Add local variables, closes #2752 #8740

codeshaunted · 2025-06-20T17:20:30Z

Adds support for immutable local variables in .slint functions using a new let keyword with an optional type annotation:

let foo: int = 1; // with type annotation
let bar = 2; // without type annotation

Shadowing is allowed:

let foo = 1;
let foo = foo + 1;

debug(foo); // prints "2"

let foo = "hello world";

debug(foo); // prints "hello world"

Internally an extra discriminator field was added to the StoreLocalVariable and ReadLocalVariable expressions, so that shadowing support could be added in the C++ generator, as C++ does not support variable shadowing.

ogoffart

Thanks, this is great!

Just added some comment inline, in particular, i think the "discriminator" should be part of the name in the Expression

Another thing is that the discriminator should probably be global, because otherwise we might have conflicts with inlining.

There should also be tests for error in internal/compiler/tests/syntax. For example that test that the types are respected and that it is error otherwise.

let foo : int = "1" should be an error. `property xyz : { let foo : string = 1; return foo; } should be an error, and so on.

Shadowing is allowed:

I know i said in the chat it should be allowed. But I'm thinking we shouldn't allow this at first in Slint. Make it an error for now and we can still allow it later if we think this is useful, but it is not used in other programming languages, so i wonder if we should really do that.

Anyway, great patch, thanks for contributing.

ogoffart · 2025-06-21T07:56:38Z

docs/astro/src/content/docs/guide/language/coding/functions-and-callbacks.mdx

@@ -39,6 +39,37 @@ Functions can be annotated with the `pure` keyword.
 This indicates that the function does not cause any side effects.
 More details can be found in the <Link type="Purity" /> chapter.

+## Local Variables


I think this should be documented in the expressions-and-statements section, because it is basically available everywhere, not only in functions.

ogoffart · 2025-06-21T07:59:18Z

internal/compiler/expression_tree.rs

@@ -611,13 +611,15 @@ pub enum Expression {
    /// Should be directly within a CodeBlock expression, and store the value of the expression in a local variable
    StoreLocalVariable {
        name: SmolStr,
+        discriminator: Option<usize>,


i'd rather the discriminator be part of the name.
The name should probably be prefixed as well so that it doesn't conflict with internal variable. Eg, if you have let foo = ..., the name should be local_foo_1

ogoffart · 2025-06-21T08:09:11Z

internal/compiler/parser/statements.rs

@@ -50,6 +55,11 @@ pub fn parse_statement(p: &mut impl Parser) -> bool {
        return true;
    }

+    if p.peek().as_str() == "let" {


I know we have the same problem with return, but we should check that the next token is also an identifier. So if there is a property called let, code like let = 3; or let + 1 still works.

ogoffart · 2025-06-21T08:15:03Z

tests/cases/expr/let.slint

+        if (true) {
+            let a = a + 1;
+
+            return a;


can we also have a version of the test without the return here? (to make sure that the original value is not changed.

tests/cases/expr/let.slint

tools/lsp/language/semantic_tokens.rs

ogoffart

Thanks again. This looks good to me.
Please just cleanup the patch a bit for the unnecessary changes.

I see that the C++ test is failing with a warning about unused variable. To solve this we could either:

add [[maybe_unused]] to the variable declaration. (Although this warning is valid, it shows in the generated code instead of the Slint code)
ignore this particular warning for this particular test somehow.

ogoffart · 2025-06-25T02:36:25Z

internal/compiler/generator/rust.rs

@@ -2516,7 +2516,7 @@ fn compile_expression(expr: &Expression, ctx: &EvaluationContext) -> TokenStream
            }
        }

-        Expression::StoreLocalVariable { name, value } => {
+        Expression::StoreLocalVariable { name, value, .. } => {


Since there are no more extra discriminant, the patch could be a bit cleaned up for these changes.

ogoffart · 2025-06-25T02:39:29Z

internal/compiler/generator/cpp.rs

@@ -3132,7 +3132,7 @@ fn compile_expression(expr: &llr::Expression, ctx: &EvaluationContext) -> String
        Expression::StoreLocalVariable { name, value } => {
            format!("auto {} = {};", ident(name), compile_expression(value, ctx))
        }
-        Expression::ReadLocalVariable { name, .. } => ident(name).to_string(),
+        Expression::ReadLocalVariable { name, .. } => format!("{}", ident(name)),


This can be reverted. I think clippy actually warns about use of format when to_string can be used

ogoffart · 2025-06-25T02:49:05Z

internal/compiler/passes/resolving.rs

+                    .push_error(format!("Redeclaration of local variables is not allowed"), &node);
+                return Expression::Invalid;
+            }
+            // conflicts with something else


Imho we should be careful because this means that adding anything in Skint becomes a breaking change.

Imagine someone writes

let Foo = 42;

But now if I'm adding an enum called Foo in Slint, this code would break.

So I think we should only reject if it overrides a local variables (including the function parameters) but not a global one.

ogoffart · 2025-06-25T02:52:40Z

internal/compiler/tests/syntax/basic/let.slint

+// SPDX-License-Identifier: GPL-3.0-only OR LicenseRef-Slint-Royalty-free-2.0 OR LicenseRef-Slint-Software-3.0
+
+export X := Rectangle {
+//       ^warning{':=' to declare a component is deprecated. The new syntax declare components with 'component MyComponent {'. Read the documentation for more info}


Just use the new syntax in new tests.

ogoffart · 2025-06-25T02:53:23Z

internal/compiler/tests/syntax/basic/let.slint

+//       ^warning{':=' to declare a component is deprecated. The new syntax declare components with 'component MyComponent {'. Read the documentation for more info}
+    background: {
+        let blue: brush = 42;
+//      ^error{Local variable declaration conflicts with existing name}


See other comment why I think this should actually be accepted.

ogoffart

Thanks.
Looks good to me.
I'll merge soon.

codeshaunted added 7 commits June 20, 2025 11:23

working impl of local variables for rust generator and interpreter

eb24734

add discriminator to fix cpp generator

e231a99

add type annotation support

a8d509b

update test

b0cb227

update docs

d415eaa

fix mistake in docs

495224a

add formatter

19c20ea

codeshaunted force-pushed the avery/local-variables branch from 5ef96a0 to 19c20ea Compare June 20, 2025 21:21

ogoffart requested changes Jun 21, 2025

View reviewed changes

ogoffart reviewed Jun 21, 2025

View reviewed changes

tests/cases/expr/let.slint Show resolved Hide resolved

tools/lsp/language/semantic_tokens.rs Show resolved Hide resolved

codeshaunted added 11 commits June 23, 2025 10:13

make local redeclaration an error, prefix locals with "local_"

1890fd5

add extra check for following identifier for let statement in parser

25d1041

add extra resolving error for overlapping global names, add syntax test

10321f7

rework tests, add interpreter test

90f837e

move docs

f55a551

improve tests

c4f70fe

remove unused import

5dd1814

adjust semantic tokens and add fmt test

3082bed

update syntax tests with extra case

4608667

add scoping and fix const propogation bug

5fa86c2

update comment on let test

84fd545

codeshaunted force-pushed the avery/local-variables branch from 34e4a19 to 84fd545 Compare June 24, 2025 21:36

codeshaunted marked this pull request as draft June 24, 2025 21:46

ogoffart reviewed Jun 25, 2025

View reviewed changes

codeshaunted added 6 commits June 25, 2025 13:49

add unscoped completion for locals

dff5f5d

disable completion for out of scope and undeclared locals

3063032

fix warning in cpp generator

edb0529

use new component syntax in syntax test

0192f40

remove overlapping global name check

06774d9

cleanup pass

83a738e

codeshaunted added 2 commits June 25, 2025 15:39

clippy fixes

febbd07

formatting pass

d1993da

codeshaunted force-pushed the avery/local-variables branch from 436ab75 to d1993da Compare June 25, 2025 19:50

[autofix.ci] apply automated fixes

1160572

codeshaunted marked this pull request as ready for review June 25, 2025 21:36

ogoffart approved these changes Jun 26, 2025

View reviewed changes

ogoffart merged commit d2a5ae0 into slint-ui:master Jun 26, 2025
40 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add local variables, closes #2752 #8740

Add local variables, closes #2752 #8740

Uh oh!

codeshaunted commented Jun 20, 2025

Uh oh!

ogoffart left a comment

Uh oh!

ogoffart Jun 21, 2025

Uh oh!

ogoffart Jun 21, 2025

Uh oh!

ogoffart Jun 21, 2025

Uh oh!

ogoffart Jun 21, 2025

Uh oh!

Uh oh!

Uh oh!

ogoffart left a comment

Uh oh!

ogoffart Jun 25, 2025

Uh oh!

ogoffart Jun 25, 2025

Uh oh!

ogoffart Jun 25, 2025

Uh oh!

ogoffart Jun 25, 2025

Uh oh!

ogoffart Jun 25, 2025

Uh oh!

ogoffart left a comment

Uh oh!

Uh oh!

Uh oh!

Add local variables, closes #2752 #8740

Add local variables, closes #2752 #8740

Uh oh!

Conversation

codeshaunted commented Jun 20, 2025

Uh oh!

ogoffart left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

ogoffart left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ogoffart left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!