 0, .add=TRUE)
# {a = (1L 2L 3L 4L ...), b = "foo", x = 0}
That is, using .add = TRUE
basically means, ‘replace
it, or, if it is not there, just add it’
Maybe you agree that even these simple use-cases already require some effort when using base R lists.
Extract
When extracting a parameter, you might want to be sure it exists and signal an error otherwise.
at(params, "x")
# Error: index 'x' not found
at(params, "a", "b")
# {a = (1L 2L 3L 4L ...), b = "foo"}
To extract a single raw element, use at2
at2(params, "a")
# [1] 1 2 3 4 5 6 7 8 9 10
Alternatively, you could use the standard access operators, which
behave like base R list and therefore return an empty dict (or
NULL
) if the index is not found.
params["x"]
# {}
params[["x"]]
# NULL
params["a"]
# {a = (1L 2L 3L 4L ...)}
params[["a"]]
# [1] 1 2 3 4 5 6 7 8 9 10
Default values
A nice property of the dict is that it provides an easy and flexible way to manage default values.
That is, if you peek at a non-existing parameter, by default an empty dict is returned, but with the option to explicitly set the default. This also works for multiple peeks.
peek_at(params, "a", "x", "y", .default = 3:1)
# {a = (1L 2L 3L 4L ...), x = (3L 2L 1L), y = (3L 2L 1L)}
Remove
Similar to the above examples, the user can control how removal of
existing/non-existing parameters is handled. If you expect a parameter
and want to be signaled if it was not there, use
delete
.
Otherwise to loosely delete a parameter, regardless of whether it
exists or not, use discard
.
discard_at(params, "a", "x")
# {b = "foo"}
It’s important to note, that the “base R list way” to delete elements
does not work, because it just inserts a NULL
.
params[["a"]] <- NULL
params
# {a = NULL, b = "foo"}
Merge
Last but not least, dict allows to easily merge and/or update parameter lists.
par1 = dict(a = 1, b = "foo")
par2 = dict(b = "bar", x = 2, y = 3)
update(par1, par2)
# {a = 1, b = "bar", x = 2, y = 3}
As can be seen, existing parameters are updated and new parameters
added. Using as.dict
you can also do this with ordinary
lists.
That’s it. I hope, it will free you some time and safe some bugs next time you need to manage parameter lists.
As a very last note, keep in mind that since container version 1.0.0, dict elements are always sorted by their name, while you are still able to access elements by position (based on the sorted values).
d = dict(x = 1, z = 2, a = 3)
d
# {a = 3, x = 1, z = 2}
d[[1]]
# [1] 3
d[2:3]
# {x = 1, z = 2}