Completion source for nvim-cmp using R.nvim as backend.
Use a plugin manager such as vim-plug or packer.
Register the source for nvim-cmp
:
require'cmp'.setup {
sources = {
{ name = 'cmp_r' },
}
}
cmp-r
has the following options:
-
filetypes
: The list of file types for which the source should be enabled. Default:{'r', 'rmd', 'quarto', 'rnoweb', 'rhelp'}
-
doc_width
: The expected maximum width of the floating documentation window. Valid values are between30
and160
. Default:58
.See also the option
maxwidth
in lspkind.nvim. -
trigger_characters
: List of characters to trigger completion. Default:{ " ", ":", "(", '"', "@", "$" }
. -
fun_data_1
: List of functions that receive adata.frame
as its first argument and for which thedata.frame
s columns names should be completed. This option is overridden byg:R_fun_data_1
. Default:{'select', 'rename', 'mutate', 'filter'}
. -
fun_data_2
: Dictionary with parent functions as keys. The value of each key is a list of functions that are expected to be nested in the parent function which receives a data.frame as its first arguments. Column names of thedata.frame
are completed as arguments of the nested function. This option is overridden byg:R_fun_data_1
. Default:{ggplot = {'aes'}, with = {'*'}}
. -
quarto_intel
: Path toyaml-intelligence-resources.json
which is part ofquarto
application and has all necessary information for completion of valid YAML options in an Quarto document. Default:nil
(cmp-r
will try to find the file).
Below is an example of how to set theses options:
require("cmp_r").setup({
filetypes = {"r", "rmd", "quarto"},
doc_width = 58,
quarto_intel = "~/Downloads/quarto-1.1.251/share/editor/tools/yaml/yaml-intelligence-resources.json"
})
Note: The algorithm for finding the first argument of the parent function is
very simple and would wrongly recognize data
as the name of the
data.frame
in the following code:
ggplot(data = my_data_frame, mapping = aes(
The source cmp_r
does not require any special configuration of
nvim-cmp
to work, and people have different preferences and workflows.
Anyway, I share below what worked best for me:
-
cmp-r
sets the fieldsortText
of completed items as "0" for function arguments and "9" for everything else. This means that function arguments will always be at the top of the list of suggestions if you putcmp.config.compare.sort_text
at the top of list of comparenvim-cmp
functions. -
Install a Nerd font and install and configure your terminal emulator to use it. Then install lspkind.nvim to have symbols representing the kind of object whose name is being completed.
-
The names of kind of objects are generic names of the LSP server protocol and do not match the classes of R objects. So, it is better to hide the names and display only the symbols.
An example following the above suggestions:
cmp.setup({
formatting = {
fields = {'abbr', 'kind', 'menu'},
format = lspkind.cmp_format({
mode = 'symbol', -- show only symbol annotations
maxwidth = 50, -- prevent the popup from showing more than provided characters
ellipsis_char = '...', -- the truncated part when popup menu exceed maxwidth
before = function(entry, item)
local menu_icon = {
nvim_lsp = '',
vsnip = '',
path = '',
cmp_zotcite = 'z',
cmp_r = 'R'
}
item.menu = menu_icon[entry.source.name]
return item
end,
})
},
sources = cmp.config.sources({
{ name = 'vsnip' },
{ name = 'cmp_zotcite' },
{ name = 'cmp_r' },
{ name = 'nvim_lsp' },
{ name = 'path', option = { trailing_slash = true } },
}),
})
Since cmp-r
and the R package languageserver
provide completions for
the same code, and completions from the languageserver
might be the first
ones to be displayed by nvim-cmp
, you may want to put this in your ~/.Rprofile
:
# Disable completion from the language server
options(languageserver.server_capabilities =
list(completionProvider = FALSE, completionItemResolve = FALSE))