一键导入
moonbit-practice
MoonBit code generation best practices. Use when writing MoonBit code to avoid common AI mistakes with syntax, tests, and benchmarks.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
MoonBit code generation best practices. Use when writing MoonBit code to avoid common AI mistakes with syntax, tests, and benchmarks.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
| name | moonbit-practice |
| description | MoonBit code generation best practices. Use when writing MoonBit code to avoid common AI mistakes with syntax, tests, and benchmarks. |
Best practices for AI when generating MoonBit code.
If you don't understand something here, use moonbit-docs skill to search the official documentation.
In MoonBit projects, prefer moon ide commands over Read tool or grep.
# ❌ Avoid: Reading files directly
Read src/parser.mbt
# ✅ Recommended: Find definitions from symbols
moon ide peek-def Parser::parse
moon ide goto-definition -tags 'pub fn' -query 'parse'
# ❌ Avoid: Searching with grep
grep -r "fn parse" .
# ✅ Recommended: Semantic search
moon ide find-references parse
moon ide outline src/parser.mbt
Why:
moon ide provides semantic search (distinguishes definitions from call sites)moon doc quickly reveals APIsmoon doc '<Type>' to explore APIs before implementingmoon.mod / moon.pkg over the deprecated-pending moon.mod.json / moon.pkg.json. moon fmt migrates JSON → DSL. Samples: assets/moon.mod, assets/moon.pkg. Check reference/configuration.md before editing eithermoon.work), not standalone modules. moon work init <dirs> creates the manifest; members share one build context and resolve each other locally. See reference/configuration.md "Workspace".mbtx scripts (moon run script.mbtx, stdin via moon run -, inline import { } block; nightly only)mut is only for reassignment, not field mutation - Array push doesn't need itreturn is unnecessary - last expression is the return valueType:: prefix++ -- not supported - use i = i + 1 or i += 1try needed for error propagation - automatic (unlike Swift)await keyword - just declare with async fnfor i in 0..<n {...}nobreak not else for functional for-loop exit values (else is deprecated)function_name!(...) and function_name(...)? are deprecatedfor { ... } is deprecated - use for ;; { ... } or while true { ... } for infinite loops. syntax for impl removed - . method call only works within the same packagetrait/impl methods now require the fn keyword (v0.10.0) - moon fmt migrates old code; see "Traits and impl" belowtry? is deprecated (v0.10.0) - use Ok(expr) catch { e => Err(e) } to get a Resultfn new(..) constructor removed (v0.10.0) - use a user-defined constructor fn Type::Type(..)immut/array replaced by immut/vector (v0.9) - update imports/deps; better performanceinspect superseded by debug_inspect for snapshot tests (v0.9.2) - Show for container types (Array/Map/Set/Option/Result/tuples) is deprecated; see "Snapshot Tests"///| NG: fn identity[T] is old syntax
fn identity[T](val: T) -> T { val }
///| OK: Type parameter comes right after fn
fn[T] identity(val: T) -> T { val }
///|
/// NG: -> T!Error was removed
fn parse(s: String) -> Int!Error { ... }
///|
/// OK: Use raise keyword
fn parse(s: String) -> Int raise Error { ... }
Int raise is shorthand for Int raise Error.
async fn implicitly raises by default; use noraise to enforce no errors.
///|
/// NG: ! suffix was removed
assert_true!(true)
///|
/// OK
assert_true(true)
fn (v0.10.0)///| NG: method without fn (deprecated, warns then removed)
trait I {
f(Self) -> Unit
}
impl I for Int with f(_) {}
///| OK: fn keyword on both the trait method and the impl
trait I {
fn f(Self) -> Unit
}
impl I for Int with fn f(_) {}
Polymorphic trait methods: the method's own type parameters go after fn,
the impl's own type parameters stay after impl.
trait Logger {
fn[X : Show] write_object(Self, X) -> Unit
}
impl Logger for StringBuilder with fn write_object(self, x) {
self.write_string(x.to_string())
}
impl[A] Poly for Array[A] with fn[X] f(self, x : X) { ... }
// ^^^ impl type params ^^^ method type params
moon fmt rewrites old trait/impl syntax automatically.
The removed fn new(..) special form is replaced by a plain user-defined
constructor fn Type::Type(..). Methods take the receiver as self : Self.
///| Struct with derives
struct Point {
x : Int
y : Int
} derive(Eq, Debug)
///| User-defined constructor (replaces removed `fn new`)
fn Point::Point(x : Int, y : Int) -> Point {
{ x, y } // field-punning struct literal
}
///| Method: receiver `self : Self`, last expression is the return value
fn Point::manhattan(self : Self) -> Int {
self.x.abs() + self.y.abs()
}
let text =
#|line 1
#|line 2
///| is a block separator. /// comments attach to the following ///| block.
///|
/// This function is foo
fn foo() -> Unit { ... }
///|
/// This function is bar
fn bar() -> Unit { ... }
Avoid consecutive ///| at the file beginning as they create separate blocks.
moon test -u auto-fills the empty content="" of inspect(val) /
debug_inspect(val). Which to use (v0.9.2+):
inspect (Show is fine here)Array/Map/Set/Option/Result/tuples) and custom
types → debug_inspect, because Show for these containers is deprecated and
Debug gives structured, indented output. Keep Show for human-facing
formatting only.Choose by the inspected expression's static type, not the runtime value:
Result[Int, _] is a container even when it holds Ok(123), and the rule
propagates — to snapshot/assert a custom type (even nested inside a container) the
custom type must derive(Debug), and equality via @debug.assert_eq additionally
needs derive(Eq).
test "snapshot" {
inspect("hello", content="") // primitive/string -> inspect
debug_inspect([1, 2, 3], content="") // container -> debug_inspect
}
///| Custom type: derive Debug to snapshot, derive Eq to assert_eq
struct P {
x : Int
} derive(Eq, Debug)
test "custom type" {
debug_inspect(P::{ x: 1 }, content="")
@debug.assert_eq(P::{ x: 1 }, P::{ x: 1 })
}
After moon test -u:
test "snapshot" {
inspect("hello", content="hello")
debug_inspect([1, 2, 3], content="[1, 2, 3]")
}
Available in .mbt.md files or ///| inline comments.
| Code Block | Behavior |
|---|---|
```mbt check | Checked by LSP |
```mbt test | Executed as test {...} |
```moonbit | Display only (not executed) |
Example (inline comment):
///|
/// Increment an integer by 1
/// ```mbt test
/// inspect(incr(41), content="42")
/// ```
pub fn incr(x : Int) -> Int {
x + 1
}
Run before releasing:
moon fmt # Format code
moon info # Generate type definition files
pkg.generated.mbti is auto-generated by moon info. Don't edit it directly.
moon doc StringView # StringView methods
moon doc Array # Array methods
moon doc Map # Map methods
| Topic | Command | Details |
|---|---|---|
| Test | moon test | https://docs.moonbitlang.com/en/stable/language/tests |
| Update snapshots | moon test -u | Same as above |
| Filtered test | moon test --filter 'glob' | Run specific tests |
| Benchmark | moon bench | https://docs.moonbitlang.com/en/stable/language/benchmarks |
| Doc Test | moon check / moon test | https://docs.moonbitlang.com/en/stable/language/docs |
| Format | moon fmt | - |
| Generate types | moon info | - |
| Doc reference | moon doc <Type> | - |
| Workspace init | moon work init | See reference/configuration.md |
| Workspace add | moon work use mod1 mod2 | Add modules to workspace |
| API usage analysis | moon ide analyze . | Show public API usage stats |
More accurate than grep for code navigation. See reference/ide.md for details.
# Show symbol definition
moon ide peek-def Parser::read_u32_leb128
# Package outline
moon ide outline .
# Find references
moon ide find-references TranslationUnit
# Jump to type definition (with location)
moon ide peek-def Parser -loc src/parse.mbt:46:4
# Show type signature and docs at location (unlike peek-def, shows inferred type + doc comments)
moon ide hover my_func --loc src/lib.mbt:10:4
# Rename symbol across the project
moon ide rename old_name new_name
# Analyze public API usage (v0.8.3+)
moon ide analyze . # Current package
moon ide analyze internal/* # Glob pattern
Prefer functional for loops whenever possible. More readable and easier to reason about.
// Functional for loop with state
for i = 0, sum = 0; i <= 10 {
continue i + 1, sum + i // Update state
} nobreak {
sum // Value at loop exit (nobreak, not else)
}
// Range for (recommended)
for i in 0..<n { ... }
for i, v in array { ... } // index and value
// Range for with extra loop variable (v0.8.3+)
for x in xs; sum = 0 {
continue sum + x
} nobreak {
sum
}
// Infinite loop (for { } is deprecated)
for ;; { ... }
[for .. => body] builds an array; add if <guard> before => to filter.
The same shape typed as Iter[T] yields a lazy iterator instead of an array.
let evens = [for i in 0..<100 if i % 2 == 0 => i]
///| With loop state (v0.10.0): carries p1/p2 across iterations
let fibs = [
for _ in 0..<10
p1 = 1, p2 = 0
p1 = p1 + p2, p2 = p1 => p1
]
///| Typed as Iter -> lazy
let fib_iter : Iter[Int] = [for p1 = 1, p2 = 0;; p1 = p1 + p2, p2 = p1 => p1]
<+ (v0.10.0)buf <+ expr appends to a StringBuilder — terser than write_string, pairs
well with $| interpolation for building markup/strings.
let buf = StringBuilder()
buf <+ "<ul>"
for item in items {
buf <+ $|<li>\{item}</li>
}
buf <+ "</ul>"
buf.to_string()
=~ (v0.9+)Stable regex matching (s =~ re"...") replaces the experimental lexmatch?,
with different semantics — check moon doc / docs before relying on capture,
prefix (before~), or suffix (after~) binding forms.
let matched = s =~ re"abc"
const supports string concatenation and interpolation (v0.8.3+):
const Hello : String = "Hello"
const HelloWorld : String = Hello + " world"
const Message : String =
$|========
$|\{HelloWorld}
$|========
MoonBit uses checked errors. See reference/ffi.md for details.
///| Declare error type
suberror ParseError {
InvalidEof
InvalidChar(Char)
}
///| Declare with raise, auto-propagates
fn parse(s: String) -> Int raise ParseError {
if s.is_empty() { raise ParseError::InvalidEof }
...
}
///| Convert to Result — `try?` is deprecated (v0.10.0)
let result : Result[Int, ParseError] = Ok(parse(s)) catch { e => Err(e) }
///| Single expression with a fallback: omit `try`
let n = parse(s) catch { _ => 0 }
///| Full form: `noraise` branch handles the success value
let n = try parse(s) catch {
ParseError::InvalidEof => -1
_ => 0
} noraise {
v => v
}
Prefer the official curl installer (assets/ci.yaml) or Nix with moonbit-overlay (assets/ci-nix.yaml) over third-party actions such as hustcer/setup-moonbit@v1. The installer is short and adds nothing to the action supply chain; the Nix path is fully reproducible when a flake.nix is present.
moon update is mandatoryRunners start with an empty registry index. Any registry-hosted MoonBit dependency makes moon check / moon test / moon build fail with Failed to resolve registry dependency until you run moon update. Put it immediately after installing the CLI in every workflow that touches mooncakes.
Some projects (e.g. Vite plugins, language tooling) ship as npm packages but run moon inside pnpm build — for instance pnpm build:parser that calls moon -C tools/parser build --release --target js. In that case the npm publish workflow needs both the MoonBit CLI and moon update before pnpm build, otherwise the build fails on CI.
See assets/publish-to-npm.yaml for a minimal release-triggered publish workflow that uses OIDC Trusted Publishing (no NPM_TOKEN) and sets up MoonBit correctly. Pair it with a release automation tool (see the npm-release skill for a release-please + OIDC setup) to avoid manual npm publish calls.
assets/moon.mod — module config sample in the new DSL (replaces moon.mod.json)assets/moon.pkg — package config sample in the new DSL (replaces moon.pkg.json)assets/ci.yaml — GitHub Actions CI (curl installer)assets/ci-nix.yaml — GitHub Actions CI with Nix (moonbit-overlay)assets/publish-to-npm.yaml — release-triggered npm publish with MoonBit build and OIDC Trusted Publishingmoonbit-community/moonbit-overlay provides a Nix flake overlay for reproducible MoonBit builds.
Minimal flake.nix for a MoonBit project:
{
inputs = {
nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
moonbit-overlay.url = "github:moonbit-community/moonbit-overlay";
moon-registry = {
url = "git+https://mooncakes.io/git/index";
flake = false;
};
};
outputs = { nixpkgs, moonbit-overlay, moon-registry, ... }:
let
system = "x86_64-linux"; # or aarch64-darwin, etc.
pkgs = import nixpkgs {
inherit system;
overlays = [ moonbit-overlay.overlays.default ];
};
moonHome = pkgs.moonPlatform.bundleWithRegistry {
cachedRegistry = pkgs.moonPlatform.buildCachedRegistry {
moonModJson = ./moon.mod.json;
registryIndexSrc = moon-registry;
};
};
in {
devShells.${system}.default = pkgs.mkShellNoCC {
packages = [ moonHome pkgs.git ];
env.MOON_HOME = "${moonHome}";
};
};
}
Key APIs from the overlay:
pkgs.moonPlatform.buildMoonPackage - Build a MoonBit package as a Nix derivationpkgs.moonPlatform.bundleWithRegistry - Create a MOON_HOME with cached registrypkgs.moonPlatform.buildCachedRegistry - Pre-fetch mooncakes registry dependenciesReview screenshots or other images with OpenRouter vision models via bundled Deno scripts. Use for quick VRT sanity checks, invalid-image screening, or CI gates. `scripts/review-image.ts` returns freeform feedback; `scripts/review-image-ci.ts` returns strict `pass|fail` JSON and exits non-zero on fail.
Entry-point for the `@mizchi/vlmkit` toolkit — VLM-driven frontend kit covering visual regression (snapshot / diff / regression-watch), markup synthesis from screenshots, design-token / theme / a11y / i18n audits, and a 2-stage VLM + LLM CSS auto-repair loop. Use when a coding agent has edited HTML/CSS and needs to know whether the visible output changed, where it changed, and which CSS properties drove the change — or when the task is markup-from-image / token / theme audit / fix-loop driven by VLM. This skill orients you to the 5 detailed sub-skills (vrt-visual-diff / vrt-migration-eval / vrt-markup-synth / vrt-regression-watch / vrt-css-fix-loop) and the verb-group CLI; pick the matching sub-skill once the task shape is clear.
ECS blue/green deployment patterns. Recommends ALB-native weighted target groups as the simpler default; covers CodeDeploy as a reference for teams that already use it or need pipeline-integrated rollback. Use when choosing or debugging ECS blue/green infrastructure.
Use when ECS Service Connect の DNS alias が IPv6 アドレスを返し、 IPv4-only の Fargate task から `network is unreachable` / `EAI_AGAIN` 等で接続できない問題に遭遇したとき。 OTel Collector → Tempo の OTLP gRPC、 Fargate → Service Connect の HTTP/gRPC 通信が突然失敗する症状が典型。 ECS / Fargate / Service Connect / OTLP / IPv6 dual-stack 関連の接続障害から起動して良い (user が原因を IPv6 と特定していなくても OK)。
OpenTofu/Terraform pattern for GitHub Actions OIDC trust with AWS IAM. Covers the non-obvious `job_workflow_ref` condition (vs just `sub` for repo+branch), the Bedrock inference profile ARN patterns, required `aws-marketplace` permissions alongside Bedrock, and the ReadOnlyAccess + explicit Deny pattern for AI agent roles. Use when wiring GitHub Actions to AWS via OIDC.
Use when AWS account の IAM API が MFA 必須 policy で弾かれて、 `sts:GetCallerIdentity` は通るのに `iam:*` が `InvalidClientTokenId` で拒否されるとき、 または FIDO2 passkey で MFA が CLI から使えないとき。 aws-vault / aws-cli / MFA / passkey 関連で「アクセス拒否」 「token expired」 「IAM 操作のみ失敗」 等の症状から起動 (user が aws-vault を名指しでなくても OK)。