Syntax change in macro call produces erroneous code.

[baggepinnen]

JuliaFormatter does not understand the @variables macro from ModelingToolkit, when formatting, it tries to change the syntax appearing here

sts = @variables x(t)=x_start [description = "State of SecondOrder filter $name"] xd(t)=xd_start [
        description = "Derivative state of SecondOrder filter $name",
    ]

from short to long function definition:

function xd(t)
        xd_start
    end [
        description = "Derivative state of SecondOrder filter $name",
    ]

This produces incorrect code. Perhaps JuliaFormatter should be more careful with reformatting macro calls since they may expect a certain syntax, and the alternative syntax after formatting may be invalid.

[domluna]

I guess disabling this transform in a macro block is what we should do.

[domluna]

can you post the format arguments so I can easily reproduce it

[baggepinnen]

The formatter arguments are specified by this toml file
https://github.com/SciML/ModelingToolkitStandardLibrary.jl/blob/main/.JuliaFormatter.toml

[domluna]

ok so I have a solution for this but it would apply to all macro block calls so the following would also not turn into a long form function.

    str = raw"""
       @noinline require_complete(m::Matching) =
           m.inv_match === nothing && throw(ArgumentError("Backwards matching not defined. `complete` the matching first."))
    """
    formatted_str = raw"""
    @noinline function require_complete(m::Matching)
        m.inv_match === nothing &&
            throw(ArgumentError("Backwards matching not defined. `complete` the matching first."))
    end
    """

A way to workaround this is to have a list of macros which are known to be "fine" with this sort of thing. So @noinline would be one of those.

thoughts @baggepinnen

[domluna]

18e78ae

would also have to add a case for the opposite. Not sure what other transform is problematic.