Add ResourceContentHash annotation to generated resource accessors by pjBooms · Pull Request #5357 · JetBrains/compose-multiplatform

pjBooms · 2025-07-04T14:20:58Z

Used by Compose Hot Reload to mark a related accessor as dirty if the hash is changed after reloading of new accessors.

Fixes CMP-8513

Release Notes

N/A

sellmair

Looks really promising to me!
Here are two concerns that I have:

I am curious about the feature flag; it seems to bring additional complexity and integration problems. I assume that the argument is binary size, but I fail to see how a few bytes compared to the actual resource would make a difference.
I am little concerned about the use of hash functions and the 32-bit hashes potentially colliding. I would be more relaxed if we could use 64bit and proper hashes (but thsi can be done later)

terrakok

LGTM but I'd like to change tests a bit

terrakok · 2025-08-18T12:15:46Z

+                accessorBuilder.addAnnotation(
+                    AnnotationSpec.builder(resourceContentHashAnnotationClass)
+                    .useSiteTarget(AnnotationSpec.UseSiteTarget.DELEGATE)
+                    .addMember("%L", items.fold(0){acc, item -> ((acc * 31) + item.contentHash)}).build()


fix indentions, please

terrakok · 2025-08-18T12:17:37Z

        val type = ResourceType.fromString(typeString) ?: error("Unknown resource type: '$typeString'.")
-        return listOf(ResourceItem(type, qualifiers, file.nameWithoutExtension.asUnderscoredIdentifier(), path))
+        return listOf(ResourceItem(type, qualifiers, file.nameWithoutExtension.asUnderscoredIdentifier(), path,
+            file.readBytes().contentHashCode()))


let's put every parameter on a new line?

terrakok · 2025-08-18T12:17:43Z

        val record = ValueResourceRecord.createFromString(recordString)
-        return ResourceItem(record.type, qualifiers, record.key.asUnderscoredIdentifier(), path, offset, size)
+        return ResourceItem(record.type, qualifiers, record.key.asUnderscoredIdentifier(), path,
+            record.content.hashCode(), offset, size)


here as well

terrakok · 2025-08-18T12:21:08Z

    }

+    @Test
+    fun testGeneratedAccessorsAnnotatedWithResourceContentHash(): Unit = with(testProject("misc/commonResourcesAnnotatedWithResourceContentHash")) {


Instead of the new test project, we discussed using an existed one (commonResources). You just need to add new expected directory: https://github.com/JetBrains/compose-multiplatform/tree/master/gradle-plugins/compose/src/test/test-projects/misc/commonResources

Used by Compose Hot Reload to mark a related accessor as dirty if the hash is changed after reloading of new accessors.

…em property is set for the Gradle task.

terrakok · 2025-08-18T17:09:43Z

+@delegate:ResourceContentHash(1_620_038_668)
+internal val Res.drawable._3_strange_name: DrawableResource by lazy {
+      DrawableResource("drawable:_3_strange_name", setOf(
+        ResourceItem(setOf(), "${MD}drawable/3-strange-name.xml", -1, -1),
+      ))
+    }
+
+@delegate:ResourceContentHash(1_620_038_668)
+internal val Res.drawable.camelCaseName: DrawableResource by lazy {
+      DrawableResource("drawable:camelCaseName", setOf(
+        ResourceItem(setOf(), "${MD}drawable/camelCaseName.xml", -1, -1),
+      ))
+    }
+
+@delegate:ResourceContentHash(1_620_038_668)


Is it expected that hashes are equal?

As underlying files seem to be the same by content then yes, the hashes will be equal.

…rceContentHash

terrakok · 2025-08-19T15:22:02Z

+        return if (extension.lowercase() == "xml") {
+            // Text files have different line endings on Windows/*nix,
+            // so let's ignore them in the hashcode calculation
+            Files.readAllLines(toPath()).joinToString().hashCode()
+        } else {


sorry, I don't understand it. how does it affect our usecase?

Tests on Windows fail without this because when xml files are checked out they are not equal by content with the same files on Linux/Mac (because of different line endings).

the same should happen for any text files, right? I think a correct way is to fix tests instead of hacks here

terrakok

Could you check how the build time is being changed?
I see, we count hashes even without the flag.
A project for the check: https://github.com/terrakok/fluentui-system-icons

to run the demo ./gradlew :demo:run

pjBooms · 2025-08-20T09:39:17Z

Could you check how the build time is being changed?

No differences noted: 1.9.0-beta03 and this version both build around 37s on my machine after ./gradlew clean

…etBrains#5357) Used by Compose Hot Reload to mark a related accessor as dirty if the hash is changed after reloading of new accessors. Fixes [CMP-8513](https://youtrack.jetbrains.com/issue/CMP-8513) ## Release Notes N/A

pjBooms requested a review from terrakok July 4, 2025 14:20

pjBooms force-pushed the pjBooms/resource-accessor-annotations branch from 209913f to d8f7123 Compare July 4, 2025 16:20

pjBooms force-pushed the pjBooms/resource-accessor-annotations branch from d8f7123 to 700d707 Compare August 6, 2025 16:00

pjBooms requested a review from sellmair August 6, 2025 16:03

pjBooms force-pushed the pjBooms/resource-accessor-annotations branch from 700d707 to 8bbf398 Compare August 7, 2025 10:06

sellmair approved these changes Aug 7, 2025

View reviewed changes

Comment thread ...sources/library/src/commonMain/kotlin/org/jetbrains/compose/resources/ResourceContentHash.kt

Comment thread gradle-plugins/compose/src/main/kotlin/org/jetbrains/compose/resources/GeneratedResClassSpec.kt

terrakok requested changes Aug 18, 2025

View reviewed changes

pjBooms added 3 commits August 18, 2025 17:21

Add ResourceContentHash annotation to generated resource accessors.

d506f5c

Used by Compose Hot Reload to mark a related accessor as dirty if the hash is changed after reloading of new accessors.

Generate ResourceContentHash annotation only when the respective syst…

3dec5d3

…em property is set for the Gradle task.

Review fixes

227d3ba

pjBooms force-pushed the pjBooms/resource-accessor-annotations branch 2 times, most recently from a243779 to ffd1d57 Compare August 18, 2025 16:38

terrakok reviewed Aug 18, 2025

View reviewed changes

Use misc/CommonResources for testGeneratedAccessorsAnnotatedWithResou…

81524ae

…rceContentHash

pjBooms force-pushed the pjBooms/resource-accessor-annotations branch from ffd1d57 to 81524ae Compare August 18, 2025 17:16

sellmair approved these changes Aug 19, 2025

View reviewed changes

terrakok reviewed Aug 19, 2025

View reviewed changes

Handle Windows in the test

1892dba

pjBooms force-pushed the pjBooms/resource-accessor-annotations branch from b74756c to 1892dba Compare August 20, 2025 08:49

terrakok approved these changes Aug 20, 2025

View reviewed changes

pjBooms merged commit ef3ba2e into master Aug 20, 2025
13 checks passed

pjBooms deleted the pjBooms/resource-accessor-annotations branch August 20, 2025 15:35

Conversation

pjBooms commented Jul 4, 2025

Release Notes

Uh oh!

sellmair left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

terrakok 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!

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!

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!

terrakok left a comment

Choose a reason for hiding this comment

Uh oh!

pjBooms commented Aug 20, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants