Julia 1.0/1.1 can't handle deep nested structs #31663
Comments
JeffBezanson
added
the
regression
felipenoris
changed the title
|
I did a few more tests, and it looks like Julia is able to construct the value, but it can't print it. Base.show(io::IO, n::AbstractNode) = print(io, "a node")it succeeds: But my original issue arised from a type that had customized |
felipenoris
changed the title
fredrikekre
added
the
display and printing
|
Possibly related to #31572? |
|
Yes. I'm having other side effects beyond printing, related to structs with many type parameters. I'm trying to isolate the problem, but it looks like related to #31572. |
- avoid exponential search in `is_derived_type` when parameters are used as field types - avoid inferring `show_default` - improve a fast path in subtyping
- avoid exponential search in `is_derived_type` when parameters are used as field types - avoid inferring `show_default` - improve a fast path in subtyping
|
This example exposes the problem regarding other side effects beyond printing: abstract type AbstractNode end
struct Node{N1<:AbstractNode, N2<:AbstractNode} <: AbstractNode
a::N1
b::N2
end
struct Leaf <: AbstractNode
end
function gen_nodes(qty::Integer) :: AbstractNode
@assert qty > 0
result = Leaf()
for i in 1:qty
result = Node(result, Leaf())
end
return result
end
count_leafs(l::Leaf) = 1
count_leafs(n::Node) = count_leafs(n.a) + count_leafs(n.b)
# this function call triggers the problem and does not return
count_leafs(gen_nodes(50)) |
felipenoris
changed the title
|
This is great! 8a6271b fixes both the printing and also this new example I came up with. |
|
Just for the record, With |
FWIW, this usually isn’t an issue for data-structures since trees are usually parameterized on their element type, not the tree size and shape |
|
@vtjnash, yes, you're right. this only happens on highly parametrized structs. But keep in mind this is a minimal working example of the issue, which is not useful by itself. |
I'm migrating code from v0.6 to v1.0/v1.1 , and I found this issue that breaks Julia.
Running on Julia v1.0 or v1.1, it runs ok for 10 nested elements, but it never finishes if I try to nest about 50 elements. In the example below, I had to use
CTRL+Cto cancel.The same code runs without issues on Julia v0.6, even if I increase to 1000 nested elements.
I would say that this is a major issue for any code that deals with trees or similar data structures.
The text was updated successfully, but these errors were encountered: