Mizan-Rust backend adapter: server-side substrate + three-way parity

Adds first-class Rust-backed Mizan to sit alongside mizan-django and
mizan-fastapi. A Rust dev writes:

    #[derive(Mizan, Serialize, Deserialize)]
    pub struct ProfileOutput { pub user_id: i64, pub name: String }

    #[mizan::context("user")]
    pub struct UserCtx;

    #[mizan::client(context = UserCtx)]
    pub async fn user_profile(_req: &RequestHandle<'_>, user_id: i64) -> ProfileOutput { ... }

…and gets byte-identical KDL to the Python emitters, served over the
same wire protocol the React / Rust / Vue / Svelte kernels speak.

New crates:
- cores/mizan-rust/         (Cargo: mizan-core)     — IR types, KDL emitter, traits, registry,
                                                       runtime (compute_invalidation / compute_merges
                                                       ported from mizan-fastapi), graph_check with
                                                       structural type-matching
- cores/mizan-rust-macros/  (Cargo: mizan-macros)   — #[derive(Mizan)], #[mizan::context],
                                                       #[mizan::client] proc macros
- backends/mizan-rust-axum/ (Cargo: mizan-axum)     — axum HTTP adapter: /session/, /call/, /ctx/:name/
- tests/afi/rust_app/                                — AFI fixture port + server / export-ir binaries

Substrate-shape moves required by cross-language equivalence:
- IR canonicalization: functions / contexts / context-members / shared-by
  now sort alphabetically in both Python and Rust emitters. The IR is a
  contract; linkme doesn't preserve declaration order, so canonical sort
  is the only stable mapping. afi_ir.kdl + per-target baselines regenerated.
- MizanType::TYPE_NAME is a const (with a default type_name() reader) so
  it's usable in linkme TypeEntry static initializers.
- Tree-shaken type registry: #[derive(Mizan)] only emits the trait impl;
  the #[mizan::client] macro registers canonical-named entries from
  fn signatures, including Vec<T> element types for ref resolution.
- Merge resolution is structural (NamedType shape comparison) rather than
  by name — matches the Python types_match_for_merge semantics.

Three-way forcing functions:
- tests/afi/test_codegen_parity.py — Django ≡ FastAPI ≡ Rust on KDL bytes (3 pass)
- tests/rust/run_wire_parity.py    — 12/12 probes against FastAPI + Rust (EXIT=0)

Incidental fixes surfaced by the new tests:
- Stale `from .registry import validate_registry` import removed from
  mizan-django/setup/discovery.py (referenced a function that no longer
  exists; was masking codegen-parity).
- BASE_DIR added to tests/afi/django_app/project/settings.py.
- /session/ endpoint added to mizan-fastapi for protocol-shaped readiness
  probe parity (wire-parity harness now polls /api/mizan/session/ on both
  backends rather than FastAPI's /openapi.json).
- Root .gitignore picks up Rust target/ across the tree so new crates
  don't need per-crate gitignore.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-17 22:31:26 -04:00
parent 9900f8a36f
commit 45bde51166
47 changed files with 4187 additions and 147 deletions

397
cores/mizan-rust/src/kdl.rs Normal file
View File

@@ -0,0 +1,397 @@
//! KDL emitter — byte-equivalent to `cores/mizan-python/src/mizan_core/ir.py`.
//!
//! The Python emitter is the spec; this is the second implementation under
//! the same contract. Any divergence is a bug here, not a contract change.
use crate::ir::{DefaultValue, NamedType, Primitive, StructField, TypeShape};
use crate::registry::{CONTEXTS, FUNCTIONS, TYPES};
use crate::traits::FunctionSpec;
use std::collections::BTreeMap;
const INDENT: &str = " ";
/// Escape a string for KDL — same escape set as the Python emitter.
fn kdl_string(s: &str) -> String {
let mut out = String::with_capacity(s.len() + 2);
out.push('"');
for c in s.chars() {
match c {
'\\' => out.push_str("\\\\"),
'"' => out.push_str("\\\""),
'\n' => out.push_str("\\n"),
'\r' => out.push_str("\\r"),
'\t' => out.push_str("\\t"),
other => out.push(other),
}
}
out.push('"');
out
}
fn kdl_bool(b: bool) -> &'static str {
if b {
"#true"
} else {
"#false"
}
}
fn kdl_default(v: &DefaultValue) -> String {
match v {
DefaultValue::Null => "#null".into(),
DefaultValue::Boolean(b) => kdl_bool(*b).into(),
DefaultValue::Integer(i) => i.to_string(),
DefaultValue::Number(f) => {
// Match Python's `repr(float)` for whole-number-equal-but-float
// values: e.g. 1.0 → "1.0", not "1".
if f.fract() == 0.0 && f.is_finite() {
format!("{f:.1}")
} else {
f.to_string()
}
}
DefaultValue::String(s) => kdl_string(s),
}
}
/// Convert snake_case to camelCase. Matches Python's `_snake_to_camel`.
pub fn snake_to_camel(name: &str) -> String {
let normalized = name.replace('.', "_").replace('-', "_");
let mut parts = normalized.split('_');
let mut out = String::new();
if let Some(first) = parts.next() {
out.push_str(first);
}
for part in parts {
if part.is_empty() {
continue;
}
let mut chars = part.chars();
if let Some(c) = chars.next() {
out.extend(c.to_uppercase());
out.push_str(chars.as_str());
}
}
out
}
struct Emitter {
lines: Vec<String>,
}
impl Emitter {
fn new() -> Self {
Self { lines: Vec::new() }
}
fn prefix(&self, indent: usize) -> String {
INDENT.repeat(indent)
}
fn leaf(&mut self, indent: usize, parts: &[&str]) {
let mut line = self.prefix(indent);
line.push_str(&parts.join(" "));
self.lines.push(line);
}
fn open(&mut self, indent: usize, parts: &[&str]) {
let mut line = self.prefix(indent);
line.push_str(&parts.join(" "));
line.push_str(" {");
self.lines.push(line);
}
fn close(&mut self, indent: usize) {
let mut line = self.prefix(indent);
line.push('}');
self.lines.push(line);
}
fn blank(&mut self) {
self.lines.push(String::new());
}
fn emit_type_child(&mut self, indent: usize, shape: &TypeShape) {
match shape {
TypeShape::Primitive(p) => {
let name = kdl_string(p.name());
self.leaf(indent, &["primitive", &name]);
}
TypeShape::Ref(name) => {
let n = kdl_string(name);
self.leaf(indent, &["ref", &n]);
}
TypeShape::List(inner) => {
self.open(indent, &["list"]);
self.emit_type_child(indent + 1, inner);
self.close(indent);
}
TypeShape::Optional(inner) => {
self.open(indent, &["optional"]);
self.emit_type_child(indent + 1, inner);
self.close(indent);
}
TypeShape::Enum(variants) => {
let mut parts: Vec<String> = vec!["enum".into()];
for v in variants {
parts.push(kdl_string(v));
}
let line: Vec<&str> = parts.iter().map(String::as_str).collect();
self.leaf(indent, &line);
}
TypeShape::Union(branches) => {
self.open(indent, &["union"]);
for b in branches {
self.emit_type_child(indent + 1, b);
}
self.close(indent);
}
}
}
fn emit_named_type(&mut self, indent: usize, name: &str, body: &NamedType) {
let name_lit = kdl_string(name);
self.open(indent, &["type", &name_lit]);
match body {
NamedType::Struct(fields) => {
self.open(indent + 1, &["struct"]);
for field in fields {
self.emit_struct_field(indent + 2, field);
}
self.close(indent + 1);
}
NamedType::Alias(inner) => {
self.open(indent + 1, &["alias"]);
self.emit_type_child(indent + 2, inner);
self.close(indent + 1);
}
NamedType::Enum(variants) => {
let mut parts: Vec<String> = vec!["enum".into()];
for v in variants {
parts.push(kdl_string(v));
}
let line: Vec<&str> = parts.iter().map(String::as_str).collect();
self.leaf(indent + 1, &line);
}
}
self.close(indent);
}
fn emit_struct_field(&mut self, indent: usize, field: &StructField) {
let name = kdl_string(field.name);
let mut header: Vec<String> = vec!["field".into(), name];
if !field.required {
header.push(format!("required={}", kdl_bool(false)));
if let Some(default) = &field.default {
header.push(format!("default={}", kdl_default(default)));
}
}
let line_parts: Vec<&str> = header.iter().map(String::as_str).collect();
self.open(indent, &line_parts);
self.emit_type_child(indent + 1, &field.shape);
self.close(indent);
}
fn emit_function(&mut self, indent: usize, fn_spec: &dyn FunctionSpec) {
let name = kdl_string(fn_spec.name());
self.open(indent, &["function", &name]);
let camel = kdl_string(fn_spec.camel_name());
self.leaf(indent + 1, &["camel", &camel]);
self.leaf(indent + 1, &["has-input", kdl_bool(fn_spec.has_input())]);
if let Some(input_type) = fn_spec.input_type() {
let lit = kdl_string(input_type);
self.leaf(indent + 1, &["input", &lit]);
}
let output_lit = kdl_string(fn_spec.output_type());
self.leaf(indent + 1, &["output", &output_lit]);
if fn_spec.output_nullable() {
self.leaf(indent + 1, &["output-nullable", kdl_bool(true)]);
}
let transport_lit = kdl_string(fn_spec.transport().name());
self.leaf(indent + 1, &["transport", &transport_lit]);
if let Some(ctx) = fn_spec.context() {
let lit = kdl_string(ctx);
self.leaf(indent + 1, &["context", &lit]);
}
for affect in fn_spec.affects() {
// Mirror Python's behavior: only context-typed affects make it
// into the KDL `affects` leaf. Function-typed affects are
// reserved for a future IR extension.
if let crate::ir::AffectTarget::Context(name) = affect {
let lit = kdl_string(name);
self.leaf(indent + 1, &["affects", &lit]);
}
}
for merge in fn_spec.merge() {
let lit = kdl_string(merge);
self.leaf(indent + 1, &["merge", &lit]);
}
if fn_spec.is_form() {
self.leaf(indent + 1, &["is-form", kdl_bool(true)]);
if let Some(form_name) = fn_spec.form_name() {
let lit = kdl_string(form_name);
self.leaf(indent + 1, &["form-name", &lit]);
}
if let Some(form_role) = fn_spec.form_role() {
let lit = kdl_string(form_role);
self.leaf(indent + 1, &["form-role", &lit]);
}
}
self.close(indent);
}
fn emit_context(&mut self, indent: usize, ctx_name: &str, members: &[&'static dyn FunctionSpec]) {
let name_lit = kdl_string(ctx_name);
self.open(indent, &["context", &name_lit]);
// Function membership in registration order.
for fn_spec in members {
let lit = kdl_string(fn_spec.name());
self.leaf(indent + 1, &["function", &lit]);
}
// Param info — collect across every member, then emit alphabetized
// by param name to match Python.
struct ParamSlot {
primitive: Primitive,
shared_by: Vec<&'static str>,
}
let mut params: BTreeMap<&'static str, ParamSlot> = BTreeMap::new();
for fn_spec in members {
for p in fn_spec.input_params() {
let slot = params.entry(p.name).or_insert(ParamSlot {
primitive: p.primitive,
shared_by: Vec::new(),
});
slot.primitive = p.primitive;
slot.shared_by.push(fn_spec.name());
}
}
let member_count = members.len();
for (param_name, slot) in params.iter() {
let name_lit = kdl_string(param_name);
self.open(indent + 1, &["param", &name_lit]);
let type_lit = kdl_string(slot.primitive.name());
self.leaf(indent + 2, &["type", &type_lit]);
let required = slot.shared_by.len() == member_count;
self.leaf(indent + 2, &["required", kdl_bool(required)]);
for sharer in &slot.shared_by {
let lit = kdl_string(sharer);
self.leaf(indent + 2, &["shared-by", &lit]);
}
self.close(indent + 1);
}
self.close(indent);
}
fn into_string(mut self) -> String {
// Trim trailing blanks, then add a single terminating newline.
while matches!(self.lines.last(), Some(s) if s.is_empty()) {
self.lines.pop();
}
let mut out = self.lines.join("\n");
out.push('\n');
out
}
}
/// Collected typed registries view used by `build_ir`.
pub(crate) struct IrSnapshot {
pub types: BTreeMap<&'static str, NamedType>,
pub functions: Vec<&'static dyn FunctionSpec>,
pub contexts: Vec<(&'static str, Vec<&'static dyn FunctionSpec>)>,
}
impl IrSnapshot {
pub(crate) fn collect() -> Self {
// Types: alphabetized for byte-equivalence with Python's `sorted(named_types)`.
let mut types: BTreeMap<&'static str, NamedType> = BTreeMap::new();
for entry in TYPES {
types.insert(entry.name, (entry.shape_fn)());
}
// Functions: alphabetical by wire name (canonical IR ordering,
// matches the Python emitter's `sorted(functions)`). Skip `private`.
let mut functions: Vec<&'static dyn FunctionSpec> = FUNCTIONS
.iter()
.copied()
.filter(|f| !f.private())
.collect();
functions.sort_by_key(|f| f.name());
// Contexts: alphabetical by name (canonical IR ordering), each with
// its members sorted alphabetically too.
let mut context_names: Vec<&'static str> = CONTEXTS.iter().map(|c| c.name).collect();
context_names.sort();
let mut contexts: Vec<(&'static str, Vec<&'static dyn FunctionSpec>)> = Vec::new();
for name in context_names {
let mut members: Vec<&'static dyn FunctionSpec> = functions
.iter()
.copied()
.filter(|f| f.context() == Some(name))
.collect();
members.sort_by_key(|f| f.name());
if !members.is_empty() {
contexts.push((name, members));
}
}
Self {
types,
functions,
contexts,
}
}
}
/// Build the Mizan IR for every registered type/function/context. Returns KDL.
pub fn build_ir() -> String {
crate::graph_check::verify_invariants();
let snap = IrSnapshot::collect();
let mut em = Emitter::new();
// Type definitions
let types_emitted = !snap.types.is_empty();
for (name, body) in &snap.types {
em.emit_named_type(0, name, body);
}
if types_emitted {
em.blank();
}
// Functions
let fns_emitted = !snap.functions.is_empty();
for fn_spec in &snap.functions {
em.emit_function(0, *fn_spec);
}
if fns_emitted {
em.blank();
}
// Contexts
let ctxs_emitted = !snap.contexts.is_empty();
for (ctx_name, members) in &snap.contexts {
em.emit_context(0, ctx_name, members);
}
if ctxs_emitted {
em.blank();
}
// Future: channels — once channel registry lands on the Rust side.
em.into_string()
}