cadmpeg-codec-step 0.5.5

// SPDX-License-Identifier: Apache-2.0
//! Schema identifier grammar table.
//!
//! The classifier is a total pure function over the identifier text. This table
//! states one form for each grammar rule. The `FILE_SCHEMA` diagnostic surface
//! that reads the recoverable form is pinned in `parse/tests/envelope.rs`.

use super::{split_schema_identifier, AdmittedSchemaIdentifier};

/// The admitted form of one identifier, as one comparable value.
#[derive(Debug, PartialEq, Eq)]
enum Admitted {
    /// The classifier rejects the identifier.
    Rejected,
    /// The schema name and the optional object identifier are both valid.
    Valid,
    /// The object identifier holds an out-of-range component. The fields are
    /// the schema name and the first out-of-range component in source order.
    OutOfRange(String, String),
}

fn admitted(identifier: &str) -> Admitted {
    let Some(admitted) = AdmittedSchemaIdentifier::admit(identifier.to_owned()) else {
        return Admitted::Rejected;
    };
    assert_eq!(
        admitted.text(),
        identifier,
        "the admission keeps the source text"
    );
    match admitted {
        AdmittedSchemaIdentifier::Valid { .. } => Admitted::Valid,
        AdmittedSchemaIdentifier::ObjectIdentifierOutOfRange {
            name, component, ..
        } => Admitted::OutOfRange(name, component),
    }
}

#[test]
fn classifier_reports_the_first_out_of_range_component_in_source_order() {
    // The root component states a root number as `0`, `1`, or `2`, or as a
    // registered ASN.1 identifier for one of those numbers. Every other root
    // component is out of range. Under root `0` or `1` the second number is in
    // `0..=39`. Every other position permits every non-negative number, and a
    // negative number is out of range in every position. The report names the
    // whole component text, not the number inside it.
    let cases = [
        // A minus sign is out of range in the root position.
        ("AP242 { -1 40 }", "-1"),
        // Root `1` bounds the second number, so `40` is out of range and the
        // root itself is in range.
        ("AP242 { 1 40 }", "40"),
        // Root `0` bounds the second number the same way.
        ("AP242 { 0 40 }", "40"),
        // `iso` is the registered identifier for root `1`, so it bounds the
        // second number as the number `1` does.
        ("AP242 { iso 40 }", "40"),
        // A component with an identifier and parentheses states the number
        // inside the parentheses, and the second-component bound applies to it.
        ("AP242 { 1 part(214) }", "part(214)"),
        // `3` is not a root number, so the root is out of range and no bound
        // reaches the second component.
        ("AP242 { 3 1 }", "3"),
        // An unregistered identifier states no root number, so the root is out
        // of range for the same reason that `3` is.
        ("AP242 { foo 40 }", "foo"),
        // The registered identifiers compare exactly, so a case difference
        // gives another identifier, which states no root number.
        ("AP242 { iSo 40 }", "iSo"),
        // An identifier and parentheses that hold an identifier state no
        // number, so this root is out of range even though it opens with a
        // registered identifier.
        ("AP242 { iso(standard) 40 }", "iso(standard)"),
    ];
    for (identifier, component) in cases {
        assert_eq!(
            admitted(identifier),
            Admitted::OutOfRange("AP242".to_owned(), component.to_owned()),
            "{identifier}"
        );
    }
}

#[test]
fn classifier_admits_in_range_object_identifiers() {
    let cases = [
        // Root `1` permits a second number in `0..=39`.
        "AP242 { 1 39 }",
        // Root `2` bounds no second number.
        "AP242 { 2 40 }",
        // Each registered identifier states its root number.
        "AP242 { iso 39 }",
        "AP242 { itu-t 39 }",
        "AP242 { ccitt 39 }",
        "AP242 { joint-iso-itu-t 40 }",
        "AP242 { joint-iso-ccitt 40 }",
        // A registered root bounds a numeric second component only, so an
        // unnumbered second component states no number to place out of range.
        "AP242 { iso standard 10303 part(214) version(1) }",
        // An identifier and parentheses that hold a number state that number,
        // so this root is root `1`.
        "AP242 { iso(1) 39 }",
        // A schema name alone is a complete identifier.
        "AP242",
    ];
    for identifier in cases {
        assert_eq!(admitted(identifier), Admitted::Valid, "{identifier}");
    }
}

#[test]
fn classifier_rejects_identifiers_that_do_not_parse() {
    let cases = [
        // A negative number has no leading zero either, so `-01` is not a
        // component number and the component has no form.
        "AP242 { 1 0 10303 442 -01 1 4 }",
        // An ASN.1 identifier uses letters, digits, and hyphens only.
        "AP242 { 1 0 10303 442 invalid_ 1 4 }",
        // An ASN.1 identifier starts with a lowercase letter, so an uppercase
        // spelling of a registered identifier has no component form. The
        // identifier does not parse, so the range rule does not apply to it.
        "AP242 { ISO 40 }",
        "AP242 { Iso 40 }",
        // A numeric component has no leading zero.
        "AP242 { 01 0 }",
        // An object identifier has at least two components.
        "AP242 { 1 }",
        // An identifier that opens a brace closes it at the end.
        "AP242 { 1 40",
        // A schema name is an EXPRESS simple_id: a letter, then letters,
        // digits, or underscores.
        "9AP242",
        "_AP242",
        "",
    ];
    for identifier in cases {
        assert_eq!(admitted(identifier), Admitted::Rejected, "{identifier}");
    }
}

#[test]
fn split_separates_the_schema_name_from_the_object_identifier() {
    // The split ignores whitespace around the identifier and around the schema
    // name. An identifier with no brace is a schema name alone. An identifier
    // that opens an object identifier and does not close it at the end of the
    // identifier has no schema name and no object identifier. The split reads
    // the first brace, so a brace inside the object identifier text stays in
    // that text, where no component form accepts it.
    let cases = [
        ("AP242", Some(("AP242", None))),
        ("  AP242  ", Some(("AP242", None))),
        ("AP242 { 1 40 }", Some(("AP242", Some(" 1 40 ")))),
        ("  AP242{1 40}  ", Some(("AP242", Some("1 40")))),
        ("AP242 {}", Some(("AP242", Some("")))),
        ("{ 1 40 }", Some(("", Some(" 1 40 ")))),
        ("A{B{1 2}", Some(("A", Some("B{1 2")))),
        ("AP242 { 1 40", None),
        ("A{1 2", None),
        ("AP242 { 1 40 } x", None),
    ];
    for (identifier, expected) in cases {
        assert_eq!(
            split_schema_identifier(identifier),
            expected,
            "{identifier}"
        );
    }
}