Format `Dict` definitions like `Vector` definitions
efaulhaber opened this issue · comments
We're using the SciML style guide since we found that it's very consistent with our "in-house" style.
However, we would like to format Dict definitions similar to Vector definitions like this:
x = [
"some arbitrary string bla bla bla bla bla bla bla bla bla",
"another longer arbitrary string bla bla bla bla bla bla bla bla",
]
SomeLongerTypeThanJustString = String
y = Dict{Int, SomeLongerTypeThanJustString}(
1 => "some arbitrary string bla bla bla bla bla bla bla bla bla",
2 => "another longer arbitrary string bla bla bla bla bla bla bla bla"
)After auto-formatting, however, it looks like this:
y = Dict{Int, SomeLongerTypeThanJustString}(1 => "some arbitrary string bla bla bla bla bla bla bla bla bla",
2 => "another longer arbitrary string bla bla bla bla bla bla bla bla")This also goes way beyond the 92-character limit.
Is there a way to redefine this behavior without changing the rest of the formatting rules?
currently there's no way to mix and match like that without creating your own style and overloading some calls. For the array calls you would need to overload p_vect and maybe something else as well to handle the typed case x = A[1, 2].
For the Dict definition it would be p_call and then you would check if the caller is Dict.
In both of these cases you would dispatch to DefaultStyle otherwise SciMLStyle
Thanks! The array call is already formatted the way I want it to. I just want to format Dict calls according to DefaultStyle.
From my understanding, before I even try to figure out how to check if the caller is a Dict, just redefining
function JuliaFormatter.p_call(style::SciMLStyle, cst::CSTParser.EXPR, s::JuliaFormatter.State)
JuliaFormatter.p_call(DefaultStyle(style), cst, s)
endshould already format the Dict definition the way I want to, right? But it would then also change the formatting of function calls, which I don't want to do.
However, after this redefinition, I now end up with this:
y = Dict{Int, SomeLongerTypeThanJustString}(
1 => "some arbitrary string bla bla bla bla bla bla bla bla bla",
2 => "another longer arbitrary string bla bla bla bla bla bla bla bla"
)So, one additional line break after the opening parenthesis, but still with the indent that I don't want.
yes that should work!
oh sorry you'll also have to update the nesting function as well since that determines how it's indented, etc. n_call! - here's an example https://github.com/domluna/JuliaFormatter.jl/blob/master/src/styles/yas/nest.jl#L1
but that will work for all function calls not just ones with Dict, maybe that's what you want but if not you would need to verify the caller is Dict.
Thanks, now that part works.
How can I verify that the caller is Dict?
cst[1] will be that node. It seems either it's a curly node or a leaf node
JuliaFormatter.is_leaf(cst[1])
will check this
so you can do
is_dict_call = false
if JuliaFormatter.is_leaf(cst[1]) && cst[1].val == "Dict"
is_dict_call = true
elseif !JuliaFormatter.is_leaf(cst[1]) && JuliaFormatter.is_leaf(cst[1][1]) && cst[1][1].val == "Dict"
is_dict_call = true
end
then dispatch based on is_dict_call
and then do something similar for n_call!
it should actually work the exact same for an FST node
Thank you very much! Everything works as expected. Only forwarding all functions to SciMLStyle was a bit of a hassle.
Now, is there any way to use my custom style in VS Code? It seems like the config only allows the predefined styles:
JuliaFormatter.jl/src/JuliaFormatter.jl
Lines 939 to 946 in 9092f29
Alright, I decided not to create a completely new style that only really changes this one thing, but to instead add this as an option to SciMLStyle.
I created a PR, which is now ready for review: #676