feature(cairo): Support Span<T> destructuring via fixed-size array pa…

[ilyalesokhin-starkware]

…tterns

Add support for let [a, b] = span else { ... } syntax by introducing a SliceDestructure flow control node that calls tuple_from_span to convert Span to [T; N] at runtime.

---

Note

Medium Risk
Touches semantic type-checking and match/let lowering by adding a new flow-control node and runtime conversion path; bugs here could affect pattern matching correctness and generated code for matches involving Span<T> and fixed-size array patterns.

Overview
Adds support for matching/destructuring Span<T> with fixed-size array patterns (e.g. let [a, b] = span else { ... } and match span { [a,b] => ... }).

Lowering now builds a size-dispatch chain for Span<T> patterns via a new FlowControlNode::SliceDestructure, which calls corelib tuple_from_span (with FixedSizedArrayInfoImpl) and handles success/failure branching. Semantic pattern computation is updated to accept fixed-size array patterns on Span types, and diagnostics/tests are extended to cover the new behavior.

^{Reviewed by Cursor Bugbot for commit dbd4689. Bugbot is set up for automated code reviews on this repo. Configure here.}

[reviewable-StarkWare]

This change is 

[ilyalesokhin-starkware]

• feature(cairo): Support Span<T> destructuring via fixed-size array pa… #9823 👈 (View in Graphite)
• feature(semantic): Recognize Span<T> in fixed-size array patterns #9826
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[ilyalesokhin-starkware]

@ilyalesokhin-starkware resolved 1 discussion.
Reviewable status: 0 of 11 files reviewed, 1 unresolved discussion (waiting on orizi).

[orizi]

@orizi made 2 comments.
Reviewable status: 0 of 11 files reviewed, 3 unresolved discussions (waiting on ilyalesokhin-starkware).

---

corelib/src/test/language_features/match_test.cairo line 26 at r3 (raw file):

            assert_eq!(b, @20);
            assert_eq!(c, @30);
        },

Suggestion:

        [a, b, c] => {
            assert_eq!((*a, *b, *c), (10, 20, 30));
        },

---

crates/cairo-lang-lowering/src/lower/flow_control/test_data/match line 2603 at r3 (raw file):

    match s {
        [a, b] => *a + *b,
        [a, b] => *a + *b,

add also:

Suggestion:

    match s {
        [a, b] | [a, b, _] => *a + *b,

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 465 at r5 (raw file):

                        group.filter.add(idx);
                        group.patterns.push(None);
                    }

Not sure if I should do that like in create_node_for_enum, or just use break here.

Code quote:

                    opt_wildcard_idx = Some(idx);
                    for group in size_groups.values_mut() {
                        group.filter.add(idx);
                        group.patterns.push(None);
                    }

[ilyalesokhin-starkware]

corelib/src/test/language_features/match_test.cairo line 26 at r3 (raw file):

            assert_eq!(b, @20);
            assert_eq!(c, @30);
        },

Done.

[ilyalesokhin-starkware]

@ilyalesokhin-starkware made 1 comment and resolved 1 discussion.
Reviewable status: 0 of 11 files reviewed, 3 unresolved discussions (waiting on orizi).

---

crates/cairo-lang-lowering/src/lower/flow_control/test_data/match line 2603 at r3 (raw file):

Previously, orizi wrote…

add also:

Done.

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 465 at r5 (raw file):

Previously, ilyalesokhin-starkware wrote…

Not sure if I should do that like in create_node_for_enum, or just use break here.

see r5 <-> r6.

[cursor]

PR Summary

Medium Risk
Adds new lowering/control-flow graph machinery for Span<T> fixed-size array destructuring via runtime TryInto, which changes match/let-else codegen paths and could affect pattern reachability/exhaustiveness behavior.

Overview
Adds lowering support for destructuring Span<T> using fixed-size array patterns (e.g. [a, b], []) in match/if let by introducing a new flow-control node, SliceDestructure, that tries sizes in order and falls through on length mismatch.

The lowering pipeline now builds a Span<T> size-dispatch chain when FixedSizeArray patterns are present, and lowers it by reconstructing a Span from @Array, calling the core TryInto::try_into impl to obtain a boxed fixed-size array, then unboxing/destructuring elements on success.

Updates core semantic CoreInfo to expose TryInto/try_into and box_forward_snapshot, and extends tests/fixtures to cover span matching (including empty patterns, inner-pattern mismatches, wildcard/struct catch-alls, and unreachable/non-exhaustive diagnostics).

^{Reviewed by Cursor Bugbot for commit 29abfa2. Bugbot is set up for automated code reviews on this repo. Configure here.}

[orizi]

@orizi reviewed 6 files and all commit messages, made 1 comment, and resolved 2 discussions.
Reviewable status: 6 of 11 files reviewed, 3 unresolved discussions (waiting on ilyalesokhin-starkware).

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 549 at r6 (raw file):

    /// single inner pattern and feed into [create_node_for_patterns], which wants it already
    /// unwrapped, hence the asymmetry.
    patterns: Vec<PatternOption<'a, 'db>>,

doesn't the index in the filter provide this already?

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 549 at r6 (raw file):

Previously, orizi wrote…

doesn't the index in the filter provide this already?

It can also be None, so I would need some index to include that.
Also, FilteredPatterns doesn't provide a convenient way to access the index.

This is smpler

[ilyalesokhin-starkware]

@ilyalesokhin-starkware resolved 1 discussion.
Reviewable status: 6 of 11 files reviewed, 2 unresolved discussions (waiting on orizi).

[orizi]

@orizi reviewed 1 file and made 3 comments.
Reviewable status: 7 of 11 files reviewed, 4 unresolved discussions (waiting on ilyalesokhin-starkware).

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 549 at r6 (raw file):

Previously, ilyalesokhin-starkware wrote…

It can also be None, so I would need some index to include that.
Also, FilteredPatterns doesn't provide a convenient way to access the index.

This is smpler

got it - call it inner_patterns maybe like we call it in the VariantInfo struct.

even possibly you can use the exact same struct (and change its name) and share much more code

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 401 at r7 (raw file):

/// Each size is tried in order. On failure, the next size is attempted. If all sizes fail,
/// the wildcard/otherwise patterns are used.
#[allow(clippy::too_many_arguments)]

is there no good logical grouping for params to remove this?

Code quote:

#[allow(clippy::too_many_arguments)]

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 430 at r7 (raw file):

            Some(semantic::Pattern::FixedSizeArray(p)) => Some(p),
            _ => None,
        });

Suggestion:

        let first_fsa = patterns.iter().flatten().find_map(|p| try_extract_matches!(p, semantic::Pattern::FixedSizeArray));

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 430 at r7 (raw file):

            Some(semantic::Pattern::FixedSizeArray(p)) => Some(p),
            _ => None,
        });

decided to remove it; it shouldn't have happened anyway.

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 3b35bee. Configure here.}

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 549 at r6 (raw file):

Previously, orizi wrote…

got it - call it inner_patterns maybe like we call it in the VariantInfo struct.

even possibly you can use the exact same struct (and change its name) and share much more code

/// Note: unlike [VariantInfo], which stores the inner pattern of each `EnumVariant` arm,
/// here we store the whole `FixedSizeArray` pattern. The next stage —

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 401 at r7 (raw file):

Previously, orizi wrote…

is there no good logical grouping for params to remove this?

Done.

[ilyalesokhin-starkware]

@ilyalesokhin-starkware resolved 1 discussion.
Reviewable status: 7 of 11 files reviewed, 3 unresolved discussions (waiting on orizi).

[orizi]

@orizi reviewed 1 file and all commit messages, made 3 comments, and resolved 3 discussions.
Reviewable status: 8 of 11 files reviewed, 3 unresolved discussions (waiting on ilyalesokhin-starkware).

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 396 at r9 (raw file):

/// Each size is tried in order. On failure, the next size is attempted. If all sizes fail,
/// the wildcard/otherwise patterns are used.
fn try_create_slice_destructure_chain<'db>(

can you separate this function into 2:

1. check if you want to try to match using the [] pattern.
2. actually return a node id.

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 411 at r9 (raw file):

    // the `StructConstruct(Span<T>)` it emits rely on this: `snapshot_array_var` below must be
    // exactly `@Array<T>`, not a wrapped variant.
    if wrapping_info.n_outer_snapshots != 0 || wrapping_info.n_boxed_inner_snapshots.is_some() {

can you add a test with:

match arr.span() {
      [v0] => {...},
      [v1] => {...},
      Span { snapshot: inner } =>{...}
}

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 511 at r9 (raw file):

    }));

    Some(chain)

Suggestion:

    Some(graph.add_node(FlowControlNode::Deconstruct(Deconstruct {
        input: input_var,
        outputs: vec![snapshot_array_var],
        next: failure_node,
    })))

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 411 at r9 (raw file):

Previously, orizi wrote…

can you add a test with:

match arr.span() {
      [v0] => {...},
      [v1] => {...},
      Span { snapshot: inner } =>{...}
}

Do we want to support don't catch everything?

[ilyalesokhin-starkware]

@ilyalesokhin-starkware made 3 comments.
Reviewable status: 7 of 11 files reviewed, 4 unresolved discussions (waiting on orizi).

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 396 at r9 (raw file):

Previously, orizi wrote…

can you separate this function into 2:

1. check if you want to try to match using the [] pattern.
2. actually return a node id.

Done.

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 511 at r9 (raw file):

    }));

    Some(chain)

Done.

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 503 at r10 (raw file):

                // A catch-all struct pattern makes every subsequent arm unreachable, so we stop
                // adding to `size_groups`. The match-completeness pass will flag the dead arms.
                break;

we can break here for now, right?

Code quote:

                // A catch-all struct pattern makes every subsequent arm unreachable, so we stop
                // adding to `size_groups`. The match-completeness pass will flag the dead arms.
                break;

[orizi]

@orizi reviewed 2 files and all commit messages, made 4 comments, and resolved 2 discussions.
Reviewable status: 9 of 11 files reviewed, 4 unresolved discussions (waiting on ilyalesokhin-starkware).

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 411 at r9 (raw file):

Previously, ilyalesokhin-starkware wrote…

Do we want to support don't catch everything?

this just checks that can open the type of "catch everything" object.

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 503 at r10 (raw file):

Previously, ilyalesokhin-starkware wrote…

we can break here for now, right?

assuming they are still considered merged and supply diagnostics - this make sense.

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 354 at r10 (raw file):

            elem_ty,
        );
    }

Suggestion:

    if let Some(elem_ty) =
        should_create_slice_destructure_chain(params.ctx.db, params.patterns, concrete_struct_id, wrapping_info)
        // or change signature to `should_create_slice_destructure_chain(&params, concrete_struct_id, wrapping_info)`
    {
        return create_slice_destructure_chain(
            params,
            input_var,
            concrete_struct_id,
            elem_ty,
        );
    }
    let CreateNodeParams { ctx, graph, patterns, build_node_callback, location } = params;

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 443 at r10 (raw file):

    let snapshot_member_id = members.iter().next().unwrap().1.id;
    let snapshot_array_ty = members.iter().next().unwrap().1.ty;
    let snapshot_array_var = graph.new_var(snapshot_array_ty, location);

Suggestion:

    let members = match ctx.db.concrete_struct_members(concrete_struct_id) {
        Ok(members) => members.iter().exactly_one(),
        Err(diag_added) => return graph.add_node(FlowControlNode::Missing(diag_added)),
    };
    let (snapshot_member_id, snapshot_array_ty) = match members.iter().exactly_one() {
        Ok((_, member)) => (member.id, member.ty),
        Err(e) => panic!("Got wrong number of `Span` members: `{e:?}`."),
    };
    let snapshot_array_var = graph.new_var(snapshot_array_ty, location);

[ilyalesokhin-starkware]

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 503 at r10 (raw file):

Previously, orizi wrote…

assuming they are still considered merged and supply diagnostics - this make sense.

what do you mean by considered merged?

[orizi]

@orizi made 1 comment.
Reviewable status: 9 of 11 files reviewed, 4 unresolved discussions (waiting on ilyalesokhin-starkware).

---

crates/cairo-lang-lowering/src/lower/flow_control/create_graph/patterns.rs line 503 at r10 (raw file):

Previously, ilyalesokhin-starkware wrote…

what do you mean by considered merged?

just the goto at the end the arm to the continuation of the flow.