feat: implement /say, /me, and /msg chat commands by nglmercer · Pull Request #302 · Steel-Foundation/SteelMC

nglmercer · 2026-07-30T22:39:12Z

Summary

Implements the vanilla chat messaging commands:

/say <message> — operator-only, broadcasts to all players and logs to the server console.
/me <message> — available to all players, broadcasts an emote.
/msg <targets> <message> — available to all players, with /tell and /w aliases. Sends an incoming whisper to each target and an outgoing echo back to the sender.

Command-generated chat is delivered as unsigned disguised chat (CDisguisedChat) bound to the matching vanilla chat type (say_command, emote_command, msg_command_incoming, msg_command_outgoing), so the client applies the standard decorations ([%s] %s, * %s %s, %s whispers to you: %s, You whisper to %s: %s).

Changes

steel-core/src/command/builtins/message.rs — new module with the three commands and graph-shape tests.
steel-core/src/command/execution/text.rs — adds CommandTextResolver::for_source for resolving a message once against the command source.
steel-core/src/command/execution/source.rs — removes the now-used dead_code expectation on CommandSource::sender().
steel-core/src/command/builtins/mod.rs — registers the commands and extends the command graph-shape test.

Verification

cargo check -p steel-core — pass
cargo fmt --all --check — clean
cargo clippy -p steel-core --all-targets --all-features — 0 warnings
cargo test -p steel-core --lib command::builtins — 72 passed, including 3 new tests and the updated graph-shape test

Note on vanilla verification

The chat-type bindings and decoration parameters were taken from the repository's generated data (vanilla_chat_types.rs and the generated translations) rather than transcribed from memory. minecraft-src/ was not generated locally, so the /msg outgoing-echo binding follows stable vanilla behavior corroborated by the msg_command_outgoing parameter data ([target, content]); it can be confirmed against MessageCommand.java once the vanilla source is generated.

Add the vanilla chat messaging commands /say, /me, and /msg (with /tell and /w aliases). Command-generated chat is delivered as unsigned disguised chat bound to the matching vanilla chat type (say_command, emote_command, msg_command_incoming/outgoing), so the client applies the usual decorations. /say is operator-only and broadcasts server-wide; /me and /msg are available to all players. Also adds a CommandTextResolver::for_source helper for resolving a message once against the command source.

nglmercer · 2026-07-30T23:24:23Z

no existing handling (or test) for multi-word component arguments

Plain multi-word input is now consumed whole as literal text instead of failing on the first space, so /msg, /say, and /me accept real sentences. Input that opens a structured SNBT component (a leading {, [, or quote) still parses through the SNBT path unchanged, which also lets /tellraw accept plain text. Selector and score resolution are unaffected because those only come from structured components.

nglmercer · 2026-07-31T00:20:42Z

is added a new multi-word component arguments starts_structured_component and parse_structured_component.

coco875 · 2026-08-02T14:44:33Z

        reader.restore(start);
        invalid_component(reader, error.to_string())
-    })?;
-    validate_component_syntax(&component).map_err(|error| {


why have remove the validate ?

"parse_component was not removed and remains unchanged. Since I introduced parse_structured_component, we can add validate_component_syntax there instead. What do you think?"

hmm ok not really familiar with it, will try do a more deep review later

@coco875 did you had time for the deeper review?

not really an expert in command but can try

same for me

@hthug06 if you have time and want to help, I know you did commands, so I think you are more expert than us maybe you can find something

Add validation for component syntax and error handling.

Removed redundant validation of component syntax.

JunkyDeveloper · 2026-08-05T13:32:03Z

Please fix lint

nglmercer · 2026-08-05T15:16:14Z

Please fix lint

ready

Laptop59 · 2026-08-05T16:19:05Z

+fn say_command() -> CommandNodeBuilder<CommandSource, SteelCommandRuntime> {
+    literal("say").then(argument("message", SteelArgumentType::component()).executes(say))
+}
+
+fn me_command() -> CommandNodeBuilder<CommandSource, SteelCommandRuntime> {
+    literal("me").then(argument("message", SteelArgumentType::component()).executes(me))
+}
+
+fn msg_command() -> CommandNodeBuilder<CommandSource, SteelCommandRuntime> {
+    literal("msg").then(
+        argument("targets", SteelArgumentType::players())
+            .then(argument("message", SteelArgumentType::component()).executes(msg)),
+    )
+}


I'm pretty sure the wrong argument type is being used. It should be some kind of MessageArgument (from Vanilla), not a text component.

I'm pretty sure the wrong argument type is being used. It should be some kind of MessageArgument (from Vanilla), not a text component.

ready new 'message' in SteelArgumentType

Introduces `SteelArgumentType::message()` and `MessageParser` that parse the entire remaining input as plain text with vanilla's 256-character limit, mirroring `MessageArgument.parseText`. The `say`, `me`, and `msg` commands now use this instead of the structured component parser, which matches vanilla behavior where these commands accept raw text rather than JSON components.

The spawned domain-switch storage tasks run on the test's current-thread runtime, so the loop yields to the runtime each tick to drive their async I/O. A periodic sleep lets the runtime process I/O completions; without it, slow disk writes under parallel test load could exhaust the tick budget.

Laptop59

I actually tested this PR, and I found out that when trying to run /say, /tell, or /me as a player in the server, nothing seems to happen. This is because the server does not currently know how to handle the packet SChatCommandSigned. The packet's ReadFrom implementation is already there, so it just needs to be handled when caught, which is currently not done in SteelMC, as indicated by

[Info] play packet id 8 is not known

Seems like on the server terminal, the commands work as expected, except for the selectors resolving to entity names.

Laptop59 · 2026-08-06T12:04:13Z

+/// Parses a vanilla message argument: the entire remaining input as plain text.
+///
+/// Mirrors `MessageArgument.parseText`, including the 256-character length limit.
+fn parse_message(reader: &mut StringReader<'_>) -> Result<TextComponent, CommandSyntaxError> {
+    let length = reader.remaining().len();
+    if length > 256 {
+        let message = translations::ARGUMENT_MESSAGE_TOO_LONG
+            .message([length.to_string(), "256".to_owned()])
+            .component();
+        return Err(reader.error(CommandSyntaxErrorKind::Dynamic(Box::new(message))));
+    }
+    Ok(TextComponent::plain(reader.read_remaining().to_owned()))
+}


Vanilla also resolves selectors to the names of the entities referenced by them, separated by commas, if the sender is allowed to use them.
The current code for parsing here does not do this and simply just reads the remaining string.

Laptop59 · 2026-08-06T12:14:25Z

+/// Advances chunk scheduling and polls server jobs until the queue empties.
+///
+/// The spawned domain-switch storage tasks run on the test's current-thread
+/// runtime, so the loop yields to the runtime each tick to drive their async
+/// I/O. A periodic sleep lets the runtime process I/O completions; without it,
+/// slow disk writes under parallel test load could exhaust the tick budget.
+async fn tick_until_jobs_finish(server: &Arc<Server>, worlds: &[Arc<World>]) {
+    for tick in 1..=1_000_000u64 {
+        for world in worlds {
+            world.chunk_map.advance_scheduling();
+        }
+        server.tick_jobs(tick, true);
+        if server.jobs.is_empty() {
+            return;
+        }
+        if tick.is_multiple_of(100) {
+            sleep(Duration::from_millis(1)).await;
+        } else {
+            yield_now().await;
+        }
+    }
+    panic!("server jobs did not finish within 1_000_000 ticks");
+}
+


Why do a chunk scheduling change in a command PR? Isn't it out of scope?

it isn't chunk scheduling only test, but also don't see much benefit

coco875 · 2026-08-06T12:31:31Z

@Laptop59 thanks for the help to review the pr

github-actions · 2026-08-06T14:37:02Z

This pull request has conflicts with the base branch "master". Please resolve those so we can test out your changes.

JunkyDeveloper · 2026-08-06T19:46:26Z

+/// Advances chunk scheduling and polls server jobs until the queue empties.
+///
+/// The spawned domain-switch storage tasks run on the test's current-thread
+/// runtime, so the loop yields to the runtime each tick to drive their async
+/// I/O. A periodic sleep lets the runtime process I/O completions; without it,
+/// slow disk writes under parallel test load could exhaust the tick budget.
+async fn tick_until_jobs_finish(server: &Arc<Server>, worlds: &[Arc<World>]) {
+    for tick in 1..=1_000_000u64 {
+        for world in worlds {
+            world.chunk_map.advance_scheduling();
+        }
+        server.tick_jobs(tick, true);
+        if server.jobs.is_empty() {
+            return;
+        }
+        if tick.is_multiple_of(100) {
+            sleep(Duration::from_millis(1)).await;
+        } else {
+            yield_now().await;
+        }
+    }
+    panic!("server jobs did not finish within 1_000_000 ticks");
+}
+


it isn't chunk scheduling only test, but also don't see much benefit

… import conflict

github-actions · 2026-08-08T02:57:49Z

Conflicts have been resolved! 🎉

@x

MessageValue now stores raw text and parsed selector parts, resolving them via CommandTextResolutionSource when the message is delivered. This mirrors vanilla's MessageArgument behavior where selectors are only resolved against the sender at delivery time. A new try_parse_message_selector handles message args specifically, treating invalid selector types (e.g. @x) as literal text instead of errors. Inline the tick_until_jobs_finish test helper.

Message arguments now scan for `@a`/`@p`/`@r`/`@s`/`@e` selectors during parsing when the source allows them, matching vanilla `MessageArgument.parseText` behavior. Selectors are stored as parts on `MessageValue` and resolved to entity names only at delivery time.

Replace nonexistent steel_registry::test_support::init_test_registry with the actual steel_registry::init_vanilla_registry. Suppress dead_code on CommandTextResolver::for_source which is kept as a utility API.

… passes clean:

coco875 requested changes Aug 2, 2026

View reviewed changes

nglmercer and others added 3 commits August 2, 2026 11:21

Enhance parse_structured_component with syntax validation

38e5af2

Add validation for component syntax and error handling.

Remove duplicate syntax validation for components

5e58109

Removed redundant validation of component syntax.

Merge branch 'master' into feat/message-commands

9a1a941

nglmercer added 2 commits August 5, 2026 09:36

update argument.rs

3515a05

Update argument.rs

bec525a

coco875 previously approved these changes Aug 5, 2026

View reviewed changes

Laptop59 suggested changes Aug 5, 2026

View reviewed changes

Update message argument type to use SteelArgumentType::message

4b3ee48

nglmercer dismissed coco875’s stale review via 4b3ee48 August 5, 2026 16:41

nglmercer added 5 commits August 5, 2026 11:45

Refactor message argument type assertions

02aaad2

Use component argument type for message commands

3413dfd

format code

361b3a6

nglmercer requested a review from Laptop59 August 5, 2026 17:49

Laptop59 reviewed Aug 6, 2026

View reviewed changes

github-actions Bot added the Merge Conflicts label Aug 6, 2026

JunkyDeveloper requested changes Aug 6, 2026

View reviewed changes

Merge upstream/master into feat/message-commands, resolve argument.rs…

2ba6721

… import conflict

github-actions Bot removed the Merge Conflicts label Aug 8, 2026

nglmercer added 6 commits August 7, 2026 22:12

Remove stored EntitySelector from MessagePart

15e5875

Apply rustfmt

dd61aa3

Fix CI: use init_vanilla_registry and suppress dead_code warning

7133372

Replace nonexistent steel_registry::test_support::init_test_registry with the actual steel_registry::init_vanilla_registry. Suppress dead_code on CommandTextResolver::for_source which is kept as a utility API.

All 5 clippy errors fixed — cargo clippy --all-targets -- -D warnings…

3cb8758

… passes clean:

nglmercer requested review from JunkyDeveloper and Laptop59 August 8, 2026 04:04

Uh oh!

Conversation

nglmercer commented Jul 30, 2026

Summary

Changes

Verification

Note on vanilla verification

Uh oh!

nglmercer commented Jul 30, 2026

Uh oh!

nglmercer commented Jul 31, 2026

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!

JunkyDeveloper commented Aug 5, 2026

Uh oh!

nglmercer commented Aug 5, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nglmercer Aug 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Laptop59 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!

coco875 commented Aug 6, 2026

Uh oh!

github-actions Bot commented Aug 6, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Aug 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

nglmercer Aug 5, 2026 •

edited

Loading