Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions graph/bundle/SCHEMA.md
Original file line number Diff line number Diff line change
Expand Up @@ -442,7 +442,7 @@ One row per place a call is written (or, for a synthesised edge, the construct t
- **typescript** — end_line / end_column come from the expression row; the call-site row itself records only the start.
- **javascript** — caller_id is the parser's enclosing method, or the module initializer for top-level code. end_line / end_column come from the expression row. `require()` is a module edge, not a call site.
- **typescript** — PROPERTY_READ and PROPERTY_WRITE rows are accessor invocations with no written call: the site is the property-access expression that runs the getter or setter, positioned from the expressions table, and callee_name is NULL because nothing was written; the accessor's name is on the callee's methods row. Filter them out with kind NOT IN (…) when counting calls.
- **python** — PROPERTY_READ, PROPERTY_WRITE, CONTEXT_MANAGER, ITERATION_PROTOCOL, BUILTIN_PROTOCOL, OPERATOR_PROTOCOL, TRUTH_PROTOCOL, METACLASS_CREATION and DYNAMIC_CALL rows are protocol or indirect edges with no written call: their site is the expression that triggers them, and callee_name is always NULL because nothing was written. SUBSCRIPT_CALL is NULL only when the subscript is not a written name (measured 206 of 337 rows on a Python subject). Filter them out with kind NOT IN (…) when counting calls.
- **python** — PROPERTY_READ, PROPERTY_WRITE, CONTEXT_MANAGER, ITERATION_PROTOCOL, BUILTIN_PROTOCOL, OPERATOR_PROTOCOL, TRUTH_PROTOCOL, METHOD_REFERENCE, METACLASS_CREATION and DYNAMIC_CALL rows are protocol or indirect edges with no written call: their site is the expression that triggers them, and callee_name is always NULL because nothing was written. SUBSCRIPT_CALL is NULL only when the subscript is not a written name (measured 206 of 337 rows on a Python subject). Filter them out with kind NOT IN (…) when counting calls.
- **python** — The id is an EXPRESSION hash for a written call; a DECORATOR hash (PY_DECORATOR_…) for DECORATOR_APPLICATION and DECORATOR_* sites, positioned at the decorator line; and the class's TYPE hash for METACLASS_CREATION, positioned at the class declaration.

### `call_edges`
Expand Down Expand Up @@ -541,6 +541,7 @@ THE GRAPH. One row per (site, resolved target). A site with N possible targets h
| `SUBSCRIPT_PROTOCOL` | python | `x[k]` runs `__getitem__` (and `x[k] = v` / `del x[k]` the setter and deleter) of the receiver's class. No written call; the site is the subscript expression. Its own kind so it is never counted as a written call. |
| `BUILTIN_PROTOCOL` | python | `repr(x)`, `str(x)`, `len(x)`, `hash(x)`, `bool(x)`, `iter(x)`, `next(x)`, `abs(x)`, `format(x)` and `reversed(x)` run the matching dunder of the argument's class (str falls back to `__repr__`, bool to `__len__`). The written call is to the builtin; this edge is the dunder it runs. The site is the argument expression. Its own kind so it is never counted as a written call. |
| `OPERATOR_PROTOCOL` | python | An operator runs a dunder of its operands: `a + b` the left operand's `__add__` and the right one's `__radd__`, a comparison the left slot and the right one's mirror (`a < b` -> `b.__gt__`, `!=` falls back to `__eq__`), `x in c` the right operand's `__contains__` (or `__iter__`), `-a`/`+a`/`~a` its `__neg__`/`__pos__`/`__invert__`. No written call; the site is the operator expression. |
| `METHOD_REFERENCE` | python | A method or function READ AS A VALUE, not called: an argument (`pool.submit(self.work, x)`, `map(Cls.convert, rows)`, `partial(svc.cancel, oid)`), a keyword (`Thread(target=svc.run)`), an element, a stored or returned value. The target is what the expression resolves to (a typed receiver through the MRO, a class, a scoped name, a module member); an untyped receiver gives no row. Always tier callback_registered; the site is the reference expression (call-edge-generation/method-reference.dl). |
| `TRUTH_PROTOCOL` | python | A truth test runs the tested value's `__bool__`, or `__len__` when its class has none: `if x:`, `while x:`, `assert x`, `not x`, a conditional or comprehension condition, and an operand of `and`/`or` (the right one only when the whole expression is tested). No written call; the site is the tested expression. |

**`call_edges.tier` values**
Expand All @@ -564,6 +565,7 @@ THE GRAPH. One row per (site, resolved target). A site with N possible targets h
| `dynamic_terminal` | javascript | `obj[expr]()`, `eval`, `import()`: no static target by construction; callee is NULL. |
| `fan_capped` | javascript, java, csharp | More targets than --dispatch-cap: the set was refused rather than emitted. JavaScript: callee is NULL. Java and C#: callee is the declared base method the fan would have started from; dispatch-capped-sites.csv carries the refused count. |
| `callback_registered` | javascript, typescript | The site HANDS the callee this function (`xs.forEach(f)`, `p.then(f)`, `emitter.on('x', h)`, `setTimeout(f)`), which may invoke it. Not the site's own callee; a reachability edge, labelled so it is never read as a resolved call. |
| `callback_registered` | python | A method or function named as a VALUE and handed over, stored or returned (`pool.submit(self.work)`, `Thread(target=svc.run)`, `sorted(xs, key=self.rank)`, `[h.on_event]`), always kind METHOD_REFERENCE. Its target is resolved the way a call on the same receiver would be; whoever receives the value calls it. A reachability edge, labelled so it is never read as a resolved call. |
| `event_dispatch` | javascript | `x.emit('name')` reaching a handler registered by `x.on('name', h)` on a value x may hold — name-sensitive for literal names, every handler on that value for a computed one. |
| `event_dispatch` | typescript | A publish on a project class's keyed registry reaching each function filed under the same key: `bus.publish('x', p)` where `publish` (or a method it passes the key to) calls what `this.handlers.get(name)` holds, to the function a `bus.subscribe('x', f)` stored there, or an entry `{ ['x']: f }` of a table a subscribe-all method files by key. A literal, a const or a const-object member key; a key only known at run time matches every key, and a constant the dispatch also looks up (a wildcard) matches every publish. The site is the publish call (call-edge-generation/keyed_registry.dl). |
| `event_dispatch` | java | A Spring application event: `publishEvent(e)` reaching each listener (`@EventListener`, `@TransactionalEventListener`, `ApplicationListener<E>.onApplicationEvent`) whose declared event type e's static type is, or is a subtype of. Added beside the publishEvent boundary row, never in place of it (call-edge-generation/event_dispatch.dl). Also a JPA entity write (`save`, `persist`, `merge`, `delete` on a repository or EntityManager) reaching the `@PrePersist`/`@PreUpdate`/`@PreRemove`-style callbacks of the written entity's type and of the listeners `@EntityListeners` names on it or a superclass (kind entity_callback; call-edge-generation/entity_lifecycle.dl). |
Expand All @@ -586,7 +588,7 @@ THE GRAPH. One row per (site, resolved target). A site with N possible targets h
- **python** — A `boundary_lib` edge may point at a builtin (callee_provenance builtin, callee_label `builtin:NAME`) or at an unstaged import path (callee_provenance external) — neither has a methods row.
- **java** — A `boundary_lib` edge with callee_provenance external names a method of an ancestor type no staged IR declares (callee_label `external:<type>.<name>`, no methods row). A site whose receiver is declared as such a type is multi_inferred even with one client override: the platform method itself, and the platform's own subclasses, are the other possible targets. Stage the library to replace the label with the real method.
- **python** — The reason a site is ambiguous_unknown is exported per site in ext_call_site_unresolved (site, caller, reason, detail).
- **all** — THE TRUST LINE, and it is not the same set of tiers in every language. RESOLVED (callee_method_id is set): known_edge and multi_inferred in every language, and boundary_lib where the library is staged (--library) — without it boundary_lib names the target in callee_label and leaves callee_method_id NULL; ALSO ambient_terminal in TypeScript, fan_capped in Java and C# (the declared base, the fan refused), and runtime_observed in C#. HANDED OVER (callee set, but the site passes the function rather than calling it): callback_registered and event_dispatch in JavaScript and TypeScript; event_dispatch in C# too, where a mediator Send or Publish runs the handler for the request type, beside the row for the site itself. CORRECT END (callee NULL, and nothing is missing): intrinsic_terminal in TypeScript; ambient_terminal, implicit_constructor and dynamic_terminal in JavaScript; known_implicit_ctor, known_builtin_operator and boundary_generated (callee_label set) in C#. BLIND SPOT (callee NULL; exactly the tiers named `ambiguous_*`, which are what unresolved_sites holds): ambiguous_unknown everywhere, ALSO ambiguous_anon in Java and ambiguous_dynamic in C#. CAPPED (callee NULL, not in unresolved_sites): fan_capped in JavaScript. Python emits only the four shared tiers. A filter written as `tier IN (known_edge, multi_inferred)` therefore drops resolved edges in every language but Python — derive the set from this note or from unresolved_sites, never from a hardcoded list.
- **all** — THE TRUST LINE, and it is not the same set of tiers in every language. RESOLVED (callee_method_id is set): known_edge and multi_inferred in every language, and boundary_lib where the library is staged (--library) — without it boundary_lib names the target in callee_label and leaves callee_method_id NULL; ALSO ambient_terminal in TypeScript, fan_capped in Java and C# (the declared base, the fan refused), and runtime_observed in C#. HANDED OVER (callee set, but the site passes the function rather than calling it): callback_registered and event_dispatch in JavaScript and TypeScript, callback_registered (kind METHOD_REFERENCE) in Python; event_dispatch in C# too, where a mediator Send or Publish runs the handler for the request type, beside the row for the site itself. CORRECT END (callee NULL, and nothing is missing): intrinsic_terminal in TypeScript; ambient_terminal, implicit_constructor and dynamic_terminal in JavaScript; known_implicit_ctor, known_builtin_operator and boundary_generated (callee_label set) in C#. BLIND SPOT (callee NULL; exactly the tiers named `ambiguous_*`, which are what unresolved_sites holds): ambiguous_unknown everywhere, ALSO ambiguous_anon in Java and ambiguous_dynamic in C#. CAPPED (callee NULL, not in unresolved_sites): fan_capped in JavaScript. Python emits the four shared tiers and callback_registered. A filter written as `tier IN (known_edge, multi_inferred)` therefore drops resolved edges in every language but Python — derive the set from this note or from unresolved_sites, never from a hardcoded list.
- **java** — A multi_inferred fan is CHA-wide: it is every override the hierarchy admits, bounded only by the dispatch cap. type_instantiated is computed and exported but NOT read by any rule, so the fan is not narrowed to types the program constructs. Narrow it yourself by joining dispatch_candidates to type_instantiated — see the dispatch_envelope_of query. A receiver is ALSO typed by what flows into it (a local's initializer, the arguments callers pass to a parameter, the receivers callers invoke a method on for its `this`), and each flow-in type resolves its member directly, outside the fan: that is why a `fan_capped` site still carries edges, and why they are the types the program was seen to hand over, not the whole hierarchy.
- **typescript** — A multi_inferred fan is CHA-wide, as in Java: type_instantiated is computed and exported but NOT read by any rule. The fan also has sources that are not virtual dispatch at all — an overload set or a union-typed receiver produces one too.
- **python** — A multi_inferred fan IS narrowed by the instantiation set: type_instantiated_reachable (the constructed classes and their bases) bounds dispatch in resolution/dispatch.dl. Python is the only front end where that narrowing is applied, so a fan here is tighter than the same shape would be in Java or TypeScript.
Expand Down
Loading
Loading