All 176 files in the four accumulating directories now use a zero-padded 4-digit counter prefix that reflects creation order (`NNNN-slug.md`). The counter is assigned per directory in strict git-log creation order; ties broken alphabetically by original name. The old `YYYY-MM-DD-` prefix on docs/specs/ and docs/plans/ files is dropped — the date is recoverable from git log and the counter carries the ordering. A file's counter is stable for the life of the file: never reassigned, never reused, never compacted. Deleted files retire their counter; subsequent files do not fill the gap. This is the property that lets cross-references stay literal — refs use the full filename including the counter (`design/contracts/0007-honesty-rule.md`) so they grep cleanly and resolve directly without a glob step. 313 cross-references updated across .md/.rs/.toml/.c/.json files (test pins, include_str! paths, design-INDEX entries, baseline notes, runtime C comments, inter-contract markdown links incl. bare basename and `../models/foo.md` forms). CLAUDE.md gets a new "File-naming convention" section spelling out the rule and rationale. skills/brainstorm/SKILL.md and skills/planner/SKILL.md updated so new spec/plan creation produces counter-prefixed names from the start. The full test suite (cargo test --workspace) passes.
35 KiB
Iteration ct.4: Milestone Close — Implementation Plan
Parent spec:
docs/specs/0007-canonical-type-names.mdFor agentic workers: REQUIRED SUB-SKILL: use
skills/implementto run this plan. Steps use- [ ]checkboxes for tracking.
Goal: Land the milestone-closing consolidation: amend DESIGN.md
with the canonical-form rule, pin hashes for the two migrated
cross-module fixtures, add qualifier-trim-on-print to the prose
printer with a snapshot test, and recreate
examples/compare_primitives_smoke.ail.json as the end-to-end
demonstration that closes the iter-23.3-Task-4 bug.
Architecture: Four tasks across DESIGN, hash regression, prose
printer, and end-to-end fixture. Task 1 is design writing. Task 2
adds new hash-pin tests (the existing pins on sum/list are
intra-module and unaffected; the migrated ordering_match /
test_22b1_dup_a get new pins to lock the post-migration state).
Task 3 threads module_name context through the prose printer so
intra-module qualified Type::Cons get trimmed on print, plus a
snapshot for ordering_match. Task 4 authors the
compare_primitives_smoke fixture (9 Ord.compare invocations
across Int/Bool/Str, expected stdout 1\n2\n3\n1\n2\n3\n1\n2\n3\n)
and adds an E2E test plus an IR-shape integration test asserting
the compare__T mono symbols are emitted.
Tech Stack: docs/DESIGN.md, crates/ailang-core/src/hash.rs,
crates/ailang-prose/src/lib.rs, crates/ailang-prose/tests/snapshot.rs,
examples/*.ail.json + .prose.txt, crates/ail/tests/e2e.rs.
Files this plan creates or modifies
- Modify:
docs/DESIGN.md— amend Decision 2 (around line 1067) with the canonical-form rule forType::Connames; review the surrounding hash-stability promises at lines 1121, 1136-1139 to make sure they explicitly cite the migrated fixtures as the named exception. - Modify:
crates/ailang-core/src/hash.rs— append three new pin tests near the existingiter22b1_schema_extension_preserves_pre_22b_hashesblock: one pinning the new canonical hashes of the two migrated fixtures, one assertingsum.ail.jsonandlist.ail.json(both intra-module-only) still hash bit-identically. - Modify:
crates/ailang-prose/src/lib.rs— thread anowning_module: &strcontext throughwrite_module→write_def→ ... →write_type; inwrite_type'sType::Conarm, trim a qualifier whose owner matchesowning_module. - Modify:
crates/ailang-prose/tests/snapshot.rs— add one testsnapshot_ordering_matchconsuming the newexamples/ordering_match.prose.txtsnapshot. - Create:
examples/ordering_match.prose.txt— committed prose snapshot for the migrated cross-module fixture. - Create:
examples/compare_primitives_smoke.ail.json— the iter-23.3 demonstration fixture. - Modify:
crates/ail/tests/e2e.rs— append two tests:compare_primitives_smoke_prints_1_2_3_thrice(E2E) andcompare_primitives_smoke_emits_compare_mono_symbols(IR shape).
No files are deleted.
Task 1: DESIGN.md Decision 2 amendment
Decision 2 (content-addressed defs) in DESIGN.md is the spec of
the .ail.json canonical form. It currently does not pin down
how Type::Con names are scoped — that policy lived in
JOURNAL-level prose until ct.1 made it normative. This task
writes the rule into DESIGN.md so future iters check against it
during architect drift review.
Files:
-
Modify:
docs/DESIGN.mdnear line 1067 (Decision 2 / canonical form) — add a new subsection or paragraph naming the rule. -
Modify:
docs/DESIGN.mdline 1136-1139 — update theiter19b_*regression-pin citation to also name the new ct.1-era pins for migrated fixtures. -
Step 1: Read the Decision 2 surrounding context
Read docs/DESIGN.md lines 1050-1145 (approximately). Identify
the natural insertion point for the canonical-form rule — likely
adjacent to the existing "JSON canonical hash for every existing
fixture stays bit-identical" paragraph at 1067-1070.
- Step 2: Insert the canonical-form rule
Edit docs/DESIGN.md. After the paragraph at lines 1067-1070
("JSON canonical hash for every existing fixture stays bit-
identical..."), insert a new paragraph block:
**Type::Con name scoping (canonical form, since ct.1).** Within a
`.ail.json`, a `Type::Con.name` is interpreted relative to the
file's top-level `"name"` field (the owning module). Bare names
(no `.`) refer to a TypeDef in the owning module's own `defs`.
Cross-module references MUST be qualified `<owning_module>.<TypeName>`
where `<owning_module>` is a known module in the workspace.
Primitives (`Int`, `Bool`, `Str`, `Unit`, `Float`) are bare and
have no module qualifier. Bare cross-module references are a
schema violation (`WorkspaceLoadError::BareCrossModuleTypeRef`);
qualified references whose owner is unknown are also a violation
(`WorkspaceLoadError::BadCrossModuleTypeRef`). The same rule
applies to `Term::Ctor.type_name`.
Class names (`ClassDef.name`, `InstanceDef.class`,
`SuperclassRef.class`, `Constraint.class`) are NOT module-scoped
under this rule; they remain workspace-flat with
`MethodNameCollision` enforced at load. Class-name scoping is a
future milestone with its own DESIGN amendment.
- Step 3: Update the regression-pin citation at 1136-1139
Edit docs/DESIGN.md. Find the paragraph at lines 1136-1139 that
ends with "regression-pinned by iter19b_empty_suppress_preserves_pre_19b_hashes
and iter19b_schema_extension_preserves_pre_19b_hashes."
Append at the end of that paragraph:
The canonical-form tightening in ct.1 shifted the hashes of two
cross-module fixtures (`ordering_match.ail.json` and
`test_22b1_dup_a.ail.json`); all intra-module fixtures, including
the regression-pinned `sum.ail.json` and `list.ail.json`, remain
bit-identical. The new pins are
`ct4_migrated_fixtures_have_canonical_form_hashes` (locks the
post-migration hashes) and
`ct4_unmigrated_fixtures_remain_bit_identical` (re-asserts the
existing 13a/19b/22b.1 hashes still hold).
- Step 4: Verify DESIGN.md still parses cleanly
Run: cargo doc --workspace --no-deps 2>&1 | grep -i 'DESIGN\.md\|warning' | head
Expected: no new errors (DESIGN.md is rustdoc-rendered in places; the additions are plain markdown with no special syntax).
- Step 5: Commit
git add docs/DESIGN.md
git commit -m "iter ct.4.1: DESIGN.md Decision 2 — canonical Type::Con name scoping rule"
Task 2: Hash pins for migrated fixtures
The existing pin tests in crates/ailang-core/src/hash.rs use
sum.ail.json and list.ail.json — both intra-module, neither
touched by the ct.1 migration. Their pinned hashes remain valid
and don't need re-baselining. This task ADDS new pins that lock
in the canonical-form hashes of the two migrated fixtures, plus
re-asserts the unmigrated pins still hold.
Files:
-
Modify:
crates/ailang-core/src/hash.rs— append two new tests at the end ofmod tests. -
Step 1: Capture the current hashes of the migrated fixtures
Run the following one-liner to print the canonical hash of each
migrated fixture's relevant def. From a Rust REPL is overkill;
write a single throwaway test in crates/ailang-core/src/hash.rs
that asserts known-wrong hashes (so it fails and prints the
actual hash):
Add inside mod tests:
#[test]
#[ignore] // helper: run manually with --ignored
fn print_migrated_fixture_hashes() {
let manifest_dir = std::path::PathBuf::from(env!("CARGO_MANIFEST_DIR"));
let examples = manifest_dir.join("../../examples");
let ord_src = std::fs::read(examples.join("ordering_match.ail.json")).unwrap();
let ord_mod: crate::ast::Module = serde_json::from_slice(&ord_src).unwrap();
let main_def = ord_mod.defs.iter().find(|d| d.name() == "main").unwrap();
eprintln!("ordering_match::main: {}", def_hash(main_def));
let dup_a_src = std::fs::read(examples.join("test_22b1_dup_a.ail.json")).unwrap();
let dup_a_mod: crate::ast::Module = serde_json::from_slice(&dup_a_src).unwrap();
for d in &dup_a_mod.defs {
eprintln!("test_22b1_dup_a::{}: {}", d.name(), def_hash(d));
}
panic!("captured hashes — see stderr");
}
Run: cargo test --workspace -p ailang-core print_migrated_fixture_hashes --ignored -- --nocapture 2>&1 | grep '::'
Expected: prints hash lines for ordering_match::main and each
test_22b1_dup_a::<def_name>. Record the hashes for use in Step 2.
- Step 2: Replace the helper with the real pin test
Edit crates/ailang-core/src/hash.rs — replace the throwaway
helper test with the real pin:
/// Iter ct.4 (canonical-type-names milestone close): pin the
/// canonical-form hashes of the two cross-module fixtures
/// migrated by `ail migrate-canonical-types` in ct.1. These
/// hashes are the post-migration state; pre-migration values
/// (when both fixtures carried bare cross-module Type::Con)
/// are no longer reproducible because the ct.1 validator
/// rejects that shape upstream.
#[test]
fn ct4_migrated_fixtures_have_canonical_form_hashes() {
let manifest_dir = std::path::PathBuf::from(env!("CARGO_MANIFEST_DIR"));
let examples = manifest_dir.join("../../examples");
let ord_src = std::fs::read(examples.join("ordering_match.ail.json"))
.expect("examples/ordering_match.ail.json present");
let ord_mod: crate::ast::Module = serde_json::from_slice(&ord_src).unwrap();
let main_def = ord_mod.defs.iter().find(|d| d.name() == "main").unwrap();
assert_eq!(
def_hash(main_def),
// Replace this string with the actual hash printed in Step 1.
"REPLACE_WITH_CAPTURED_HASH_FROM_STEP_1",
"ordering_match::main canonical hash must match captured post-migration value"
);
let dup_a_src = std::fs::read(examples.join("test_22b1_dup_a.ail.json"))
.expect("examples/test_22b1_dup_a.ail.json present");
let dup_a_mod: crate::ast::Module = serde_json::from_slice(&dup_a_src).unwrap();
// Pin the first def's hash. If test_22b1_dup_a has multiple defs,
// pin only the type def that carries the migrated qualifier.
let first_def = dup_a_mod.defs.first().expect("at least one def");
assert_eq!(
def_hash(first_def),
"REPLACE_WITH_SECOND_CAPTURED_HASH_FROM_STEP_1",
"test_22b1_dup_a::{} canonical hash must match captured post-migration value",
first_def.name(),
);
}
You'll need to substitute the actual hash strings from Step 1's
captured output and remove the placeholder strings. If
test_22b1_dup_a has multiple defs of relevance, pin each (one
assert_eq! per def).
- Step 3: Add a re-assertion pin for unmigrated fixtures
Edit crates/ailang-core/src/hash.rs — append a second new test:
/// Iter ct.4: re-assert that the canonical-form tightening
/// did NOT change hashes of intra-module-only fixtures. The
/// existing iter-13a / iter-19b / iter-22b.1 pin tests already
/// assert these; this one names the canonical-type-names
/// milestone explicitly so future archaeology finds the
/// connection.
#[test]
fn ct4_unmigrated_fixtures_remain_bit_identical() {
let manifest_dir = std::path::PathBuf::from(env!("CARGO_MANIFEST_DIR"));
let examples = manifest_dir.join("../../examples");
let sum_src = std::fs::read(examples.join("sum.ail.json")).unwrap();
let sum_mod: crate::ast::Module = serde_json::from_slice(&sum_src).unwrap();
let sum_def = sum_mod.defs.iter().find(|d| d.name() == "sum").unwrap();
assert_eq!(def_hash(sum_def), "db33f57cb329935e",
"sum.sum hash drifted across canonical-form tightening — unexpected");
let list_src = std::fs::read(examples.join("list.ail.json")).unwrap();
let list_mod: crate::ast::Module = serde_json::from_slice(&list_src).unwrap();
let int_list_def = list_mod.defs.iter().find(|d| d.name() == "IntList").unwrap();
assert_eq!(def_hash(int_list_def), "b082192bd0c99202",
"list.IntList hash drifted across canonical-form tightening — unexpected");
}
- Step 4: Run the pin tests
Run: cargo test --workspace -p ailang-core ct4_migrated_fixtures_have_canonical_form_hashes ct4_unmigrated_fixtures_remain_bit_identical
Expected: BOTH PASS.
- Step 5: Run the full test suite
Run: cargo test --workspace
Expected: all pass (test count grows by 2).
- Step 6: Commit
git add crates/ailang-core/src/hash.rs
git commit -m "iter ct.4.2: pin canonical-form hashes for migrated fixtures + unmigrated sanity"
Task 3: Prose qualifier-trim-on-print + snapshot
The prose printer at crates/ailang-prose/src/lib.rs:462 writes
Type::Con.name verbatim. When the canonical .ail.json already
has bare names for intra-module Type::Cons (the canonical
convention), this is fine. But the validator does not reject
qualified intra-module refs (it only rejects bare cross-module
and bad-qualified); a hand-authored fixture could legally have
prelude.Ordering inside prelude's own file. The prose form
should print this as bare Ordering so the prose surface stays
LLM-author-friendly (qualifier omitted because it follows from
the file's module name).
This task threads owning_module context through the prose
printer and trims at write_type's Type::Con arm. Plus a
snapshot test that pins the print behaviour for the migrated
cross-module fixture (ordering_match.ail.json, which uses
prelude.Ordering cross-module qualifier — the trim does NOT
fire here because the qualifier's owner differs from the file's
module).
Files:
-
Modify:
crates/ailang-prose/src/lib.rs— threadowning_modulethrough everywrite_*fn signature fromwrite_moduledown towrite_type; trim inwrite_type'sType::Conarm. -
Create:
examples/ordering_match.prose.txt— committed snapshot. -
Modify:
crates/ailang-prose/tests/snapshot.rs— register one new snapshot test. -
Step 1: Write the RED test for trim behaviour
Append to crates/ailang-prose/src/lib.rs mod tests (or wherever
the existing unit tests live — search for mod tests near the
bottom of the file):
/// ct.4 Task 3: when a `Type::Con.name` is qualified with the
/// owning module's own name (e.g. `prelude.Ordering` inside
/// `prelude`'s own file), the prose printer must trim the
/// qualifier and emit bare `Ordering`. Cross-module qualified
/// refs (`prelude.Ordering` inside any non-prelude file)
/// round-trip verbatim.
#[test]
fn type_con_qualifier_trimmed_when_owner_matches_module() {
use ailang_core::ast::{Module, Def, FnDef, Type, Term, Literal, ParamMode};
let m = Module {
schema: "ailang/v0".into(),
name: "prelude".into(),
imports: vec![],
defs: vec![Def::Fn(FnDef {
name: "noop".into(),
ty: Type::Fn {
params: vec![],
ret: Box::new(Type::Con {
name: "prelude.Ordering".into(),
args: vec![],
}),
effects: vec![],
param_modes: vec![],
ret_mode: ParamMode::Implicit,
},
params: vec![],
body: Term::Ctor {
type_name: "prelude.Ordering".into(),
ctor: "EQ".into(),
args: vec![],
},
doc: None,
suppress: vec![],
})],
};
let prose = module_to_prose(&m);
assert!(
prose.contains("-> Ordering"),
"expected bare `Ordering` in fn return type (owner == file's \
module); got prose:\n{}",
prose
);
assert!(
!prose.contains("prelude.Ordering"),
"expected NO `prelude.Ordering` after trim; got prose:\n{}",
prose
);
}
/// ct.4 Task 3: a cross-module Type::Con (owner != current
/// file's module) must round-trip verbatim — no trim.
#[test]
fn type_con_qualifier_preserved_when_owner_differs() {
use ailang_core::ast::{Module, Def, FnDef, Type, Term, Literal, ParamMode, Import};
let m = Module {
schema: "ailang/v0".into(),
name: "user".into(),
imports: vec![Import { module: "prelude".into(), alias: None }],
defs: vec![Def::Fn(FnDef {
name: "lt".into(),
ty: Type::Fn {
params: vec![],
ret: Box::new(Type::Con {
name: "prelude.Ordering".into(),
args: vec![],
}),
effects: vec![],
param_modes: vec![],
ret_mode: ParamMode::Implicit,
},
params: vec![],
body: Term::Ctor {
type_name: "prelude.Ordering".into(),
ctor: "LT".into(),
args: vec![],
},
doc: None,
suppress: vec![],
})],
};
let prose = module_to_prose(&m);
assert!(
prose.contains("prelude.Ordering"),
"expected qualified `prelude.Ordering` preserved (owner != file's \
module); got prose:\n{}",
prose
);
}
- Step 2: Run the RED tests to verify they fail
Run: cargo test --workspace -p ailang-prose type_con_qualifier_trimmed type_con_qualifier_preserved
Expected:
-
type_con_qualifier_trimmed_when_owner_matches_module: FAIL (current printer writesprelude.Orderingverbatim). -
type_con_qualifier_preserved_when_owner_differs: PASS (current printer already writes qualified verbatim). -
Step 3: Thread
owning_modulethrough the print signatures
Edit crates/ailang-prose/src/lib.rs. Modify the following
signatures to add owning_module: &str as the last parameter
(after level: usize where present, before out-mutating return
of ()):
write_defwrite_type_defwrite_fn_defwrite_const_defwrite_class_defwrite_instance_defwrite_typewrite_mode_type(if it exists; check)- any other
write_*that consumes aTypeorTerm
For each call site within the same file, add the propagated
m.name.as_str() argument from the entry point (module_to_prose
passes m.name.as_str() to write_module's helper, which then
threads it to every write_def call).
Refactor the entry point module_to_prose:
pub fn module_to_prose(m: &Module) -> String {
let mut out = String::new();
write_module(&mut out, m);
out
}
stays the same; the threading begins inside write_module.
write_module's body — the m.name.as_str() is captured locally
(let module_name = m.name.as_str();) and passed to each
write_def(&mut out, def, level, module_name) call.
- Step 4: Apply the trim in
write_type'sType::Conarm
Edit crates/ailang-prose/src/lib.rs around line 465. The Type::Con
arm currently reads:
Type::Con { name, args } => {
out.push_str(name);
// ... args writing
}
Replace with:
Type::Con { name, args } => {
// Trim a qualifier whose owner matches the current
// file's module: `prelude.Ordering` printed from
// inside `prelude` becomes bare `Ordering`. Cross-
// module qualifiers (`prelude.Ordering` printed from
// inside any non-prelude module) survive verbatim.
let display_name = match name.split_once('.') {
Some((owner, suffix)) if owner == owning_module => suffix,
_ => name.as_str(),
};
out.push_str(display_name);
// ... args writing (preserved verbatim, recursing
// into write_type with the same owning_module)
}
Where the args recursion calls write_type(out, a), update to
write_type(out, a, owning_module).
- Step 5: Apply the same trim in
Term::Ctor's type_name (in write_term, wherever it lives)
The prose printer also writes Term::Ctor.type_name somewhere
in its term-lowering. Search for Term::Ctor in crates/ailang-prose/src/lib.rs
and find the analogous out.push_str(type_name) site. Apply the
same trim logic:
// in the Term::Ctor write arm — find the existing string-push site
let display_type = match type_name.split_once('.') {
Some((owner, suffix)) if owner == owning_module => suffix,
_ => type_name.as_str(),
};
out.push_str(display_type);
- Step 6: Run the RED tests now expecting GREEN
Run: cargo test --workspace -p ailang-prose type_con_qualifier_trimmed type_con_qualifier_preserved
Expected: BOTH PASS.
- Step 7: Regenerate existing prose snapshots
The trim changes how prelude.Ordering (and any other
qualifier-matching pattern) renders. Most existing prose
snapshots are intra-module-only fixtures with no qualified
Type::Cons — they should be byte-identical. Verify:
Run: cargo test --workspace -p ailang-prose
Expected: all existing snapshot tests still pass. If any fail
with a .prose.txt.actual byte diff, manually inspect the diff:
- If the diff is ONLY the qualifier-trim behaviour we just added
(and the snapshot input file does have qualified intra-module
Type::Cons), regenerate the snapshot:
cargo run -p ail -- prose examples/<stem>.ail.json > examples/<stem>.prose.txt. Confirm the diff post-regenerate is what you expect. - If the diff is OTHERWISE unexpected, that is a bug — investigate.
For each snapshot whose .actual file appeared, decide whether
to regenerate or revert.
- Step 8: Create the
ordering_matchsnapshot
Run: cargo run -p ail -- prose examples/ordering_match.ail.json > examples/ordering_match.prose.txt
Expected: command succeeds, the output file is non-empty and
contains prelude.Ordering (cross-module qualifier preserved
because prelude ≠ ordering_match).
- Step 9: Register the new snapshot test
Edit crates/ailang-prose/tests/snapshot.rs. Append:
#[test]
fn snapshot_ordering_match() {
check_snapshot("ordering_match");
}
- Step 10: Run the prose test suite
Run: cargo test --workspace -p ailang-prose
Expected: all pass, including the new snapshot_ordering_match.
- Step 11: Run the full test suite
Run: cargo test --workspace
Expected: all pass.
- Step 12: Commit
git add crates/ailang-prose/src/lib.rs \
crates/ailang-prose/tests/snapshot.rs \
examples/ordering_match.prose.txt
git commit -m "iter ct.4.3: prose qualifier-trim-on-print + ordering_match snapshot"
Task 4: Recreate compare_primitives_smoke.ail.json + E2E
This task resumes the paused iter 23.3 work (TaskList #15+#16): recreate the end-to-end fixture that demonstrates the iter-23.1/23.3 bug closure, run it through the full pipeline, and lock the IR-shape integration assertion.
The fixture: consumer module compare_primitives_smoke imports
prelude, calls Ord.compare on 9 pairs (3 each of Int/Bool/Str
arranged as LT-case / EQ-case / GT-case), pattern-matches each
result on LT | EQ | GT, prints 1 / 2 / 3 respectively. Expected
stdout: 1\n2\n3\n1\n2\n3\n1\n2\n3\n.
Files:
-
Create:
examples/compare_primitives_smoke.ail.json— the fixture. -
Modify:
crates/ail/tests/e2e.rs— append the E2E test and the IR-shape integration test. -
Step 1: Author the canonical-form fixture
Create examples/compare_primitives_smoke.ail.json with the
structure below. The fixture has one fn main with a body
that is a sequence of 9 do io/print_int statements; each
print_int wraps a match on (call prelude.compare A B)
returning a prelude.Ordering. Use qualified
prelude.Ordering throughout (canonical post-ct.1 since the
type's owner ≠ this file's module).
For brevity, build the fixture incrementally — start with one case, verify it loads, then add the rest. The shape of a single case (Int, LT):
{
"t": "do",
"op": "io/print_int",
"args": [
{
"t": "match",
"scrutinee": {
"t": "app",
"fn": { "t": "var", "name": "prelude.compare" },
"args": [
{ "t": "lit", "lit": { "kind": "int", "value": 1 } },
{ "t": "lit", "lit": { "kind": "int", "value": 2 } }
]
},
"arms": [
{
"pat": { "p": "ctor", "ctor": "LT", "fields": [] },
"body": { "t": "lit", "lit": { "kind": "int", "value": 1 } }
},
{
"pat": { "p": "ctor", "ctor": "EQ", "fields": [] },
"body": { "t": "lit", "lit": { "kind": "int", "value": 2 } }
},
{
"pat": { "p": "ctor", "ctor": "GT", "fields": [] },
"body": { "t": "lit", "lit": { "kind": "int", "value": 3 } }
}
]
}
]
}
Compose 9 such blocks (Int LT/EQ/GT, Bool LT/EQ/GT, Str LT/EQ/GT)
in a seq chain. For Bool, use bool literals ({"kind": "bool", "value": false});
for Str, use str literals ({"kind": "str", "value": "a"}).
The fn main returns Unit with IO effect:
{
"schema": "ailang/v0",
"name": "compare_primitives_smoke",
"imports": [{ "module": "prelude" }],
"defs": [
{
"kind": "fn",
"name": "main",
"type": {
"k": "fn",
"params": [],
"ret": { "k": "con", "name": "Unit" },
"effects": ["IO"]
},
"params": [],
"body": { "t": "seq", "lhs": <case 1>, "rhs": { "t": "seq", "lhs": <case 2>, "rhs": ... <case 9> ... } }
}
]
}
The 9 cases pair-pattern (which value, which compare, expected output):
| # | type | A | B | expected match | |
|---|---|---|---|---|---|
| 1 | Int | 1 | 2 | LT | 1 |
| 2 | Int | 1 | 1 | EQ | 2 |
| 3 | Int | 2 | 1 | GT | 3 |
| 4 | Bool | false | true | LT | 1 |
| 5 | Bool | false | false | EQ | 2 |
| 6 | Bool | true | false | GT | 3 |
| 7 | Str | "a" | "b" | LT | 1 |
| 8 | Str | "a" | "a" | EQ | 2 |
| 9 | Str | "b" | "a" | GT | 3 |
You'll write this manually as one big JSON file. Use 2-space
indentation and standard serde_json-canonical key ordering
(the validator + manifest test will fail otherwise; alternatively,
run the file through cargo run -p ail -- manifest examples/compare_primitives_smoke.ail.json
to canonicalise once it loads, then commit the canonicalised
form).
- Step 2: Verify the fixture loads + typechecks
Run: cargo run -p ail -- check examples/compare_primitives_smoke.ail.json
Expected: exit 0, no diagnostics. If validator complains about
canonical form (e.g. bare Ordering somewhere), fix and re-run.
If typechecker complains about an unknown ctor or pattern type
mismatch, that is the iter-23.1 bug — investigate (it should
NOT happen post-ct.2).
- Step 3: Verify the fixture runs end-to-end
Run: cargo run -p ail -- run examples/compare_primitives_smoke.ail.json
Expected: stdout exactly 1\n2\n3\n1\n2\n3\n1\n2\n3\n (nine
lines, no extra newline at the very end depending on the runtime
printer convention — match whatever the existing
ordering_match_via_prelude_prints_1 test asserts for trailing
newline).
If the run fails, the iter-23.1 bug closure is NOT complete — escalate to orchestrator.
- Step 4: Append the E2E test to e2e.rs
Edit crates/ail/tests/e2e.rs. Append after the
ordering_match_via_prelude_prints_1 test (around line 920):
/// Iter 23.3 Task 4 (resumed in ct.4): the end-to-end
/// demonstration that the iter-23.1 cross-module Type::Con
/// mismatch bug is closed. `examples/compare_primitives_smoke.ail.json`
/// imports prelude, calls `prelude.compare` on three pairs each
/// of Int / Bool / Str (arranged LT / EQ / GT per type), and
/// pattern-matches the result. Expected stdout encodes the
/// 9-case decision lattice as `1\n2\n3` × 3.
///
/// Pre-ct.1 / pre-ct.2, the bare `Ordering` returned by
/// `compare` mismatched the qualified `prelude.Ordering`
/// scrutinee at the user-side pattern lookup. Post-canonical-
/// type-names, the scrutinee is qualified throughout, the
/// pattern lookup is type-driven, and the run completes cleanly.
#[test]
fn compare_primitives_smoke_prints_1_2_3_thrice() {
let stdout = build_and_run("compare_primitives_smoke.ail.json");
assert_eq!(stdout, "1\n2\n3\n1\n2\n3\n1\n2\n3\n");
}
- Step 5: Append the IR-shape integration test
Append immediately after the E2E test:
/// Iter 23.3 Task 4 / IR-shape: the mono pass must emit
/// `compare__Int`, `compare__Bool`, and `compare__Str` symbols
/// driven by the three `prelude.compare` call sites in the
/// smoke fixture. Codegen's `try_emit_primitive_instance_body`
/// intercepts each one (iter 23.3.2). This test pins the
/// presence of those symbols in the emitted IR so a future
/// refactor cannot silently drop one.
#[test]
fn compare_primitives_smoke_emits_compare_mono_symbols() {
let ir = emit_ir("compare_primitives_smoke.ail.json");
assert!(
ir.contains("compare__Int") || ir.contains("@compare__Int"),
"compare__Int symbol missing from emitted IR"
);
assert!(
ir.contains("compare__Bool") || ir.contains("@compare__Bool"),
"compare__Bool symbol missing from emitted IR"
);
assert!(
ir.contains("compare__Str") || ir.contains("@compare__Str"),
"compare__Str symbol missing from emitted IR"
);
}
If a helper emit_ir(fixture: &str) -> String does not already
exist in e2e.rs, add it adjacent to build_and_run (it should
invoke cargo run -p ail -- emit-ir <fixture> and capture
stdout). Use the existing build_and_run's implementation as a
template.
- Step 6: Run the new E2E + IR tests
Run: cargo test --workspace -p ail compare_primitives_smoke_prints compare_primitives_smoke_emits
Expected: BOTH PASS.
- Step 7: Run the full test suite
Run: cargo test --workspace
Expected: all pass.
- Step 8: Commit
git add examples/compare_primitives_smoke.ail.json crates/ail/tests/e2e.rs
git commit -m "iter ct.4.4: compare_primitives_smoke E2E fixture + IR-shape test (closes iter 23.3 Task 4)"
Closing checks
- Step C1: Bench script smoke run
Run: bash bench/run.sh (or whichever entry point the project
uses; consult bench/README.md or skills/audit/SKILL.md).
Expected: exit 0, or a ratified non-zero per the audit-skill policy. No new regressions vs. baseline.
- Step C2: Confirm all acceptance criteria from the spec
Verify each item in docs/specs/0007-canonical-type-names.md
§ "Acceptance criteria":
- ✅
load_workspacevalidates everyType::Con(ct.1). - ✅ Every fixture migrated to qualified form (ct.1).
- ✅ Four obsolete mechanisms deleted (ct.2.2, ct.2.3, ct.3.1, ct.3.2).
- ✅
qualify_local_typesaudited and applied uniformly (ct.2.1). - ✅
cargo build --workspace && cargo test --workspacegreen (this commit). - ✅ Iter 23.3 Task 4+5 resumed and complete (ct.4 Task 4 above).
- ✅ DESIGN.md amended (ct.4 Task 1).
- ✅ JOURNAL entry (ct.4 Task C3 below).
- ⏳ Bench scripts exit 0 (verify via C1).
- Step C3: JOURNAL entry
Append to docs/JOURNAL.md:
## 2026-05-11 — Iteration ct.4: canonical-type-names milestone close
Four tasks landed, closing the milestone:
- **ct.4.1**: amended DESIGN.md Decision 2 with the canonical
Type::Con name scoping rule (bare = local, qualified =
cross-module, primitives bare). Updated the regression-pin
citation at the surrounding "hashes stay bit-identical"
paragraph to name the two new ct.4 pins.
- **ct.4.2**: added two new pin tests in
`crates/ailang-core/src/hash.rs` —
`ct4_migrated_fixtures_have_canonical_form_hashes` (locks the
post-migration hashes of `ordering_match.ail.json` and
`test_22b1_dup_a.ail.json`) and
`ct4_unmigrated_fixtures_remain_bit_identical` (re-asserts the
unchanged `sum`/`list` hashes).
- **ct.4.3**: added qualifier-trim-on-print to the prose
printer. `write_type` (and `Term::Ctor.type_name` writing)
now trims a qualifier whose owner matches the current file's
module. The `owning_module` context is threaded from
`write_module` down through the call chain. Two unit tests
pin the trim and the cross-module preserve behaviours; the
`ordering_match.prose.txt` snapshot is committed.
- **ct.4.4**: recreated `examples/compare_primitives_smoke.ail.json`
(the iter-23.3 Task 4 demonstration fixture). 9 cases across
Int / Bool / Str, expected stdout `1\n2\n3` × 3. Two new
E2E tests in `crates/ail/tests/e2e.rs` —
`compare_primitives_smoke_prints_1_2_3_thrice` (output
assertion) and `compare_primitives_smoke_emits_compare_mono_symbols`
(IR-shape assertion that mono emits compare__Int /
compare__Bool / compare__Str symbols).
The canonical-type-names milestone closes:
- The structural invariant: every `Type::Con` in canonical
`.ail.json` is bare (= local) or qualified
(`<owner>.<TypeName>`, with primitives bare).
- The four obsolete imports-fallback / overlay mechanisms named
in the ct.1 catalogue are gone (Pattern::Ctor, Term::Ctor
synth, codegen `lookup_ctor_by_type`, mono
`apply_per_module_ctor_index_overlay`'s ctor_index half).
- The iter-23.1-Task-4 bug is empirically closed: the smoke
fixture compiles, runs, and prints the expected 9-line output
end-to-end without any further compiler changes.
Workspace at milestone close: full `cargo build --workspace` and
`cargo test --workspace` green, 453+ tests passing
(ct.4 adds 4 new tests: 2 pin + 2 prose + 2 E2E, minus or plus
any shifts in other test count).
The skill-system pipeline calls for `audit` next (drift review +
bench regression + rustdoc), then `fieldtest` (real-world `.ailx`
examples authored against DESIGN.md only).
Out of scope, deferred:
- The codegen `lookup_ctor_in_pattern` helper (imports-walk, but
no scrutinee type to type-anchor; a later iter can plumb Type
through pattern lowering if desired).
- `check_in_workspace`'s analogous per-module overlay in
`crates/ailang-check/src/lib.rs:1234` (its env.ctor_index half
serves the duplicate-detection diagnostic at workspace-build
time, not the runtime ctor lookup; not in the canonical-type-
names scope).
- Class names (`ClassDef.name`, `InstanceDef.class`, etc.) are
intentionally bare and workspace-flat post-ct.1; type-driven
method dispatch is its own milestone (already named in
`docs/roadmap.md` as a P2 item).
- Step C4: Commit the JOURNAL entry
git add docs/JOURNAL.md
git commit -m "journal: ct.4 canonical-type-names milestone close"
- Step C5: Optional roadmap update
If the user wants the milestone removed from
docs/roadmap.md (or marked done with a checkmark), make that
edit now and commit alongside the JOURNAL entry. Otherwise leave
the roadmap as-is — the orchestrator typically prunes finished
entries in a small follow-up commit.