feat(aura-engine, aura-cli): the blueprint name op — name_gate on every authored intake

An eleventh op-script op {"op":"name","name":"<n>"} sets the composite
render name (engine Op::Name + CLI OpDoc mirror); omitting it keeps the
default "graph" byte-identically. The default leaked everywhere one
research project has more than one strategy: every store entry named
graph, every default tap recording sharing traces/graph/, and two
use-splices of unnamed blueprints colliding on the default instance
identifier — the authored name dissolves all three through existing
mechanics (no downstream site edited).

Mechanics: at-most-once per script (doc-op precedent), position-free
(read only at finish). A shared deterministic name_gate (aura-engine:
non-empty, no path separators, not . or ..) guards every seam where a
name is born from authored data, because the name keys a trace
directory unsanitized: the op intake (GraphSession::set_name) and all
four CLI fresh-file envelope intakes (register, introspect
--content-id FILE, the bare graph FILE viewer, aura run FILE — the run
route reached begin_run(signal.name()) ungated, and introspect
--params FILE rode the same parse). Store read-back (reproduce, use
resolution, introspect/params by content id) stays deliberately
ungated — C29: registered artifacts are never retroactively
invalidated — pinned by a test that plants a bad-root-name blueprint
via the registry API and asserts introspect --params still answers.
Refusal prose is single-sourced (name_gate_fault_prose) so every seam
reads identically, op-indexed on the op route.

Identity semantics: the authored name hashes into the content id (a
named document is a different document) and never into the identity id
(names are stripped as debug symbols, C23) — pinned by a twin test.
The previously untested use-splice instance-name default
(construction.rs) got its ratifying pin before the collision claim
leans on it. The registry label (register --name) stays orthogonal.

Review rounds caught and fixed: the envelope gate initially fired on
store read-back (C29 violation at the introspect surface), and the
run/params fresh-file routes were unenumerated intake seams — closed
with the call-site classification now recorded in the wrapper docs.

Verification: cargo build/test/clippy -D warnings all green (99 test
targets, 0 failures, independently re-run); the new run-route test was
RED-verified by hand-removing the gate.

closes #331
refs #328, #311

Spec: blueprint-name-op (fork minutes on #331)
This commit is contained in:
2026-07-25 04:33:18 +02:00
parent bbac29db2d
commit a851af993a
8 changed files with 542 additions and 19 deletions
+113 -13
View File
@@ -8,9 +8,9 @@ use std::collections::BTreeMap;
use std::path::Path;
use aura_engine::{
blueprint_from_json, blueprint_identity_json, blueprint_to_json, replay, ArgOpError,
BindOpError, BlueprintDoc, CompileError, Composite, GraphSession, LoadError, Op, OpError,
Scalar, ScalarKind,
blueprint_from_json, blueprint_identity_json, blueprint_to_json, name_gate, replay, ArgOpError,
BindOpError, BlueprintDoc, CompileError, Composite, GraphSession, LoadError, NameGateFault, Op,
OpError, Scalar, ScalarKind,
};
use aura_runner::runner::render_value;
use serde::Deserialize;
@@ -22,9 +22,10 @@ use crate::research_docs::resolve_id_prefix;
#[cfg(test)]
use aura_vocabulary::std_vocabulary;
/// The op-list reference `aura graph build --help` appends (#323): the ten
/// op kinds with their fields and one worked element each. Lives beside
/// [`OpDoc`] so a new op variant is one screen away from its help line.
/// The op-list reference `aura graph build --help` appends (#323): the eleven
/// op kinds with their fields and one worked element each (#331: `name` joins
/// the roster, ten -> eleven). Lives beside [`OpDoc`] so a new op variant is
/// one screen away from its help line.
pub const OP_REFERENCE: &str = r#"Op-list reference (stdin: a JSON array of op objects, applied in order):
{"op":"source","role":"price","kind":"F64"}
declare a bound root input role of a scalar kind
@@ -50,6 +51,9 @@ pub const OP_REFERENCE: &str = r#"Op-list reference (stdin: a JSON array of op o
{"op":"use","ref":{"name":"agree"},"name":"gate","bind":{"sma.length":{"I64":9}}}
splice a registered blueprint (by "content_id" or "name") under an
instance name ("bind" path-qualifies the spliced instance's params)
{"op":"name","name":"ny_momentum"}
set the composite's render name, at most once per script (default
"graph" if omitted)
Node types and their ports: aura graph introspect --vocabulary | --node <T>"#;
@@ -111,6 +115,9 @@ enum OpDoc {
#[serde(default)]
bind: BTreeMap<String, Scalar>,
},
/// Set the composite's render name (#331) — script-level, at most once;
/// the op-script twin of `replay`'s seeded default name (`"graph"`).
Name { name: String },
}
/// A `use` op's reference (#317) — exactly one of a store content id (full
@@ -143,6 +150,7 @@ impl OpDoc {
OpDoc::Doc { .. } => "doc".to_string(),
OpDoc::Use { name: Some(n), .. } => format!("use {n:?}"),
OpDoc::Use { name: None, .. } => "use".to_string(),
OpDoc::Name { .. } => "name".to_string(),
}
}
}
@@ -182,6 +190,7 @@ impl From<OpDoc> for Op {
name,
bind: bind.into_iter().collect(),
},
OpDoc::Name { name } => Op::Name { name },
}
}
}
@@ -257,9 +266,25 @@ fn format_op_error(e: &OpError) -> String {
OpError::UnknownSubgraph { ref_id } => {
format!("use: no subgraph for content id {ref_id:?}")
}
OpError::DuplicateName => "a script names its blueprint at most once".to_string(),
OpError::BadName { name, fault } => name_gate_fault_prose(name, fault),
}
}
/// Phrase a `name_gate` shape violation as prose (#331): shared by this
/// module's op-intake `format_op_error` `BadName` arm and the
/// blueprint-envelope intake's root-name gate (`composite_from_authored_text`
/// below) — the shape rule has exactly one seam-independent cause per fault,
/// so both data-borne birth routes for a name read identically.
fn name_gate_fault_prose(name: &str, fault: &NameGateFault) -> String {
let cause = match fault {
NameGateFault::Empty => "must be non-empty",
NameGateFault::ContainsSeparator => "must not contain '/' or '\\'",
NameGateFault::DotSegment => "must not be \".\" or \"..\"",
};
format!("blueprint name {name:?} is invalid: {cause} (a single path segment)")
}
/// Resolve one `use` op's [`UseRef`] to the full store content id (#317):
/// verbatim if it already IS a 64-hex content id, else a unique content-id
/// prefix (#302 semantics, reusing [`resolve_id_prefix`]) or a registry name
@@ -649,7 +674,7 @@ pub fn introspect_cmd(cmd: crate::GraphIntrospectCmd, env: &aura_runner::project
std::process::exit(1);
}
};
match composite_from_any(&text, env) {
match composite_from_authored_text(&text, env) {
Ok(c) => c,
Err(m) => {
eprintln!("aura: {m}");
@@ -819,6 +844,18 @@ pub(crate) fn blueprint_slot_prose(doc: &str, env: &aura_runner::project::Env) -
/// envelope (a JSON object: `format_version` + `blueprint`) OR a construction
/// op-list (a JSON array) — shape-discriminated on the top-level JSON type,
/// each canonicalized by its own rules.
///
/// **Ungated by design (#331 review finding).** This is the STORE READ-BACK
/// shape: `params_lines`'s content-id fetch (`resolve_blueprint_text`'s
/// non-file branch) — `introspect --params <ID>` and, by the same
/// convention, `validate_campaign_refs`'s already-ungated `blueprint_from_json`
/// call — reads whatever is already sitting in the store, and C29 says a
/// registered artifact is never retroactively invalidated. Freshly authored
/// FILE text (a hand-edited document that has not yet passed through a gated
/// intake) goes through [`composite_from_authored_text`] instead, never here
/// — `params_lines`'s OWN file branch is such a case (#331 delta re-review:
/// it used to call straight through to this fn, missing the gate; fixed by
/// branching on `resolve_blueprint_text`'s file-vs-store flag).
pub(crate) fn composite_from_any(text: &str, env: &aura_runner::project::Env) -> Result<Composite, String> {
let value: serde_json::Value =
serde_json::from_str(text).map_err(|e| format!("invalid document: {e}"))?;
@@ -836,12 +873,68 @@ pub(crate) fn composite_from_any(text: &str, env: &aura_runner::project::Env) ->
}
}
/// The FILE-intake counterpart of [`composite_from_any`] (#331 review
/// finding): identical parse, plus the blueprint-envelope root-name shape
/// gate. Every call site that builds a `Composite` from text freshly read
/// off disk (`graph register`, `graph introspect --content-id <FILE>`, the
/// bare `aura graph <FILE>` viewer, and `graph introspect --params <FILE>`'s
/// file branch) goes through this wrapper — never the ungated
/// `composite_from_any` — because a hand-edited envelope with
/// `"name":"../x"` would otherwise register/build cleanly and write
/// `traces/../x/` at run time (`trace_store.rs` joins the name unsanitized).
/// Gating the composite's name unconditionally (not just on the Object
/// branch) is harmless: an op-script-built composite's name already passed
/// the op-intake gate (`GraphSession::set_name`), so re-checking it here is
/// redundant, not restrictive — it keeps this wrapper a single shared choke
/// point rather than one that has to re-discriminate the JSON shape. Only
/// the ROOT name is checked: the filesystem seam consumes exclusively the
/// root name. Store read-back (reproduce, `use`-splice resolution from the
/// registry, and `params_lines`'s content-id branch) never reaches this
/// function — C29's "registered artifacts are never retroactively
/// invalidated" stays intact.
///
/// **One deliberate exception:** `aura run <blueprint.json>`'s loaded-blueprint
/// branch (`dispatch_run`, main.rs) also reads a fresh FILE and reaches the
/// same unsanitized `traces/<name>/` seam (via `run_signal_r`/`run_measurement`
/// -> `bind_tap_plan` -> `TraceStore::begin_run`), so it too must gate the root
/// name — but it does NOT route through this wrapper, because its grammar is
/// deliberately narrower than `composite_from_any`'s (envelope-only, no
/// op-script array fallback; a bare `blueprint_from_json` call with its own
/// load-error prose/hint convention). Routing it through this wrapper would
/// silently loosen that grammar to also accept op-scripts. It instead calls
/// [`gate_authored_root_name`] directly, sharing the identical `name_gate` +
/// `name_gate_fault_prose` primitives this wrapper uses — so the refusal wording
/// is byte-identical across all five authored-file-intake routes even though
/// the shape-discrimination step is not shared by `dispatch_run`.
pub(crate) fn composite_from_authored_text(
text: &str,
env: &aura_runner::project::Env,
) -> Result<Composite, String> {
let composite = composite_from_any(text, env)?;
gate_authored_root_name(composite.name())?;
Ok(composite)
}
/// The root-name shape gate itself (#331 delta re-review), factored out of
/// [`composite_from_authored_text`] so `dispatch_run`'s narrower-grammar file
/// intake (see that fn's doc comment) can share the exact `name_gate` call and
/// `name_gate_fault_prose` wording without going through the shape-discriminating
/// wrapper.
pub(crate) fn gate_authored_root_name(name: &str) -> Result<(), String> {
name_gate(name).map_err(|fault| name_gate_fault_prose(name, &fault))
}
/// Resolve a blueprint document's bytes from a file path or a 64-hex content
/// id in the project store (the campaign-run target-addressing convention).
/// The id shape is `aura_runner::axes::is_content_id` — the same predicate
/// `campaign run`'s target resolution uses, so the two FILE-or-id surfaces
/// cannot drift apart on what counts as a store address.
fn resolve_blueprint_text(target: &str, env: &aura_runner::project::Env) -> Result<String, String> {
/// cannot drift apart on what counts as a store address. The `bool` names
/// which branch fired: `true` for a FRESH FILE read, `false` for a STORE
/// (content-id) fetch — `params_lines` (#331 delta re-review) uses it to
/// decide whether the root-name gate applies (a fresh file must gate; a
/// store read-back must not, C29) without re-deriving the FILE-vs-id
/// distinction a second time.
fn resolve_blueprint_text(target: &str, env: &aura_runner::project::Env) -> Result<(String, bool), String> {
// A CLI arg tolerates the `content:` display prefix (#194); doc ref
// fields stay bare-only.
let target = target
@@ -851,11 +944,12 @@ fn resolve_blueprint_text(target: &str, env: &aura_runner::project::Env) -> Resu
let path = Path::new(target);
if path.is_file() {
return std::fs::read_to_string(path)
.map(|text| (text, true))
.map_err(|e| format!("cannot read {}: {e}", path.display()));
}
if aura_runner::axes::is_content_id(target) {
return match env.registry().get_blueprint(target) {
Ok(Some(json)) => Ok(json),
Ok(Some(json)) => Ok((json, false)),
Ok(None) => Err(format!("no blueprint {target} in the project store")),
Err(e) => Err(e.to_string()),
};
@@ -872,12 +966,16 @@ fn resolve_blueprint_text(target: &str, env: &aura_runner::project::Env) -> Resu
/// `bound_param_space()`'s names are already RAW (#203), so no blueprint-name
/// concatenation is needed — line-identical to the reconciled `--list-axes`
/// bound pass (`list_blueprint_axes`, main.rs), same `render_value` lexicon.
/// #331 delta re-review: the FILE target is a FOURTH freshly-authored-file
/// intake this fn's own `composite_from_any` call had missed gating — a FILE
/// routes through [`composite_from_authored_text`] instead; a content id
/// (STORE read-back) keeps the ungated `composite_from_any` (C29).
fn params_lines(target: &str, env: &aura_runner::project::Env) -> Result<String, String> {
use std::fmt::Write as _;
let text = resolve_blueprint_text(target, env)?;
let (text, is_file) = resolve_blueprint_text(target, env)?;
// Shape-discriminated like file-mode --content-id: an op-script (array)
// builds through the one-build tail, an envelope (object) loads (#202).
let composite = composite_from_any(&text, env)?;
let composite = if is_file { composite_from_authored_text(&text, env)? } else { composite_from_any(&text, env)? };
let mut out = String::new();
for p in composite.param_space() {
let _ = writeln!(out, "{}:{:?}", p.name, p.kind);
@@ -918,7 +1016,9 @@ fn register_blueprint(
.map_err(|e| format!("cannot read {}: {e}", file.display()))?;
// Shape-discriminated like file-mode --content-id (#202): either shape
// canonicalizes to the envelope form, so the stored id is shape-invariant.
let composite = composite_from_any(&text, env)?;
// Gated (#331 review finding): this text is freshly authored FILE bytes,
// not a store read-back, so `composite_from_authored_text` applies.
let composite = composite_from_authored_text(&text, env)?;
let canonical = blueprint_to_json(&composite).map_err(|e| format!("serialize error: {e:?}"))?;
let id = crate::content_id(&canonical);
let registry = env.registry();
+14 -1
View File
@@ -1884,6 +1884,19 @@ fn dispatch_run(a: RunCmd, env: &aura_runner::project::Env) {
eprintln!("aura: {path}: {msg}");
std::process::exit(2);
});
// #331 delta re-review: this is the third freshly-authored-FILE
// intake (`register`, `introspect --content-id <FILE>`, and this
// one) that must gate the root name before it ever reaches the
// filesystem — both run arms below feed `signal.name()` straight
// into `bind_tap_plan`'s `TraceStore::begin_run`, which joins it
// unsanitized (`trace_store.rs`). Not routed through
// `composite_from_authored_text` itself (see that fn's doc
// comment on why); same `name_gate`/`name_gate_fault_prose`
// primitives, this arm's own load-error exit/prose convention.
if let Err(msg) = graph_construct::gate_authored_root_name(signal.name()) {
eprintln!("aura: {path}: {msg}");
std::process::exit(2);
}
// Shape dispatch (C28 phase 3): a `bias` output → the strategy path
// (wrap_r + R evaluation), byte-identical; else ≥1 declared tap → a
// bare measurement run (no wrap_r); else an inert blueprint → refuse.
@@ -1981,7 +1994,7 @@ fn dispatch_graph(a: GraphCmd, env: &aura_runner::project::Env) {
eprintln!("aura: {path}: {e}");
std::process::exit(2);
});
let bp = graph_construct::composite_from_any(&doc, env).unwrap_or_else(|msg| {
let bp = graph_construct::composite_from_authored_text(&doc, env).unwrap_or_else(|msg| {
eprintln!("aura: {path}: {msg}");
std::process::exit(2);
});
+180
View File
@@ -88,6 +88,43 @@ const SIGNAL_DOC_NAME_RESTATED: &str = r#"[
{"op":"expose","from":"bias.bias","as":"bias"}
]"#;
/// SIGNAL_DOC with a leading `{"op":"name",...}` (#331) — authors a render
/// name distinct from the CLI's own default seed ("graph"), otherwise
/// structurally identical to SIGNAL_DOC.
const SIGNAL_DOC_NAMED: &str = r#"[
{"op":"name","name":"ny_momentum"},
{"op":"source","role":"price","kind":"F64"},
{"op":"add","type":"SMA","name":"fast","bind":{"length":{"I64":2}}},
{"op":"add","type":"SMA","name":"slow","bind":{"length":{"I64":4}}},
{"op":"add","type":"Sub"},
{"op":"add","type":"Bias"},
{"op":"feed","role":"price","into":["fast.series","slow.series"]},
{"op":"connect","from":"fast.value","to":"sub.lhs"},
{"op":"connect","from":"slow.value","to":"sub.rhs"},
{"op":"connect","from":"sub.value","to":"bias.signal"},
{"op":"expose","from":"bias.bias","as":"bias"}
]"#;
/// SIGNAL_DOC_NAME_RESTATED with an authored name (#331 spec test 7) instead
/// of the CLI's default seed: the `doc` op's text restates the *authored*
/// name ("ny_momentum") rather than the literal "graph" — the same
/// RestatesName arm of the C29 shape gate must fire against whichever name
/// the composite actually carries, not a hardcoded "graph" comparison.
const SIGNAL_DOC_NAMED_NAME_RESTATED: &str = r#"[
{"op":"name","name":"ny_momentum"},
{"op":"doc","text":"Ny Momentum"},
{"op":"source","role":"price","kind":"F64"},
{"op":"add","type":"SMA","name":"fast","bind":{"length":{"I64":2}}},
{"op":"add","type":"SMA","name":"slow","bind":{"length":{"I64":4}}},
{"op":"add","type":"Sub"},
{"op":"add","type":"Bias"},
{"op":"feed","role":"price","into":["fast.series","slow.series"]},
{"op":"connect","from":"fast.value","to":"sub.lhs"},
{"op":"connect","from":"slow.value","to":"sub.rhs"},
{"op":"connect","from":"sub.value","to":"bias.signal"},
{"op":"expose","from":"bias.bias","as":"bias"}
]"#;
/// Every op below applies cleanly (all eager checks pass), but `sub.rhs` is
/// never wired — a 0-cover input slot only the holistic finalize gate can
/// reject. The smallest document that is op-valid yet whole-document-invalid.
@@ -929,6 +966,120 @@ fn graph_register_refuses_an_op_script_doc_that_restates_the_composite_name() {
assert!(err.contains("C29"), "stderr cites the contract: {err}");
}
/// (#331 spec test 7) The C29 restates-name gate checks against the
/// *authored* name, not a hardcoded default: a `doc` op restating the name
/// set by a `name` op ("ny_momentum") refuses at register exactly like the
/// default-name case above (`graph_register_refuses_an_op_script_doc_that_
/// restates_the_composite_name`, which must stay green byte-identical).
#[test]
fn graph_register_refuses_a_doc_that_restates_an_authored_name() {
let dir = temp_cwd("register-op-script-restated-authored-name-doc");
let restated_script = dir.join("restated-authored-name-doc-script.json");
std::fs::write(&restated_script, SIGNAL_DOC_NAMED_NAME_RESTATED)
.expect("write op-script fixture");
let (out, err, code) = run_in(&dir, &["graph", "register", restated_script.to_str().unwrap()]);
assert_eq!(code, Some(1), "authored-name-restating doc refuses: {out} {err}");
assert!(err.contains("merely restates"), "stderr names the rule: {err}");
assert!(err.contains("C29"), "stderr cites the contract: {err}");
}
/// (#331 spec test 6) `graph build` on an ops doc with a `name` op emits
/// blueprint JSON carrying the authored name, not the CLI's own seeded
/// default (`"graph"`, `replay("graph", ...)` in `composite_from_str`).
#[test]
fn graph_build_emits_the_authored_name_from_a_name_op() {
let (stdout, stderr, ok) = run(&["graph", "build"], SIGNAL_DOC_NAMED);
assert!(ok, "build succeeds: {stderr}");
assert!(stdout.contains("\"name\":\"ny_momentum\""), "authored name carries into the blueprint: {stdout}");
assert!(!stdout.contains("\"name\":\"graph\""), "the default seed name must not leak in: {stdout}");
}
/// (#331 spec test 8) Refusal prose for a duplicate `name` op and a
/// shape-violating name is op-indexed (`op N (name): ...`) — the same
/// by-identifier labeling every other op fault gets.
#[test]
fn graph_build_name_op_refusals_are_op_indexed() {
let twice_named = r#"[
{"op":"source","role":"price","kind":"F64"},
{"op":"add","type":"SMA","name":"fast","bind":{"length":{"I64":2}}},
{"op":"name","name":"a"},
{"op":"name","name":"b"},
{"op":"feed","role":"price","into":["fast.series"]},
{"op":"expose","from":"fast.value","as":"out"}
]"#;
let (_out, stderr, ok) = run(&["graph", "build"], twice_named);
assert!(!ok, "a second name op refuses");
assert!(stderr.contains("op 3 (name):"), "op-indexed at the second name op: {stderr}");
let slash_named = r#"[{"op":"name","name":"a/b"}]"#;
let (_out2, stderr2, ok2) = run(&["graph", "build"], slash_named);
assert!(!ok2, "a shape-violating name refuses");
assert!(stderr2.contains("op 0 (name):"), "op-indexed at the offending op: {stderr2}");
}
/// (#331 spec test 9, at the binary seam — mirrors
/// `graph_introspect_identity_id_bridges_renamed_op_scripts`'s pattern for
/// the debug-name field): a named and an unnamed build of the SAME
/// structural script share the identity id (the name is stripped as a
/// debug symbol before identity hashing, C23) and differ in the content id
/// (the name is part of the serialized bytes).
#[test]
fn graph_named_and_unnamed_twins_share_identity_id_and_differ_content_id() {
let (ia, _e, ok) = run(&["graph", "introspect", "--identity-id"], SIGNAL_DOC);
assert!(ok, "exit success");
let (ib, _e2, ok2) = run(&["graph", "introspect", "--identity-id"], SIGNAL_DOC_NAMED);
assert!(ok2, "exit success");
assert_eq!(ia.trim(), ib.trim(), "the name is stripped before identity hashing (C23)");
let (ca, _e3, ok3) = run(&["graph", "introspect", "--content-id"], SIGNAL_DOC);
let (cb, _e4, ok4) = run(&["graph", "introspect", "--content-id"], SIGNAL_DOC_NAMED);
assert!(ok3 && ok4, "exit success");
assert_ne!(ca.trim(), cb.trim(), "the name is part of the serialized bytes -> different content id");
}
/// (#331 spec test 10, the skeptic's bypass route): the op intake
/// (`GraphSession::set_name`) is not the name's only data-borne birth
/// route — a hand-edited blueprint ENVELOPE (not an op-script) can carry a
/// shape-violating root name directly. `aura graph register` on such a file
/// is refused naming the offending name (exit 1, the refusal convention
/// every sibling test in this module asserts), not silently stored under a
/// name that would write `traces/../x/` at run time. Shipped envelope
/// fixtures with plain root names (`doc_blueprint.json`'s "sig",
/// `nested_doc_blueprint.json`'s "root", every op-script's default "graph")
/// stay accepted — the surrounding suite is that regression gate.
///
/// Spec test 10 also names `build` as a gated envelope-intake route
/// alongside `register`. The `graph build` SUBCOMMAND takes only a stdin
/// op-script (`Vec<OpDoc>`, never a JSON object) — it cannot ingest this
/// envelope fixture at all (a shape mismatch fires before any name is ever
/// read). The file-consuming envelope-intake sibling `build` route is
/// `graph introspect --content-id <FILE>` (its own doc comment: "one
/// build, then each requested id…") — the other call site this iteration
/// moved onto the same `composite_from_authored_text` choke point as
/// `register`, so it is the leg exercised here, sharing the identical
/// `name_gate_fault_prose` text.
#[test]
fn graph_build_and_register_refuse_a_hand_crafted_envelope_with_a_bad_root_name() {
let dir = temp_cwd("register-bad-root-name");
let (envelope_bytes, _e, built) = run(&["graph", "build"], SIGNAL_DOC_DESCRIBED);
assert!(built, "graph build produces the envelope");
let bad = envelope_bytes.replacen("\"name\":\"graph\"", "\"name\":\"../x\"", 1);
assert!(bad.contains("\"name\":\"../x\""), "the root-name replace landed: {bad}");
let file = dir.join("bad-root-name.json");
std::fs::write(&file, &bad).expect("write envelope fixture");
let (out, err, code) = run_in(&dir, &["graph", "register", file.to_str().unwrap()]);
assert_eq!(code, Some(1), "a shape-violating root name refuses (not a panic): {out} {err}");
assert!(err.contains("../x"), "names the offending root name: {err}");
// The other file-consuming envelope intake (same fixture, same choke
// point) refuses identically.
let (out2, err2, code2) =
run_in(&dir, &["graph", "introspect", "--content-id", file.to_str().unwrap()]);
assert_eq!(code2, Some(1), "the sibling envelope intake refuses too: {out2} {err2}");
assert_eq!(err2, err, "both file-consuming envelope intakes share the identical fault prose");
}
/// Property (#202, the same on-ramp seam at the sibling verb): `aura graph
/// introspect --params` accepts an op-script exactly as it accepts a blueprint
/// envelope — the raw axis namespace it prints for the op-script equals the one it
@@ -954,6 +1105,35 @@ fn graph_params_accepts_an_op_script_matching_its_envelope() {
assert_eq!(op_out, env_out, "op-script and envelope agree on the raw axis namespace");
}
/// (#331 review finding, the C29 exemption's ratifying pin): the root-name
/// shape gate lives at the CLI's file-consuming envelope intakes (`register`,
/// `introspect --content-id <FILE>`) — never at STORE read-back. A blueprint
/// with a shape-violating root name registered DIRECTLY through the
/// `aura-registry` store API (bypassing the CLI's gate entirely, the way a
/// pre-#331 artifact would already sit in a project's store) must still be
/// introspectable: `introspect --params <ID>` reads the store, not a file, so
/// it must never re-punish an artifact the store already accepted — C29,
/// "registered artifacts are never retroactively invalidated".
#[test]
fn graph_introspect_params_accepts_a_pre_331_store_artifact_with_a_bad_root_name() {
let dir = temp_cwd("introspect-store-readback-bad-root-name");
let (envelope_bytes, _e, built) = run(&["graph", "build"], SIGNAL_DOC_DESCRIBED);
assert!(built, "graph build produces the envelope");
let bad = envelope_bytes.replacen("\"name\":\"graph\"", "\"name\":\"../x\"", 1);
assert!(bad.contains("\"name\":\"../x\""), "the root-name replace landed: {bad}");
// Bypass the CLI's gate entirely: the registry's own store API is the
// simulated pre-#331 write path.
let id = aura_research::content_id_of(&bad);
let registry = aura_registry::Registry::open(dir.join("runs/runs.jsonl"));
registry
.put_blueprint(&id, &bad)
.expect("the registry's own store API has no root-name gate to bypass");
let (out, err, code) = run_in(&dir, &["graph", "introspect", "--params", &id]);
assert_eq!(code, Some(0), "store read-back stays ungated (C29): {out} {err}");
}
/// Property (#226, the render positional closes the #202 on-ramp family):
/// `aura graph <op-script.json>` (the render positional) accepts an op-script
/// document (a JSON array) exactly as it accepts a #155 blueprint envelope (a
@@ -58,7 +58,8 @@ fn sugar_verbs_name_their_document_shape() {
/// #323: `graph build --help` carries the op-list reference — the op kinds
/// and fields are learnable from the binary, not only from serde refusals.
/// #317: the `use` op joins the roster (nine -> ten).
/// #317: the `use` op joins the roster (nine -> ten). #331: the `name` op
/// joins the roster (ten -> eleven).
#[test]
fn graph_build_help_carries_the_op_reference() {
let (out, err, ok) = run(&["graph", "build", "--help"]);
@@ -74,6 +75,7 @@ fn graph_build_help_carries_the_op_reference() {
r#"{"op":"gang""#,
r#"{"op":"doc""#,
r#"{"op":"use""#,
r#"{"op":"name""#,
] {
assert!(out.contains(op), "op reference carries {op}: {out}");
}
+57
View File
@@ -81,3 +81,60 @@ fn measurement_run_honours_the_tap_selector() {
let expected = (1.0097_f64 + 1.0092) / 2.0;
assert!((got - expected).abs() < 1e-9, "last row: got {got}, expected {expected}");
}
/// #331 delta re-review, the third authored-file intake route: `dispatch_run`
/// loads a FRESH FILE via `blueprint_from_json` and, on this measurement arm,
/// feeds `signal.name()` straight into `bind_tap_plan` -> `TraceStore::begin_run`
/// (`trace_store.rs` joins the name unsanitized) — a hand-crafted envelope with
/// `"name":"../x"` used to reach `--tap fast_tap=record` and escape
/// `runs/traces/<name>/` for `runs/x/` (one level up, since `traces/../x`
/// normalizes there) instead of refusing at the same root-name choke point
/// `register`/`introspect --content-id <FILE>` already gate through. This pins
/// the closed route: exit nonzero (dispatch_run's own bad-input-file
/// convention, exit 2), the shared `name_gate_fault_prose` wording, and — the
/// property that actually matters — the escaped directory is never created.
#[test]
fn run_refuses_a_hand_crafted_envelope_with_a_bad_root_name_before_any_trace_write() {
let cwd = temp_cwd("bad-root-name");
let mut v: serde_json::Value =
serde_json::from_str(&measurement_blueprint_json()).expect("parse measurement blueprint");
v["blueprint"]["name"] = serde_json::json!("../x");
let bad = serde_json::to_string(&v).expect("re-serialize with a shape-violating root name");
let bp_path = cwd.join("bad-root-name.json");
std::fs::write(&bp_path, &bad).expect("write bad-root-name blueprint");
let out = Command::new(BIN)
.args(["run", bp_path.to_str().expect("utf-8 path"), "--tap", "fast_tap=record"])
.current_dir(&cwd)
.output()
.expect("spawn aura run");
assert_eq!(
out.status.code(),
Some(2),
"a shape-violating root name refuses at dispatch_run's own bad-file exit code: {:?} stderr={}",
out.status,
String::from_utf8_lossy(&out.stderr)
);
let stderr = String::from_utf8_lossy(&out.stderr);
assert!(
stderr.contains(r#"blueprint name "../x" is invalid"#),
"the shared name_gate_fault_prose wording names the offending root name: {stderr}"
);
assert!(
out.stdout.is_empty(),
"a refused run must not print a report: {:?}",
String::from_utf8_lossy(&out.stdout)
);
// The property that actually matters: the escaped write target (one level
// up from `traces/`, since `trace_store.rs` joins the name unsanitized)
// was never created — the gate fires before `begin_run` ever touches disk.
assert!(
!cwd.join("runs/x").exists(),
"the escaped trace directory must never be created"
);
assert!(
!cwd.join("runs/traces").exists(),
"no trace directory of any shape is written on a refused run"
);
}
+45
View File
@@ -195,6 +195,38 @@ pub struct GangMember {
pub name: String,
}
/// Deterministic shape gate for authored composite render names (#331) —
/// shape only, never content judgement (C29 discipline): non-empty, no path
/// separator (`/` or `\`), and not the special segments `.` or `..`. The rule
/// exists because the name keys a trace directory on disk unsanitized
/// (`traces/<name>/`, `crates/aura-registry/src/trace_store.rs`). Applied at
/// both data-borne birth routes for a name (the skeptic's two-seam finding):
/// the `Op::Name` op intake (`construction.rs`) and the CLI's
/// blueprint-envelope root-name intake — never at store read-back (C29:
/// registered artifacts are never retroactively invalidated), and never on
/// the Rust builder API (`GraphBuilder::new`, role-2 native authoring).
pub fn name_gate(name: &str) -> Result<(), NameGateFault> {
if name.is_empty() {
return Err(NameGateFault::Empty);
}
if name.contains('/') || name.contains('\\') {
return Err(NameGateFault::ContainsSeparator);
}
if name == "." || name == ".." {
return Err(NameGateFault::DotSegment);
}
Ok(())
}
/// The [`name_gate`] shape violation — WHY the name was refused; the caller
/// already holds the offending name (WHAT) to build a message from.
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum NameGateFault {
Empty,
ContainsSeparator,
DotSegment,
}
/// A reusable sub-graph fragment compiled away by inlining (C9/C23). It is **not**
/// a [`Node`]: it is never `eval`'d. It holds interior items (local indices),
/// interior edges (local indices), input roles (role `r` fans into the interior
@@ -1519,6 +1551,19 @@ mod tests {
use aura_strategy::Bias;
use std::sync::mpsc;
/// `name_gate` (#331) unit table: an ordinary name passes; an empty
/// name, one containing either path separator, and the two dot segments
/// each fault with their own fault variant.
#[test]
fn name_gate_accepts_ordinary_names_and_refuses_shape_violations() {
assert_eq!(name_gate("ny_momentum"), Ok(()));
assert_eq!(name_gate(""), Err(NameGateFault::Empty));
assert_eq!(name_gate("a/b"), Err(NameGateFault::ContainsSeparator));
assert_eq!(name_gate("a\\b"), Err(NameGateFault::ContainsSeparator));
assert_eq!(name_gate("."), Err(NameGateFault::DotSegment));
assert_eq!(name_gate(".."), Err(NameGateFault::DotSegment));
}
/// One knob fanning into two sibling open params passes the gate; the
/// value carries the gang table.
#[test]
+127 -2
View File
@@ -16,8 +16,8 @@ use std::collections::{HashMap, HashSet};
use aura_core::{ArgOpError, BindOpError, NodeSchema, PrimitiveBuilder, Scalar, ScalarKind};
use crate::blueprint::{
check_param_namespace_injective, edge_kind_check, validate_wiring, BlueprintNode, Composite,
Gang, GangMember, OutField, Role, Tap, TapWire,
check_param_namespace_injective, edge_kind_check, name_gate, validate_wiring, BlueprintNode,
Composite, Gang, GangMember, NameGateFault, OutField, Role, Tap, TapWire,
};
use crate::builder::{resolve_input_slot, resolve_output_field, PortResolveError};
use crate::harness::{Edge, Target};
@@ -64,6 +64,8 @@ pub enum Op {
/// `bind` is applied path-qualified (e.g. `"sma.length"`), after the
/// splice.
Use { ref_id: String, name: Option<String>, bind: Vec<(String, Scalar)> },
/// Set the composite's render name (#331). At most once per script.
Name { name: String },
}
/// A per-op construction fault, by-identifier so the cause names the op.
@@ -124,6 +126,13 @@ pub enum OpError {
/// resolves and fetches before replay, refusing there on a miss), but
/// total: a bare `replay`/`GraphSession` caller can still hit it.
UnknownSubgraph { ref_id: String },
/// A second `name` op — the render name is declared at most once (mirrors
/// `DuplicateDoc`).
DuplicateName,
/// A `name` op's `name` (or, at the CLI's blueprint-envelope intake, an
/// envelope's root name) failed the deterministic `name_gate` shape check
/// (#331). Carries the offending name and the shape fault detail.
BadName { name: String, fault: NameGateFault },
}
/// A per-op-fallible blueprint accumulator. Holds the same interior data a
@@ -153,6 +162,11 @@ pub struct GraphSession<'v> {
coverage: HashMap<(usize, usize), usize>,
gangs: Vec<Gang>,
doc: Option<String>,
/// At-most-once guard for `Op::Name` (#331), mirroring `doc`'s own
/// uniqueness posture but tracked separately (`name` itself is always
/// populated, seeded from `::new`'s `name` argument, so it cannot double
/// as its own "was it authored" flag).
name_authored: bool,
}
impl<'v> GraphSession<'v> {
@@ -182,6 +196,7 @@ impl<'v> GraphSession<'v> {
coverage: HashMap::new(),
gangs: Vec::new(),
doc: None,
name_authored: false,
}
}
@@ -205,9 +220,29 @@ impl<'v> GraphSession<'v> {
Ok(())
}
Op::Use { ref_id, name, bind } => self.use_subgraph(ref_id, name, bind),
Op::Name { name } => self.set_name(name),
}
}
/// `Op::Name` (#331): set the composite's render name — the op-script's
/// data-borne twin of `replay`'s seeded default name. Guarded by (a)
/// at-most-once (mirrors the `doc` op's uniqueness guard just above) and
/// (b) the shared deterministic shape gate (`name_gate`) — the SAME gate
/// the CLI's blueprint-envelope intake applies to a hand-edited
/// envelope's root name (the other data-borne birth route, skeptic
/// finding minuted on #331).
fn set_name(&mut self, name: String) -> Result<(), OpError> {
if self.name_authored {
return Err(OpError::DuplicateName);
}
if let Err(fault) = name_gate(&name) {
return Err(OpError::BadName { name, fault });
}
self.name = name;
self.name_authored = true;
Ok(())
}
/// `Op::Use` (#317): fetch a registered blueprint through the injected
/// `subgraph` resolver, rename it to the instance identifier, apply its
/// path-qualified `bind` entries, and push it as an ordinary
@@ -1002,6 +1037,72 @@ mod tests {
);
}
// ---- Op::Name (#331) -----------------------------------------------
/// A `name` op sets the finished composite's render name — the property
/// the whole op exists for.
#[test]
fn name_op_sets_the_finished_composite_name() {
let mut s = scaffold();
s.apply(Op::Name { name: "x".into() }).unwrap();
s.apply(Op::Feed { role: "price".into(), into: vec!["fast.series".into(), "slow.series".into()] })
.unwrap();
s.apply(Op::Connect { from: "fast.value".into(), to: "sub.lhs".into() }).unwrap();
s.apply(Op::Connect { from: "slow.value".into(), to: "sub.rhs".into() }).unwrap();
s.apply(Op::Expose { from: "sub.value".into(), as_name: "out".into() }).unwrap();
let c = s.finish().expect("finishes");
assert_eq!(c.name(), "x");
}
/// Byte-compat (#331 AC1): a script that never authors a `name` op keeps
/// the seed name `replay`/`GraphSession::new` was given — the CLI's own
/// seed stays `"graph"`, unchanged by this op's addition.
#[test]
fn no_name_op_keeps_the_seeded_replay_name() {
let ops = vec![
Op::Source { role: "price".into(), kind: ScalarKind::F64 },
Op::Add { type_id: "SMA".into(), as_name: Some("a".into()), args: vec![], bind: vec![] },
Op::Feed { role: "price".into(), into: vec!["a.series".into()] },
Op::Expose { from: "a.value".into(), as_name: "out".into() },
];
let c = super::replay("graph", ops, &std_vocabulary, &|_: &str| None).expect("replays");
assert_eq!(c.name(), "graph");
}
/// A second `name` op is a fault, not a silent last-wins overwrite —
/// mirrors the `doc` op's own at-most-once guard.
#[test]
fn duplicate_name_op_is_refused() {
let mut s = session("g");
s.apply(Op::Name { name: "x".into() }).unwrap();
assert_eq!(
s.apply(Op::Name { name: "y".into() }),
Err(OpError::DuplicateName)
);
}
/// Shape violations (#331): empty, containing a path separator, and the
/// dot segments all fault at the op via the shared `name_gate`.
#[test]
fn name_op_shape_violations_are_refused() {
use crate::blueprint::NameGateFault;
let mut s = session("g");
assert_eq!(
s.apply(Op::Name { name: "".into() }),
Err(OpError::BadName { name: "".into(), fault: NameGateFault::Empty })
);
let mut s = session("g");
assert_eq!(
s.apply(Op::Name { name: "a/b".into() }),
Err(OpError::BadName { name: "a/b".into(), fault: NameGateFault::ContainsSeparator })
);
let mut s = session("g");
assert_eq!(
s.apply(Op::Name { name: "..".into() }),
Err(OpError::BadName { name: "..".into(), fault: NameGateFault::DotSegment })
);
}
#[test]
fn finish_reports_incomplete_on_unwired_slot() {
// a fully-named scaffold missing the sub.rhs connection -> holistic 0-arm
@@ -1501,6 +1602,30 @@ mod tests {
);
}
/// #331 ratifying pin (spec test 5): `use_subgraph`'s instance-identifier
/// default (`name: None` -> the fetched composite's own `name()`,
/// `construction.rs:221-230` — code-inspected, previously untested since
/// every other test in this section passes an explicit instance name)
/// actually reaches the session's identifier namespace: `use_fixture()`'s
/// own name is "fixture", so with no instance name the spliced ports
/// resolve under "fixture.*", not some other placeholder.
#[test]
fn use_op_with_no_name_defaults_instance_identifier_to_the_fetched_composites_name() {
let subgraph = |_: &str| Some(use_fixture());
let mut s = GraphSession::new("g", &std_vocabulary, &subgraph);
s.apply(Op::Source { role: "price".into(), kind: ScalarKind::F64 }).unwrap();
assert_eq!(
s.apply(Op::Use { ref_id: "fixture-id".into(), name: None, bind: vec![] }),
Ok(())
);
s.apply(Op::Feed { role: "price".into(), into: vec!["fixture.price".into()] }).unwrap();
assert_eq!(
s.apply(Op::Expose { from: "fixture.out".into(), as_name: "bias".into() }),
Ok(()),
"the spliced instance resolves under \"fixture\" — the fetched composite's own name"
);
}
/// A post-splice `bind` path with no matching open param anywhere in the
/// instance faults through the existing bind-fault shape (`BadParam`),
/// naming the instance and the qualified within-instance path (#317).
+3 -2
View File
@@ -52,8 +52,9 @@ mod sweep;
mod walkforward;
pub use blueprint::{
BindError, Binder, BlueprintNode, BoundSpec, CompileError, Composite, Gang, GangFault,
GangMember, OutField, RandomBinder, ReopenError, Role, SweepBinder, Tap, TapWire,
name_gate, BindError, Binder, BlueprintNode, BoundSpec, CompileError, Composite, Gang,
GangFault, GangMember, NameGateFault, OutField, RandomBinder, ReopenError, Role, SweepBinder,
Tap, TapWire,
};
pub use blueprint_serde::{
blueprint_from_json, blueprint_identity_json, blueprint_to_json, ArgData, BlueprintDoc,