Indent interferes with code chunk

[kescobo]

It looks like an indent at the end of a code chunk interferes with the splitting of cells in the resulting notebook. MWE:

In "test.jl":

# Some text

a1 = [1 10 2 20; 3 30 0 0]
#- this works to split
a2 = [1 10 2 20;
    3 30 0 0]
#- this doesn't
a2 = [1 10 2 20;
    3 30 0 0]
a2
#- This does
println("What?")

julia> Literate.notebook("test.jl", "./")
[ Info: generating notebook from `~/computation/julia_playground/intro_julia/test.jl`
[ Info: not running on Travis, skipping links will not be correct.
[ Info: executing notebook `test.ipynb`
[ Info: writing result to `~/computation/julia_playground/intro_julia/test.ipynb`

[fredrikekre]

Hmm.. This kinda clashes with an undocumented feature; that we can continue code after a section of markdown, see e.g. http://kristofferc.github.io/JuAFEM.jl/latest/examples/generated/heat_equation/#Assembling-the-linear-system-1. I naively implemented this as "a code line that starts with whitespace is not a finished code block" which is what happens here. Maybe an explicit #- like here should not be allowed to be overridden.

[kescobo]

Maybe an explicit #- like here should not be allowed to be overridden.

This makes a lot of sense to me as a heuristic. If someone uses this to manually split a code chunk, there's a good chance that there's a reason for it.

That said, the ability to continue code after a chunk seems like a niche but very cool use case. If it's easier - perhaps disable it by default and allow an option to get that behavior where needed (with the understanding that it will break cases like this).