change from v0.4: children are `nothing` instead of `()`

[ericphanson]

on JuliaSyntax 0.4.10:

julia> using JuliaSyntax

julia> node = JuliaSyntax.parsestmt(SyntaxNode, "module A end")
line:col│ tree                                   │ file_name
   1:1  │[module]
   1:8  │  A
   1:9  │  [block]


julia> JuliaSyntax.children(JuliaSyntax.children(node)[1])
()

on 1.0.2:

julia> node = JuliaSyntax.parsestmt(SyntaxNode, "module A end")
SyntaxNode:
[module]
  A                                      :: Identifier
  [block]


julia> isnothing(JuliaSyntax.children(JuliaSyntax.children(node)[1]))
true

is this intentional? not sure if it is a bug or an intentional change so I thought I'd ask just in case. (I'm trying to update ExplicitImports to JuliaSyntax v1)

[Keno]

I do think the old behavior was better, since () is just as efficient a sentinel and iterates properly, but I'm not sure we can just change it back since newer clients may have started depending on the new behavior.

[ericphanson]

it is easy to work around at least: https://github.com/JuliaTesting/ExplicitImports.jl/blob/f98b7dd144e8e2270c7d247065cf5e45f726b21f/src/parse_utilities.jl#L101

I'll close this since there's probably nothing to do here

[mlechu]

Not sure I like this behaviour, but it does seem intentional. node.children == nothing means is_leaf(node) is true and node.val should be defined. There are nodes without children that are also not leaves, like from your example:

julia> node = JuliaSyntax.parsestmt(SyntaxNode, "module A end")
SyntaxNode:
[module]
  A                                      :: Identifier
  [block]

julia> node[1]
SyntaxNode:
A                                        :: Identifier

julia> isnothing(node[1].children) # is leaf
true

julia> node[2]
SyntaxNode:
[block]


julia> node[2].children # empty block
SyntaxNode[]