"@inferred requires a call expression" for do notation

[aplavin]

Lambda functions work fine, while the same do-block throws an error:

julia> using Test

julia> @inferred map(x->x, [1,2])
2-element Vector{Int64}:
 1
 2

julia> @inferred map([1,2]) do x
       x
       end
ERROR: LoadError: @inferred requires a call expression

[Keno]

gen_call_with_extracted_types already supports the required rewrite, just needs a few small tweaks in the macro itself.

[gsiddharthrao]

@aplavin Hi! 👋 I'd like to work on this issue.

The behavior here appears to come from the fact that @inferred only accepts a call expression in its macro expansion, and the do-block form (map([1,2]) do x ... end) lowers into a different AST shape than a standard call. As a result, the macro rejects the input before inference is even attempted.

I’m interested in implementing a fix so that @inferred supports do-block syntax in the same way it supports anonymous function calls. My plan is to update the macro to normalize or expand do-block forms into an equivalent call expression before the argument checks, allowing the expected inference test to run.

Before I begin preparing a PR, are there any prior discussions, preferred patterns for macro normalization, or compatibility concerns I should be aware of? If the approach sounds reasonable, please assign the issue to me—I’d be happy to take it on.

Thanks!

[KristofferC]

@gsiddharthrao, Note that while LLMs and AI can be useful help when coding, in this repo, it is quite important to state when such a tool is extensively used, even if it is for writing comments.

Also, we don't, in general, assign issues to people to avoid cookie licking. You can still make a PR without anything being assigned to you.

Note https://github.com/JuliaLang/julia/pull/59463 seems to be an existing PR along this approach.

[Mehul-Lakhwani]

Hey !! I would like to fix this issue

could you please redirect me to the source code from where it belongs

[codermayankgit]

Hi! I tried reproducing this and it seems related to macro expansion order. @inferred expects a call expression, but do-block syntax is lowered after macros run, so the macro sees a non-call expression and throws an error. Wrapping the expression in parentheses works: julia @inferred (map([1,2]) do x x end)

[arkgupta17]

Thanks for the clarification — that makes sense.

If this is considered expected behavior, I’d be happy to work on a small follow-up: either a documentation note mentioning the need for parentheses with do-blocks, or an improved error message pointing this out.

Please let me know what would be preferred.

[adienes]

are these accounts all from the same entity ...?

[codermayankgit]

are these accounts all from the same entity ...?

no