mib_rs/

lib.rs

//! SNMP MIB parsing, resolution, query, and tooling APIs.
//!
//! # What is a MIB?
//!
//! A MIB (Management Information Base) is a text file that describes the
//! structure of data available from an SNMP-managed device. Each piece of
//! data (a counter, a name, a status flag, a table row) is identified by an
//! OID (Object Identifier), a dotted-decimal path like `1.3.6.1.2.1.1.1`.
//! MIB files give those numeric OIDs human-readable names, types, and
//! descriptions, so instead of `1.3.6.1.2.1.1.1` you can say `sysDescr`.
//!
//! MIBs are written in a language called SMI (Structure of Management
//! Information), which has two versions: SMIv1 (RFC 1155/1212) and SMIv2
//! (RFC 2578/2579/2580). This crate handles both transparently.
//!
//! # API Layers
//!
//! Most callers should use the **handle API**: start with [`Loader`],
//! get a [`Mib`], and navigate the resolved model through borrowed
//! handle types ([`Node`], [`Object`], [`Type`], [`Module`],
//! [`Notification`], [`Group`], [`Compliance`], [`Capability`]).
//! Handles wrap an arena ID and a `&Mib` reference. Methods on
//! handles return further handles, so typical usage looks like
//! `object.ty()?.effective_base()` without touching IDs directly.
//! The handle API covers OID resolution, type chain introspection,
//! table/index navigation, module iteration, diagnostics, and
//! everything else documented in the sections below.
//!
//! Every handle exposes its arena ID via `.id()`. IDs are
//! `Copy + Eq + Hash + Ord`, so you can store them in collections
//! for deduplication or cross-referencing, then convert back to
//! handles with `mib.*_by_id()` when you need to query again.
//!
//! [`Mib`] also has query methods that work with IDs directly:
//! [`Mib::modules_defining`] and [`Mib::modules_importing`] find
//! modules by symbol name, [`Mib::objects_by_base_type`] and
//! [`Mib::objects_by_type_name`] filter objects by type, and
//! [`Mib::available_symbols`] returns everything visible in a
//! module's scope (own definitions plus resolved imports).
//!
//! ## Raw data access
//!
//! [`Mib::raw()`] returns a [`RawMib`](raw::RawMib) view that
//! exposes the arena-backed data records directly. This is useful
//! when you need things the handle API doesn't surface:
//!
//! - Per-clause source spans on [`ObjectData`](raw::ObjectData)
//!   and [`TypeData`](raw::TypeData) (e.g. `syntax_span`,
//!   `access_span`) for pointing diagnostics at specific clauses.
//! - Import metadata ([`ModuleData::is_import_used`](raw::ModuleData::is_import_used),
//!   [`ModuleData::import_source`](raw::ModuleData::import_source)).
//! - Symbolic OID references via `oid_refs()` on entity records.
//! - Bulk arena slices (`raw.*_slice()`) for batch analysis.
//!
//! See the [`raw`] module and the `raw` example.
//!
//! ## Compiler pipeline
//!
//! The [`ast`], [`parser`], [`lower`], [`ir`], and [`token`]
//! modules expose pre-resolution stages for callers that need
//! syntax-aware analysis before full resolution. The parser
//! produces partial ASTs from broken input, which matters for
//! editor integration where the user is mid-edit. Token types
//! carry classification predicates for syntax highlighting.
//! See the [`compile`] module and the `tokens` example.
//!
//! # Loading MIBs
//!
//! Use [`Loader`] to configure sources, select modules, and run the pipeline:
//!
//! ```rust
//! use mib_rs::{BaseType, Loader};
//!
//! fn example_mib() -> mib_rs::Mib {
//!     let source = mib_rs::source::memory(
//!         "DOC-EXAMPLE-MIB",
//!         r#"DOC-EXAMPLE-MIB DEFINITIONS ::= BEGIN
//! IMPORTS
//!     MODULE-IDENTITY, OBJECT-TYPE, Integer32, enterprises
//!         FROM SNMPv2-SMI
//!     TEXTUAL-CONVENTION, DisplayString
//!         FROM SNMPv2-TC;
//!
//! docExampleMib MODULE-IDENTITY
//!     LAST-UPDATED "202603120000Z"
//!     ORGANIZATION "Example"
//!     CONTACT-INFO "Example"
//!     DESCRIPTION "Example module used in crate docs."
//!     ::= { enterprises 99999 }
//!
//! DocName ::= TEXTUAL-CONVENTION
//!     DISPLAY-HINT "255a"
//!     STATUS current
//!     DESCRIPTION "Example display string type."
//!     SYNTAX DisplayString (SIZE (0..255))
//!
//! docScalars OBJECT IDENTIFIER ::= { docExampleMib 1 }
//! docTables OBJECT IDENTIFIER ::= { docExampleMib 2 }
//!
//! docDeviceName OBJECT-TYPE
//!     SYNTAX DocName
//!     MAX-ACCESS read-only
//!     STATUS current
//!     DESCRIPTION "A scalar object."
//!     ::= { docScalars 1 }
//!
//! docTable OBJECT-TYPE
//!     SYNTAX SEQUENCE OF DocEntry
//!     MAX-ACCESS not-accessible
//!     STATUS current
//!     DESCRIPTION "Example table."
//!     ::= { docTables 1 }
//!
//! docEntry OBJECT-TYPE
//!     SYNTAX DocEntry
//!     MAX-ACCESS not-accessible
//!     STATUS current
//!     DESCRIPTION "Example row."
//!     INDEX { docIndex }
//!     ::= { docTable 1 }
//!
//! DocEntry ::= SEQUENCE {
//!     docIndex Integer32,
//!     docDescr DisplayString
//! }
//!
//! docIndex OBJECT-TYPE
//!     SYNTAX Integer32 (1..2147483647)
//!     MAX-ACCESS not-accessible
//!     STATUS current
//!     DESCRIPTION "Example index."
//!     ::= { docEntry 1 }
//!
//! docDescr OBJECT-TYPE
//!     SYNTAX DisplayString
//!     MAX-ACCESS read-only
//!     STATUS current
//!     DESCRIPTION "Example column."
//!     ::= { docEntry 2 }
//!
//! END
//! "#,
//!     );
//!
//!     Loader::new()
//!         .source(source)
//!         .modules(["DOC-EXAMPLE-MIB"])
//!         .load()
//!         .expect("example MIB should load")
//! }
//!
//! let mib = example_mib();
//! let object = mib.object("docDeviceName").expect("object should exist");
//! let ty = object.ty().expect("object should have a type");
//!
//! assert_eq!(object.name(), "docDeviceName");
//! assert_eq!(ty.name(), "DocName");
//! assert_eq!(ty.effective_base(), BaseType::OctetString);
//! assert_eq!(ty.effective_display_hint(), "255a");
//! ```
//!
//! # OIDs and Resolution
//!
//! Every named element in a MIB has an OID, a path through a global tree
//! shared by all SNMP devices. OIDs are written as dotted decimal
//! (`1.3.6.1.2.1.1.1`) or symbolically (`sysDescr`). The tree is
//! hierarchical: `enterprises` is `1.3.6.1.4.1`, and a vendor's subtree
//! hangs beneath that.
//!
//! **Instance OIDs** extend a base OID with a suffix that identifies a
//! specific value. For a scalar like `sysDescr`, the instance is always
//! `sysDescr.0`. For table columns, the suffix encodes the row's index
//! values, e.g. `ifDescr.7` for interface 7.
//!
//! This crate resolves both directions: name to numeric OID, and numeric
//! OID back to its closest named node.
//!
//! ```rust
//! fn example_mib() -> mib_rs::Mib {
//!     let source = mib_rs::source::memory(
//!         "DOC-EXAMPLE-MIB",
//!         include_bytes!("../tests/data/doc-example-mib.txt").as_slice(),
//!     );
//!
//!     mib_rs::Loader::new()
//!         .source(source)
//!         .modules(["DOC-EXAMPLE-MIB"])
//!         .load()
//!         .expect("example MIB should load")
//! }
//!
//! let mib = example_mib();
//!
//! let column_oid = mib.resolve_oid("docDescr").expect("OID should resolve");
//! assert_eq!(column_oid.to_string(), "1.3.6.1.4.1.99999.2.1.1.2");
//!
//! let node = mib
//!     .exact_node_by_oid(&column_oid)
//!     .expect("exact node should exist");
//! assert_eq!(node.name(), "docDescr");
//!
//! let instance_node = mib
//!     .resolve_node("docDescr.7")
//!     .expect("instance OID should resolve to its base node");
//! assert_eq!(instance_node.name(), "docDescr");
//!
//! let instance_oid = mib.resolve_oid("docDescr.7").expect("instance OID should resolve");
//! assert_eq!(instance_oid.to_string(), "1.3.6.1.4.1.99999.2.1.1.2.7");
//! assert_eq!(mib.lookup_oid(&instance_oid).name(), "docDescr");
//! assert_eq!(mib.lookup_oid(&"1.3.6.1.4.1.99999.2.1.1.2.99".parse().unwrap()).name(), "docDescr");
//! ```
//!
//! # Tables and Indexes
//!
//! SNMP models tabular data as three nested objects:
//!
//! - A **table** (`SEQUENCE OF`) is a container, not directly readable.
//! - A **row** (entry) represents one row, also not directly readable.
//!   It declares which columns are **index** columns, whose values
//!   together form the instance suffix that identifies each row.
//! - **Columns** are the actual readable/writable values. Each column's
//!   full OID is the column OID plus the index suffix.
//!
//! For example, `ifTable` contains `ifEntry` rows indexed by `ifIndex`.
//! The column `ifDescr` for interface 7 has OID `ifDescr.7` (i.e.
//! the column's base OID with the index value `7` appended).
//!
//! ## AUGMENTS
//!
//! Some rows use `AUGMENTS` instead of `INDEX`. An augmenting row
//! extends another table's rows with additional columns, sharing the
//! same index structure. For example, `ifXEntry AUGMENTS ifEntry`
//! adds columns like `ifHighSpeed` to each `ifEntry` row, using the
//! same `ifIndex` to identify rows. Use [`Object::augments`] to find
//! the target row and [`Object::augmented_by`] to find extending rows.
//! [`Object::effective_indexes`] follows the augment chain
//! automatically, returning the inherited index list.
//!
//! ## Index encoding
//!
//! Each index component has an [`IndexEncoding`] that describes how
//! its value maps to OID sub-identifiers in the instance suffix.
//! Integer indexes use a single sub-identifier. Fixed-length strings
//! (with a single-value SIZE constraint) use one sub-identifier per
//! octet. Variable-length strings are length-prefixed. The `IMPLIED`
//! keyword omits the length prefix, relying on the index being the
//! last component. [`Index::encoding`] returns the derived encoding.
//!
//! Use [`mib::index::decode_suffix`] to decode raw OID suffix arcs
//! into typed [`IndexValue`]s, or call [`OidLookup::decode_indexes`]
//! for the common case of processing a varbind OID.
//!
//! Use [`Object::is_table`], [`Object::is_row`], [`Object::is_column`],
//! and [`Object::is_scalar`] to distinguish these, or use the filtered
//! iterators like [`Mib::tables`] and [`Mib::scalars`].
//!
//! ```rust
//! fn example_mib() -> mib_rs::Mib {
//!     let source = mib_rs::source::memory(
//!         "DOC-EXAMPLE-MIB",
//!         include_bytes!("../tests/data/doc-example-mib.txt").as_slice(),
//!     );
//!
//!     mib_rs::Loader::new()
//!         .source(source)
//!         .modules(["DOC-EXAMPLE-MIB"])
//!         .load()
//!         .expect("example MIB should load")
//! }
//!
//! let mib = example_mib();
//! let table = mib.object("docTable").expect("table should exist");
//! let row = table.row().expect("table should have a row");
//!
//! let column_names: Vec<_> = table.columns().map(|col| col.name()).collect();
//! assert_eq!(column_names, vec!["docIndex", "docDescr"]);
//!
//! let indexes: Vec<_> = row.effective_indexes().collect();
//! assert_eq!(indexes.len(), 1);
//! assert_eq!(indexes[0].row().name(), "docEntry");
//! let index_object = indexes[0].object().expect("index object");
//! let index_type = indexes[0].ty().expect("index type");
//! assert_eq!(index_object.name(), "docIndex");
//! assert_eq!(indexes[0].name(), "docIndex");
//! assert_eq!(index_type.name(), "Integer32");
//! ```
//!
//! # Modules
//!
//! A MIB file contains one module (e.g. `IF-MIB`, `SNMPv2-MIB`). Modules
//! import symbols from other modules, so loading one module typically
//! pulls in its dependencies automatically.
//!
//! ## Base modules
//!
//! Seven **base modules** are built into the library and always available:
//!
//! | Module | SMI version | Defines |
//! |--------|-------------|---------|
//! | `SNMPv2-SMI` | SMIv2 | Core types (`Integer32`, `Counter32`, etc.), OID roots (`internet`, `enterprises`, `mib-2`), macros (`MODULE-IDENTITY`, `OBJECT-TYPE`, `NOTIFICATION-TYPE`, `OBJECT-IDENTITY`) |
//! | `SNMPv2-TC` | SMIv2 | `TEXTUAL-CONVENTION` macro, standard TCs (`DisplayString`, `TruthValue`, `RowStatus`, etc.) |
//! | `SNMPv2-CONF` | SMIv2 | Conformance macros (`MODULE-COMPLIANCE`, `OBJECT-GROUP`, `NOTIFICATION-GROUP`, `AGENT-CAPABILITIES`) |
//! | `RFC1155-SMI` | SMIv1 | SMIv1 base types and OID roots |
//! | `RFC1065-SMI` | SMIv1 | Earlier SMIv1 base (predecessor to RFC1155-SMI) |
//! | `RFC-1212` | SMIv1 | SMIv1 `OBJECT-TYPE` macro definition |
//! | `RFC-1215` | SMIv1 | SMIv1 `TRAP-TYPE` macro definition |
//!
//! These modules define the SMI language itself, specifically the ASN.1
//! macros (`OBJECT-TYPE`, `MODULE-IDENTITY`, `TEXTUAL-CONVENTION`, etc.)
//! that all other MIB modules use. The library constructs them
//! programmatically rather than parsing them from files, because they
//! contain ASN.1 MACRO definitions that require a general ASN.1 macro
//! parser to process. Since RFC 2578 Section 3 explicitly prohibits
//! user-defined macros in MIB modules ("Additional ASN.1 macros must not
//! be defined in SMIv2 information modules"), the library only needs to
//! handle the fixed set of macros defined by the SMI RFCs.
//!
//! Implications for users:
//!
//! - **No files needed:** You do not need to supply these modules as source
//!   files. If they exist on disk in a source directory, the synthetic
//!   versions take priority and the files are not parsed.
//! - **Always present:** Base modules are included in every loaded [`Mib`],
//!   even if nothing imports them. Use [`Module::is_base`] to distinguish
//!   them from user-supplied modules (e.g. when iterating modules).
//! - **No source spans:** Definitions from base modules carry synthetic
//!   span values ([`Span::SYNTHETIC`](crate::types::Span::SYNTHETIC))
//!   rather than real byte offsets, since there is no parsed source text.
//!   The `source_path` for base modules is empty.
//! - **Included in iteration:** [`Mib::modules`], [`Mib::objects`],
//!   [`Mib::types`], and [`Mib::nodes`] all include base module content.
//!   Filter with [`Module::is_base`] when you only want user-supplied
//!   definitions. Module-scoped iterators (e.g. `module.objects()`) are
//!   naturally limited to a single module.
//!
//! ## OID ownership
//!
//! Several base modules define overlapping OID trees. For example, both
//! `RFC1155-SMI` (SMIv1) and `SNMPv2-SMI` (SMIv2) define `internet`,
//! `enterprises`, and other well-known roots. When multiple modules
//! register the same OID, the resolver determines which module "owns"
//! the node using these tiebreakers, in order:
//!
//! - Base modules take priority over user modules.
//! - SMIv2 modules are preferred over SMIv1.
//! - Among modules with the same SMI version, newer `LAST-UPDATED`
//!   timestamps win.
//! - Lexicographic module name as a final deterministic fallback.
//!
//! In practice this means `SNMPv2-SMI` owns nodes like `enterprises`
//! even though `RFC1155-SMI` also defines them. [`Node::module`] returns
//! the winning module. Both modules still function normally for imports,
//! so SMIv1 MIBs that `IMPORTS ... FROM RFC1155-SMI` continue to work.
//!
//! Use [`Module`] handles to scope lookups and iteration to a single
//! module:
//!
//! ```rust
//! fn example_mib() -> mib_rs::Mib {
//!     let source = mib_rs::source::memory(
//!         "DOC-EXAMPLE-MIB",
//!         include_bytes!("../tests/data/doc-example-mib.txt").as_slice(),
//!     );
//!
//!     mib_rs::Loader::new()
//!         .source(source)
//!         .modules(["DOC-EXAMPLE-MIB"])
//!         .load()
//!         .expect("example MIB should load")
//! }
//!
//! let mib = example_mib();
//! let module = mib.module("DOC-EXAMPLE-MIB").expect("module should exist");
//!
//! assert_eq!(module.object("docDeviceName").unwrap().module(), Some(module));
//!
//! let object_names: Vec<_> = module.objects().map(|obj| obj.name()).collect();
//! assert!(object_names.contains(&"docTable"));
//! assert!(object_names.contains(&"docDescr"));
//!
//! let type_names: Vec<_> = module.types().map(|ty| ty.name()).collect();
//! assert!(type_names.contains(&"DocName"));
//! ```
//!
//! # Notifications and Conformance
//!
//! Beyond objects and types, SMI defines several constructs for
//! event reporting and conformance testing:
//!
//! - **NOTIFICATION-TYPE** (SMIv2) / **TRAP-TYPE** (SMIv1) - defines
//!   an asynchronous event an agent can send. Each notification lists
//!   the objects it carries as payload via its OBJECTS clause. SMIv1
//!   traps additionally carry an enterprise OID and trap number.
//!   See [`Notification`] and [`Notification::objects`].
//!
//! - **OBJECT-GROUP** / **NOTIFICATION-GROUP** - bundles related
//!   objects or notifications into a named set. Groups are the unit
//!   of conformance: a compliance statement says "you must implement
//!   these groups". See [`Group`] and [`Group::members`].
//!
//! - **MODULE-COMPLIANCE** - declares which groups a compliant
//!   implementation must support, with optional per-object refinements
//!   that can narrow syntax or access requirements. See [`Compliance`].
//!
//! - **AGENT-CAPABILITIES** - declares what an actual agent
//!   implementation supports, including which groups it includes and
//!   any per-object variations (restricted syntax, different defaults).
//!   See [`Capability`].
//!
//! These are less commonly needed than objects and types, but matter
//! for MIB validation tooling, compliance checking, and understanding
//! which objects are required vs optional. The `notifications` example
//! demonstrates querying all four.
//!
//! # Query Formats
//!
//! Once a MIB is loaded, you can look up nodes and OIDs using several
//! formats. Qualified names (`MODULE::name`) are useful when multiple
//! modules define the same name. [`Mib::resolve_oid`],
//! [`Mib::resolve_node`], and [`RawMib::resolve`](raw::RawMib::resolve) all accept these forms:
//!
//! | Form | Example | Description |
//! |------|---------|-------------|
//! | Plain name | `sysDescr` | Looks up by object/node name across all modules |
//! | Qualified name | `SNMPv2-MIB::sysDescr` | Scoped to a specific module |
//! | Instance OID | `ifDescr.7` | Name with numeric suffix appended |
//! | Numeric OID | `1.3.6.1.2.1.1.1` | Dotted decimal, leading dot optional |
//!
//! For instance OIDs (both symbolic and numeric), [`Mib::resolve_node`] returns
//! the deepest matching tree node, while [`Mib::resolve_oid`] returns the full
//! numeric OID with the suffix included.
//!
//! [`Mib::format_oid`] converts a numeric [`Oid`] back to `MODULE::name.suffix`
//! form using longest-prefix matching.
//!
//! # Sources
//!
//! Sources tell the loader where to find MIB files. For testing and
//! embedding, use in-memory sources. For production use, point at
//! directories on disk or use system path auto-discovery to find MIBs
//! installed by net-snmp or libsmi. The [`source`] module has several
//! constructors:
//!
//! | Constructor | Description |
//! |-------------|-------------|
//! | [`source::file()`] | Single file on disk, module name auto-detected |
//! | [`source::files()`] | Multiple files on disk, module names auto-detected |
//! | [`source::dir`] | Recursively indexes a directory tree on disk |
//! | [`source::dirs`] | Chains multiple directory trees |
//! | [`source::memory`] | Single in-memory module (for tests or embedding) |
//! | [`source::memory_modules`] | Multiple in-memory modules |
//! | [`source::chain`] | Combines multiple sources; first match wins |
//!
//! [`Loader::system_paths`](load::Loader::system_paths) auto-discovers
//! net-snmp and libsmi MIB directories from config files and environment
//! variables (see [`searchpath`]).
//!
//! Module names are derived from file content (scanning for `DEFINITIONS`
//! headers), not from filenames. Files are matched by extension using
//! [`source::DEFAULT_EXTENSIONS`] (`.mib`, `.smi`, `.txt`, `.my`, or no
//! extension).
//!
//! # Type Introspection
//!
//! SMI types form parent chains. A MIB might define `HostName` as a
//! refinement of `DisplayString`, which is itself a textual convention
//! over `OCTET STRING`. Each link in the chain can add constraints
//! (size limits, value ranges), a display hint (how to render the value
//! as text), or enumeration labels.
//!
//! A **textual convention** (TC) is the standard way to define reusable
//! types in SMIv2 (RFC 2579). A TC wraps a base type with a name,
//! description, and optional DISPLAY-HINT and constraints. For example,
//! `DisplayString` is a TC over `OCTET STRING (SIZE (0..255))` with
//! display hint `"255a"`. Use [`Type::is_textual_convention`] to check
//! whether a type was defined as a TC.
//!
//! ## Constraints: SIZE vs range
//!
//! Both [`Type::sizes`] and [`Type::ranges`] return `&[Range]`, but
//! they constrain different things:
//!
//! - **SIZE** constrains the length (in octets) of string-like types
//!   (`OCTET STRING`, `Opaque`). Example: `SIZE (0..255)` means
//!   at most 255 bytes.
//! - **Range** constrains the numeric value of integer-like types.
//!   Example: `(1..2147483647)` means the value must be at least 1.
//!
//! The `effective_*` variants walk the parent chain to find inherited
//! constraints.
//!
//! ## Display hints
//!
//! A DISPLAY-HINT string (RFC 2579, Section 3) tells a MIB browser or
//! SNMP tool how to render a raw value as human-readable text. Common
//! examples:
//!
//! - `"255a"` - up to 255 ASCII characters (used by `DisplayString`)
//! - `"1x:"` - hex bytes separated by colons (used by `MacAddress`)
//! - `"2d-1d-1d,1d:1d:1d.1d"` - date-time components (used by
//!   `DateAndTime`)
//!
//! [`Type::effective_display_hint`] and
//! [`Object::effective_display_hint`] return the hint string.
//! [`Object::format_integer`], [`Object::format_octets`], and
//! [`Object::scale_integer`] apply the hint directly to raw values.
//! The [`display_hint`](mib::display_hint) module exposes the same
//! formatting functions for use without an Object handle.
//!
//! ## Direct vs effective accessors
//!
//! Each [`Type`] handle exposes two families of accessors:
//!
//! - **Direct** (`base`, `display_hint`, `enums`, `sizes`, `ranges`) -
//!   return only what this specific type declares. These are empty/None
//!   if the type inherits everything from its parent.
//! - **Effective** (`effective_base`, `effective_display_hint`,
//!   `effective_enums`, `effective_sizes`, `effective_ranges`) - walk up
//!   the parent chain and return the first non-empty value. These give
//!   you the "resolved" answer.
//!
//! **In most cases, use the `effective_*` methods.** They give you the
//! answer you actually want: "what base type does this ultimately
//! represent?", "how should I format this value?", "what are the valid
//! enum labels?". The direct methods are mainly useful when you need to
//! know exactly where in the chain a property was introduced, for
//! instance when building a MIB browser that shows the full type
//! derivation.
//!
//! | Method | Description |
//! |--------|-------------|
//! | [`Type::base`] | Directly assigned base type (may be `Unknown` for derived types) |
//! | [`Type::effective_base`] | Resolved base type - use this one |
//! | [`Type::parent`] | Immediate parent type (if derived) |
//! | [`Type::display_hint`] | This type's own DISPLAY-HINT (often empty) |
//! | [`Type::effective_display_hint`] | First non-empty hint in the chain - use this one |
//! | [`Type::enums`] | This type's own enum values |
//! | [`Type::effective_enums`] | First non-empty enums in the chain - use this one |
//! | [`Type::sizes`] / [`Type::ranges`] | This type's own constraints |
//! | [`Type::effective_sizes`] / [`Type::effective_ranges`] | Inherited constraints - use these |
//! | [`Type::is_textual_convention`] | Whether defined as a TEXTUAL-CONVENTION |
//!
//! Convenience predicates: [`Type::is_counter`], [`Type::is_gauge`],
//! [`Type::is_string`], [`Type::is_enumeration`], [`Type::is_bits`].
//! These all use the effective base type internally.
//!
//! Objects expose the same effective accessors directly (e.g.
//! [`Object::effective_display_hint`], [`Object::effective_enums`]) so
//! you don't need to go through the type handle for common lookups.
//!
//! # Diagnostics and Configuration
//!
//! Real-world MIB files frequently contain errors, vendor-specific
//! extensions, or references to modules you don't have. Rather than
//! failing on the first problem, this library collects diagnostics and
//! continues, producing as much useful output as possible. After
//! loading, check [`Mib::has_errors`] and inspect [`Mib::diagnostics`]
//! for details.
//!
//! There are two independent knobs that control loading behavior.
//! They can seem redundant at first ("don't both of them control how
//! strict loading is?"), but they operate at different levels.
//!
//! Think of it like a Rust analogy: [`ResolverStrictness`] is like
//! controlling how `use` imports are resolved. In Rust, `use foo::Bar`
//! must name the exact path. In MIBs, imports work similarly - a module
//! declares `IMPORTS DisplayString FROM SNMPv2-TC`. But many real-world
//! MIBs get this wrong: they import from the wrong module, forget to
//! import well-known names they assume are built-in, or don't declare
//! imports at all. `ResolverStrictness` controls how hard the resolver
//! tries to recover from these mistakes, from following only explicit
//! import chains (`Strict`) to searching all loaded modules by name
//! (`Permissive`).
//!
//! [`DiagnosticConfig`], on the other hand, is like compiler warnings
//! and `-Werror`. It controls what gets reported across the entire
//! pipeline (lexing, parsing, and resolution), and whether problems
//! cause `load()` to fail. It doesn't change what gets resolved.
//!
//! The key tradeoff with `ResolverStrictness` is **correctness vs
//! completeness**. The more permissive you go, the more things get
//! resolved, but the higher the risk of incorrect results. At
//! `Permissive`, the resolver falls back to searching all loaded
//! modules for a matching symbol name. If multiple modules define a
//! symbol with the same name, you're essentially guessing which one
//! was intended. At `Strict`, the resolver only follows deterministic
//! strategies where the source is unambiguous (explicit imports,
//! import forwarding chains, ASN.1 primitives), so resolved symbols
//! are traceable back to their origin.
//!
//! ## ResolverStrictness - what the resolver attempts
//!
//! [`ResolverStrictness`] controls how aggressively the resolver tries
//! to recover when it can't find a symbol through explicit imports. Set
//! via [`Loader::resolver_strictness`](load::Loader::resolver_strictness).
//!
//! | Level | Behavior | Correctness risk | When to use |
//! |-------|----------|-----------------|-------------|
//! | `Strict` | Minimal fallbacks. Only deterministic strategies that don't guess the source module. | Lowest - if it resolves, the import chain is traceable. | Validating MIBs for correctness, linting, CI checks. |
//! | `Normal` (default) | Constrained fallbacks: searches well-known base modules for unimported types and OID roots, and resolves module name aliases. | Low - fallbacks are limited to safe, unambiguous cases. | General use. Handles sloppy imports that are obviously resolvable. |
//! | `Permissive` | All fallbacks, including searching every loaded module for a symbol by name. | Higher - if two modules define `FooStatus`, the resolver picks one. | Loading badly-written vendor MIBs that you can't fix. |
//!
//! ### Specific behaviors by level
//!
//! **All levels (including Strict):**
//! - Direct import resolution (symbol found in the named source module).
//! - Import forwarding: MIB authors often import a symbol from a
//!   module that uses it, not realizing that module doesn't define
//!   the symbol - it imports it from somewhere else. SMI imports
//!   are not transitive (importing from a module only gives you
//!   what that module defines, not what it imports), but many MIB
//!   authors treat them as if they were, similar to how programmers
//!   sometimes confuse which scope a variable is visible in.
//!
//!   For example, suppose `ACME-TC` defines a textual convention
//!   `AcmeStatus`, and `ACME-MIB` imports and uses it:
//!
//!   ```text
//!   ACME-TC DEFINITIONS ::= BEGIN
//!     AcmeStatus ::= TEXTUAL-CONVENTION ...
//!   END
//!
//!   ACME-MIB DEFINITIONS ::= BEGIN
//!     IMPORTS AcmeStatus FROM ACME-TC;
//!     -- uses AcmeStatus in OBJECT-TYPE definitions
//!   END
//!   ```
//!
//!   A third module might then mistakenly import `AcmeStatus` from
//!   `ACME-MIB` instead of from `ACME-TC`:
//!
//!   ```text
//!   ACME-EXTENSION-MIB DEFINITIONS ::= BEGIN
//!     IMPORTS AcmeStatus FROM ACME-MIB;  -- wrong: ACME-MIB doesn't define it
//!   END
//!   ```
//!
//!   The resolver handles this by checking `ACME-MIB`'s own IMPORTS,
//!   finding that it declares `AcmeStatus FROM ACME-TC`, and
//!   following that chain. This is deterministic - the intermediate
//!   module explicitly names its source - so it is enabled at all
//!   strictness levels.
//! - Partial import resolution: when a source module has some but not
//!   all of the requested symbols, the ones that exist are resolved
//!   individually and the rest are reported as unresolved.
//! - ASN.1 primitive type fallback: `INTEGER`, `OCTET STRING`,
//!   `OBJECT IDENTIFIER`, and `BITS` always resolve from SNMPv2-SMI
//!   even without an explicit import.
//! - Well-known OID roots: `iso`, `ccitt`, and `joint-iso-ccitt`
//!   always resolve to their fixed arc values.
//!
//! **Normal and Permissive (constrained fallbacks):**
//! - Module name aliases: maps alternate module names to their
//!   canonical form (e.g. `SNMPv2-SMI-v1` to `SNMPv2-SMI`,
//!   `RFC-1213` to `RFC1213-MIB`). These aliases exist because
//!   modules have been renamed over time as RFCs were revised,
//!   and some vendors use non-standard names in their IMPORTS.
//! - Unimported well-known symbol fallback: names like `enterprises`,
//!   `Counter64`, and `DisplayString` feel like built-in language
//!   keywords, but they're actually defined in specific base modules
//!   (`SNMPv2-SMI`, `SNMPv2-TC`, etc.) and formally need to be
//!   imported. Many MIB authors skip the import, treating these names
//!   as globally available:
//!
//!   ```text
//!   ACME-MIB DEFINITIONS ::= BEGIN
//!     IMPORTS
//!       MODULE-IDENTITY, OBJECT-TYPE
//!         FROM SNMPv2-SMI;
//!     -- no import for enterprises or Counter64
//!
//!     acmeMib MODULE-IDENTITY ... ::= { enterprises 12345 }
//!
//!     acmeCounter OBJECT-TYPE
//!       SYNTAX Counter64   -- not imported
//!       ...
//!   END
//!   ```
//!
//!   When a type or OID parent is not found via imports, the resolver
//!   searches the well-known base modules (SNMPv2-SMI, RFC1155-SMI,
//!   SNMPv2-TC). This is limited to those specific modules, so there
//!   is no ambiguity about which definition is meant.
//! - TRAP-TYPE enterprise global lookup: the ENTERPRISE reference in
//!   TRAP-TYPE definitions is searched across all modules when not
//!   found via imports.
//!
//! **Permissive only (global fallbacks):**
//! - Global object lookup: INDEX objects, AUGMENTS targets,
//!   NOTIFICATION-TYPE OBJECTS members, and DEFVAL object references
//!   are searched across all loaded modules when not found via imports.
//! - Global group/compliance member lookup: OBJECT-GROUP members,
//!   MODULE-COMPLIANCE mandatory groups, and AGENT-CAPABILITIES
//!   variation targets are searched globally.
//!
//! **Which should I use?** Start with `Normal` (the default). If you
//! get unresolved-reference diagnostics, it's usually better to fix
//! the MIB file directly (correcting the `IMPORTS` statement to name
//! the right source module) rather than reaching for `Permissive`.
//! MIB files are plain text, and import fixes are usually obvious from
//! the diagnostic message. Reserve `Permissive` for cases where you
//! can't modify the MIB files, such as vendor-supplied MIBs loaded
//! from a read-only path. `Strict` is useful for validation tooling
//! or CI, where you want broken imports to surface as unresolved
//! references rather than being silently fixed up.
//!
//! ## DiagnosticConfig - what gets reported
//!
//! [`DiagnosticConfig`] controls which diagnostics are collected and
//! which severity level causes `load()` to fail. This is purely about
//! reporting - it does not change what the resolver does. Set via
//! [`Loader::diagnostic_config`](load::Loader::diagnostic_config).
//!
//! It has four preset constructors:
//!
//! | Preset | What's reported | `load()` fails at | When to use |
//! |--------|-----------------|-------------------|-------------|
//! | [`DiagnosticConfig::verbose()`] | Everything (style, info, warnings) | Severe | Debugging MIB issues, understanding what the resolver did. |
//! | [`DiagnosticConfig::default()`] | Minor and above | Severe | General use. |
//! | [`DiagnosticConfig::quiet()`] | Errors and above only | Severe | Production code that just wants to know about real problems. |
//! | [`DiagnosticConfig::silent()`] | Nothing | Fatal only | When you don't care about diagnostics at all and want `load()` to succeed unless something is truly broken. |
//!
//! **Which should I use?** The default is fine for most cases. Use
//! `quiet()` in production if you don't want to surface minor issues
//! to users. Use `silent()` when loading untrusted or messy vendor
//! MIBs where you just want whatever data you can get. Use `verbose()`
//! when diagnosing why something isn't resolving correctly.
//!
//! ## Combining the two
//!
//! Since strictness controls resolution behavior and diagnostics
//! controls reporting across the whole pipeline, they can be mixed
//! freely:
//!
//! - `Normal` + `default()` - good general-purpose defaults.
//! - `Permissive` + `silent()` - maximum tolerance. Tries every
//!   fallback, suppresses all diagnostics, only fails on fatal errors.
//!   Good for loading a pile of vendor MIBs where you want whatever
//!   data you can get. Be aware that some resolved symbols may be
//!   incorrect due to ambiguous fallback matches.
//! - `Strict` + `verbose()` - maximum strictness. Minimal fallbacks,
//!   all diagnostics reported (including parse warnings and style
//!   issues). Good for validating MIBs you author.
//! - `Normal` + `quiet()` - reasonable for a production SNMP tool that
//!   loads user-provided MIBs. Safe fallbacks, but only real errors
//!   are surfaced.
//!
//! ## Fine-tuning
//!
//! For more control, [`DiagnosticConfig`] also supports:
//!
//! - `fail_at` - change which severity causes `load()` to return an
//!   error. For example, set to [`Severity::Minor`] to fail on any
//!   minor issue.
//! - `overrides` - promote or demote specific diagnostic codes (e.g.
//!   turn a warning into an error).
//! - `ignore` - glob patterns to suppress specific diagnostic codes
//!   entirely (e.g. `"import-*"` to ignore all import-related
//!   diagnostics).
//!
//! # Feature Flags
//!
//! | Feature | Default | Description |
//! |---------|---------|-------------|
//! | `serde` | yes | Serde support and JSON export via [`export`] |
//! | `cli` | yes | CLI binary (`mib-rs`) |
//!
//! # Examples
//!
//! Runnable examples live in the `examples/` directory. Each one can be run
//! with `cargo run --example <name>`.
//!
//! ## Basic usage
//!
//! Load a MIB from memory, query objects, and display module metadata.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/basic.rs")]
//! ```
//!
//! ## OID tree walking
//!
//! Root traversal, subtree iteration, depth-first walk, and node navigation.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/walk.rs")]
//! ```
//!
//! ## Type introspection
//!
//! Type chains, effective values, constraints, enums, display hints,
//! and classification predicates.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/types.rs")]
//! ```
//!
//! ## Table navigation
//!
//! Tables, rows, columns, indexes, and object kind predicates.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/tables.rs")]
//! ```
//!
//! ## Module metadata
//!
//! Module metadata, imports, revisions, base modules, and module-scoped
//! iteration.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/modules.rs")]
//! ```
//!
//! ## JSON export
//!
//! JSON export of a resolved MIB using the serde-based export API.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/export.rs")]
//! ```
//!
//! ## Notifications, groups, and compliance
//!
//! Notifications, object groups, notification groups, and compliance statements.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/notifications.rs")]
//! ```
//!
//! ## Query formats
//!
//! Plain names, qualified names, numeric OIDs, instance OIDs, and OID formatting.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/query.rs")]
//! ```
//!
//! ## Diagnostics
//!
//! Diagnostic collection, strictness levels, reporting configuration,
//! filtering, and severity overrides.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/diagnostics.rs")]
//! ```
//!
//! ## Raw data access
//!
//! Low-level raw data access: sub-clause spans, import metadata,
//! OID references, symbol tables, and bulk arena access.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/raw.rs")]
//! ```
//!
//! ## Tokenization
//!
//! Lexical tokenization of MIB source text for syntax highlighting,
//! linting, or custom tooling.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/tokens.rs")]
//! ```
//!
//! ## Sources
//!
//! Source types: in-memory modules, directory sources, chaining,
//! and module listing.
//!
//! ```rust,no_run
#![doc = include_str!("../examples/sources.rs")]
//! ```
pub mod ast;
pub mod error;
#[cfg(feature = "serde")]
pub mod export;
pub(crate) mod graph;
pub mod ir;
pub(crate) mod lexer;
pub mod load;
pub mod lower;
pub mod mib;
pub mod parser;
pub(crate) mod scan;
pub mod searchpath;
pub mod source;
pub mod token;
pub mod types;

// Re-exports for convenience
pub use error::LoadError;
pub use load::{Loader, load};
pub use mib::{
    Capability, Compliance, Group, Index, Mib, Module, Node, Notification, Object, Oid, OidLookup,
    ParseOidError, ResolveOidError, Type,
    index::{DecodedIndex, IndexValue},
};
pub use source::{FindResult, Source};
pub use token::{Token, TokenKind};
pub use types::{
    Access, AccessKeyword, BaseType, DiagCode, Diagnostic, DiagnosticConfig, IndexEncoding, Kind,
    Language, ReportingLevel, ResolverStrictness, Severity, Status,
};

/// Low-level resolved data access.
///
/// This module exposes arena ids, backing records, and the explicit
/// [`RawMib`](raw::RawMib) view returned by [`Mib::raw()`].
pub mod raw {
    pub use crate::mib::{
        CapabilityData, CapabilityId, ComplianceData, ComplianceId, GroupData, GroupId, ModuleData,
        ModuleId, NodeData, NodeId, NotificationData, NotificationId, ObjectData, ObjectId,
        OidTree, RawMib, Symbol, TypeData, TypeId,
    };
}

/// Compiler pipeline APIs exposed before final resolution.
///
/// These modules are useful when building syntax-aware tooling or diagnostics
/// that need direct access to tokens, parsed AST, lowered IR, or the parser
/// entry points themselves.
pub mod compile {
    pub use crate::{ast, ir, lower, parser, token};
}