automapper_validation/expr/

parser.rs

//! Recursive descent parser for condition expressions.
//!
//! Grammar (from lowest to highest precedence):
//!
//! ```text
//! expression  = xor_expr
//! xor_expr    = or_expr (XOR or_expr)*
//! or_expr     = and_expr (OR and_expr)*
//! and_expr    = not_expr ((AND | implicit) not_expr)*
//! not_expr    = NOT not_expr | primary
//! primary     = CONDITION_ID | '(' expression ')'
//! ```
//!
//! Implicit AND: two adjacent condition references or a condition followed by
//! `(` without an intervening operator are treated as AND.

use std::collections::BTreeMap;

use super::ast::ConditionExpr;
use super::status::{status_lines, StatusBranch, StatusKind};
use super::token::{strip_status_prefix, tokenize, SpannedToken, Token};
use crate::error::ParseError;

/// Parser for AHB condition expressions.
pub struct ConditionParser;

impl ConditionParser {
    /// Parse an AHB status string into a condition expression.
    ///
    /// Returns `Ok(None)` if the input contains no condition references
    /// (e.g., bare `"Muss"` or empty string).
    ///
    /// # Examples
    ///
    /// ```
    /// use automapper_validation::expr::ConditionParser;
    /// use automapper_validation::expr::ConditionExpr;
    ///
    /// let expr = ConditionParser::parse("Muss [494]").unwrap().unwrap();
    /// assert_eq!(expr, ConditionExpr::Ref(494));
    /// ```
    pub fn parse(input: &str) -> Result<Option<ConditionExpr>, ParseError> {
        Self::parse_with_ub(input, &BTreeMap::new())
    }

    /// Parse an AHB status string, expanding UB condition references inline.
    ///
    /// When `[UB1]` is encountered and `ub_definitions` contains "UB1", the
    /// corresponding pre-parsed expression is substituted. Unknown UB references
    /// fall back to the normal `parse_condition_id` behavior.
    pub fn parse_with_ub(
        input: &str,
        ub_definitions: &BTreeMap<String, ConditionExpr>,
    ) -> Result<Option<ConditionExpr>, ParseError> {
        let input = input.trim();
        if input.is_empty() {
            return Ok(None);
        }

        // AHB_Status can carry alternative branches on separate lines, each with its
        // own status prefix, e.g.
        //     Muss [315] ∧ [707]
        //     Soll [8] ∧ [301] ∧ [707]
        // As one expression the branches form a disjunction: the part applies if
        // any branch's condition holds. Which status then applies is
        // `ConditionExprEvaluator::resolve_status`'s business.
        let alternatives: Vec<ConditionExpr> = Self::parse_status_branches(input, ub_definitions)?
            .into_iter()
            .filter_map(|b| b.condition)
            .collect();

        match alternatives.len() {
            0 => Ok(None),
            1 => Ok(alternatives.into_iter().next()),
            _ => Ok(Some(ConditionExpr::Or(alternatives))),
        }
    }

    /// Parse an AHB status into one [`StatusBranch`] per status line.
    ///
    /// Validation asks for the same few hundred statuses of a format version
    /// for every field of every group instance (an MSCONS Lastgang: thousands
    /// of times each), so each thread keeps what it parsed. A status naming a
    /// UB condition depends on `ub_definitions` as well and is not kept.
    pub fn parse_status_branches(
        input: &str,
        ub_definitions: &BTreeMap<String, ConditionExpr>,
    ) -> Result<Vec<StatusBranch>, ParseError> {
        thread_local! {
            static PARSED: std::cell::RefCell<
                std::collections::HashMap<String, Result<Vec<StatusBranch>, ParseError>>,
            > = Default::default();
        }
        // More distinct statuses than the format versions hold together
        // means the input is not an AHB's; stop keeping them.
        const KEPT: usize = 16_384;
        if input.contains("UB") {
            return Self::parse_status_branches_uncached(input, ub_definitions);
        }
        if let Some(hit) = PARSED.with(|p| p.borrow().get(input).cloned()) {
            return hit;
        }
        let parsed = Self::parse_status_branches_uncached(input, ub_definitions);
        PARSED.with(|p| {
            let mut p = p.borrow_mut();
            if p.len() < KEPT {
                p.insert(input.to_string(), parsed.clone());
            }
        });
        parsed
    }

    fn parse_status_branches_uncached(
        input: &str,
        ub_definitions: &BTreeMap<String, ConditionExpr>,
    ) -> Result<Vec<StatusBranch>, ParseError> {
        let mut branches = Vec::new();
        for line in status_lines(input) {
            let kind = StatusKind::of(line);
            let stripped = strip_status_prefix(line);
            // A bare status word ("Kann") is left as it is: no condition.
            let condition = if kind.is_some() && stripped == line {
                None
            } else {
                let tokens = tokenize(stripped)?;
                let mut pos = 0;
                parse_expression(&tokens, &mut pos, ub_definitions)?
            };
            branches.push(StatusBranch { kind, condition });
        }
        Ok(branches)
    }

    /// Parse an expression that is known to contain conditions (no prefix stripping).
    ///
    /// Returns `Err` if the input cannot be parsed. Returns `Ok(None)` if empty.
    pub fn parse_raw(input: &str) -> Result<Option<ConditionExpr>, ParseError> {
        let input = input.trim();
        if input.is_empty() {
            return Ok(None);
        }

        let tokens = tokenize(input)?;
        if tokens.is_empty() {
            return Ok(None);
        }

        let mut pos = 0;
        let expr = parse_expression(&tokens, &mut pos, &BTreeMap::new())?;

        Ok(expr)
    }
}

/// Parse a full expression (entry point for precedence climbing).
fn parse_expression(
    tokens: &[SpannedToken],
    pos: &mut usize,
    ub_definitions: &BTreeMap<String, ConditionExpr>,
) -> Result<Option<ConditionExpr>, ParseError> {
    parse_xor(tokens, pos, ub_definitions)
}

/// XOR has the lowest precedence.
fn parse_xor(
    tokens: &[SpannedToken],
    pos: &mut usize,
    ub_definitions: &BTreeMap<String, ConditionExpr>,
) -> Result<Option<ConditionExpr>, ParseError> {
    let mut left = match parse_or(tokens, pos, ub_definitions)? {
        Some(expr) => expr,
        None => return Ok(None),
    };

    while *pos < tokens.len() && tokens[*pos].token == Token::Xor {
        *pos += 1; // consume XOR
        let right = match parse_or(tokens, pos, ub_definitions)? {
            Some(expr) => expr,
            None => return Ok(Some(left)),
        };
        left = ConditionExpr::Xor(Box::new(left), Box::new(right));
    }

    Ok(Some(left))
}

/// OR has middle-low precedence.
fn parse_or(
    tokens: &[SpannedToken],
    pos: &mut usize,
    ub_definitions: &BTreeMap<String, ConditionExpr>,
) -> Result<Option<ConditionExpr>, ParseError> {
    let mut left = match parse_and(tokens, pos, ub_definitions)? {
        Some(expr) => expr,
        None => return Ok(None),
    };

    while *pos < tokens.len() && tokens[*pos].token == Token::Or {
        *pos += 1; // consume OR
        let right = match parse_and(tokens, pos, ub_definitions)? {
            Some(expr) => expr,
            None => return Ok(Some(left)),
        };
        // Flatten nested ORs into a single Or(vec![...])
        left = match left {
            ConditionExpr::Or(mut exprs) => {
                exprs.push(right);
                ConditionExpr::Or(exprs)
            }
            _ => ConditionExpr::Or(vec![left, right]),
        };
    }

    Ok(Some(left))
}

/// AND has middle-high precedence. Also handles implicit AND between adjacent
/// conditions or parenthesized groups.
fn parse_and(
    tokens: &[SpannedToken],
    pos: &mut usize,
    ub_definitions: &BTreeMap<String, ConditionExpr>,
) -> Result<Option<ConditionExpr>, ParseError> {
    let mut left = match parse_not(tokens, pos, ub_definitions)? {
        Some(expr) => expr,
        None => return Ok(None),
    };

    while *pos < tokens.len() {
        if tokens[*pos].token == Token::And {
            *pos += 1; // consume explicit AND
            let right = match parse_not(tokens, pos, ub_definitions)? {
                Some(expr) => expr,
                None => return Ok(Some(left)),
            };
            left = flatten_and(left, right);
        } else if matches!(
            tokens[*pos].token,
            Token::ConditionId(_) | Token::LeftParen | Token::Not
        ) {
            // Implicit AND: adjacent condition, paren, or NOT without operator
            let right = match parse_not(tokens, pos, ub_definitions)? {
                Some(expr) => expr,
                None => return Ok(Some(left)),
            };
            left = flatten_and(left, right);
        } else {
            break;
        }
    }

    Ok(Some(left))
}

/// Flatten nested ANDs into a single And(vec![...]).
fn flatten_and(left: ConditionExpr, right: ConditionExpr) -> ConditionExpr {
    match left {
        ConditionExpr::And(mut exprs) => {
            exprs.push(right);
            ConditionExpr::And(exprs)
        }
        _ => ConditionExpr::And(vec![left, right]),
    }
}

/// NOT has the highest precedence (unary prefix).
fn parse_not(
    tokens: &[SpannedToken],
    pos: &mut usize,
    ub_definitions: &BTreeMap<String, ConditionExpr>,
) -> Result<Option<ConditionExpr>, ParseError> {
    if *pos < tokens.len() && tokens[*pos].token == Token::Not {
        *pos += 1; // consume NOT
        let inner = match parse_not(tokens, pos, ub_definitions)? {
            Some(expr) => expr,
            None => {
                return Err(ParseError::UnexpectedToken {
                    position: if *pos < tokens.len() {
                        tokens[*pos].position
                    } else {
                        0
                    },
                    expected: "expression after NOT".to_string(),
                    found: "end of input".to_string(),
                });
            }
        };
        return Ok(Some(ConditionExpr::Not(Box::new(inner))));
    }
    parse_primary(tokens, pos, ub_definitions)
}

/// Primary: a condition reference or a parenthesized expression.
///
/// When a `ConditionId` matches a key in `ub_definitions`, the pre-parsed
/// UB expression is substituted inline instead of calling `parse_condition_id`.
fn parse_primary(
    tokens: &[SpannedToken],
    pos: &mut usize,
    ub_definitions: &BTreeMap<String, ConditionExpr>,
) -> Result<Option<ConditionExpr>, ParseError> {
    if *pos >= tokens.len() {
        return Ok(None);
    }

    match &tokens[*pos].token {
        Token::ConditionId(id) => {
            *pos += 1;
            // Check for UB expansion before falling back to parse_condition_id
            if let Some(ub_expr) = ub_definitions.get(id.as_str()) {
                return Ok(Some(ub_expr.clone()));
            }
            Ok(Some(parse_condition_id(id)))
        }
        Token::LeftParen => {
            *pos += 1; // consume (
            let expr = parse_expression(tokens, pos, ub_definitions)?;
            // Consume closing paren if present (graceful handling of missing)
            if *pos < tokens.len() && tokens[*pos].token == Token::RightParen {
                *pos += 1;
            }
            Ok(expr)
        }
        _ => Ok(None),
    }
}

/// Parse a condition ID string into a ConditionExpr.
///
/// Numeric IDs become `Ref(n)`. Non-numeric IDs (like `UB1`, `10P1..5`)
/// are kept as-is by extracting numeric portions. For the Rust port,
/// pure numeric IDs use `Ref(u32)`. Non-numeric IDs extract leading
/// digits if present, otherwise use 0 as a sentinel.
fn parse_condition_id(id: &str) -> ConditionExpr {
    // Try pure numeric ID first
    if let Ok(num) = id.parse::<u32>() {
        return ConditionExpr::Ref(num);
    }

    // Check for package condition: NP or NPmin..max
    if let Some(p_pos) = id.find('P') {
        let num_part = &id[..p_pos];
        let range_part = &id[p_pos + 1..];
        if let Ok(pkg_id) = num_part.parse::<u32>() {
            let (min, max) = parse_package_range(range_part);
            return ConditionExpr::Package {
                id: pkg_id,
                min,
                max,
            };
        }
    }

    // For non-numeric, non-package IDs (e.g., "UB1"), extract leading digits
    let numeric_part: String = id.chars().take_while(|c| c.is_ascii_digit()).collect();
    if let Ok(num) = numeric_part.parse::<u32>() {
        ConditionExpr::Ref(num)
    } else {
        ConditionExpr::Ref(0)
    }
}

/// Parse the range part of a package condition: `"0..1"` → `(0, 1)`, `""` → `(0, MAX)`.
fn parse_package_range(range: &str) -> (u32, u32) {
    if range.is_empty() {
        return (0, u32::MAX);
    }
    if let Some((min_str, max_str)) = range.split_once("..") {
        let min = min_str.parse::<u32>().unwrap_or(0);
        let max = max_str.parse::<u32>().unwrap_or(u32::MAX);
        (min, max)
    } else {
        let n = range.parse::<u32>().unwrap_or(0);
        (n, n)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use pretty_assertions::assert_eq;

    // === Basic parsing ===

    #[test]
    fn test_parse_single_condition() {
        let result = ConditionParser::parse("[931]").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Ref(931));
    }

    #[test]
    fn test_parse_with_muss_prefix() {
        let result = ConditionParser::parse("Muss [494]").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Ref(494));
    }

    #[test]
    fn test_parse_with_soll_prefix() {
        let result = ConditionParser::parse("Soll [494]").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Ref(494));
    }

    #[test]
    fn test_parse_with_kann_prefix() {
        let result = ConditionParser::parse("Kann [182]").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Ref(182));
    }

    #[test]
    fn test_parse_with_x_prefix() {
        let result = ConditionParser::parse("X [567]").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Ref(567));
    }

    // === Binary operators ===

    #[test]
    fn test_parse_simple_and() {
        let result = ConditionParser::parse("[182] ∧ [152]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(182), ConditionExpr::Ref(152)])
        );
    }

    #[test]
    fn test_parse_simple_or() {
        let result = ConditionParser::parse("[1] ∨ [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)])
        );
    }

    #[test]
    fn test_parse_simple_xor() {
        let result = ConditionParser::parse("[1] ⊻ [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Xor(
                Box::new(ConditionExpr::Ref(1)),
                Box::new(ConditionExpr::Ref(2)),
            )
        );
    }

    // === Chained operators ===

    #[test]
    fn test_parse_three_way_and() {
        let result = ConditionParser::parse("[1] ∧ [2] ∧ [3]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![
                ConditionExpr::Ref(1),
                ConditionExpr::Ref(2),
                ConditionExpr::Ref(3),
            ])
        );
    }

    #[test]
    fn test_parse_three_way_and_with_prefix() {
        let result = ConditionParser::parse("Kann [182] ∧ [6] ∧ [570]")
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![
                ConditionExpr::Ref(182),
                ConditionExpr::Ref(6),
                ConditionExpr::Ref(570),
            ])
        );
        assert_eq!(result.condition_ids(), [6, 182, 570].into());
    }

    #[test]
    fn test_parse_multiple_xor() {
        let result = ConditionParser::parse("[1] ⊻ [2] ⊻ [3] ⊻ [4]")
            .unwrap()
            .unwrap();
        assert_eq!(result.condition_ids(), [1, 2, 3, 4].into());
    }

    // === Parentheses ===

    #[test]
    fn test_parse_parenthesized_expression() {
        let result = ConditionParser::parse("([1] ∨ [2]) ∧ [3]")
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![
                ConditionExpr::Or(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)]),
                ConditionExpr::Ref(3),
            ])
        );
    }

    #[test]
    fn test_parse_nested_parentheses() {
        // (([1] ∧ [2]) ∨ ([3] ∧ [4])) ∧ [5]
        let result = ConditionParser::parse("(([1] ∧ [2]) ∨ ([3] ∧ [4])) ∧ [5]")
            .unwrap()
            .unwrap();
        assert_eq!(result.condition_ids(), [1, 2, 3, 4, 5].into());
        // Outer is AND
        match &result {
            ConditionExpr::And(exprs) => {
                assert_eq!(exprs.len(), 2);
                assert!(matches!(&exprs[0], ConditionExpr::Or(_)));
                assert_eq!(exprs[1], ConditionExpr::Ref(5));
            }
            other => panic!("Expected And, got {other:?}"),
        }
    }

    // === Operator precedence ===

    #[test]
    fn test_and_has_higher_precedence_than_or() {
        // [1] ∨ [2] ∧ [3] should parse as [1] ∨ ([2] ∧ [3])
        let result = ConditionParser::parse("[1] ∨ [2] ∧ [3]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![
                ConditionExpr::Ref(1),
                ConditionExpr::And(vec![ConditionExpr::Ref(2), ConditionExpr::Ref(3)]),
            ])
        );
    }

    #[test]
    fn test_or_has_higher_precedence_than_xor() {
        // [1] ⊻ [2] ∨ [3] should parse as [1] ⊻ ([2] ∨ [3])
        let result = ConditionParser::parse("[1] ⊻ [2] ∨ [3]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Xor(
                Box::new(ConditionExpr::Ref(1)),
                Box::new(ConditionExpr::Or(vec![
                    ConditionExpr::Ref(2),
                    ConditionExpr::Ref(3),
                ])),
            )
        );
    }

    // === Implicit AND ===

    #[test]
    fn test_adjacent_conditions_implicit_and() {
        // "[1] [2]" is equivalent to "[1] ∧ [2]"
        let result = ConditionParser::parse("[1] [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)])
        );
    }

    #[test]
    fn test_adjacent_conditions_no_space_implicit_and() {
        // "[939][14]" from real AHB XML
        let result = ConditionParser::parse("[939][14]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(939), ConditionExpr::Ref(14)])
        );
    }

    // === NOT operator ===

    #[test]
    fn test_parse_not() {
        let result = ConditionParser::parse("NOT [1]").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Not(Box::new(ConditionExpr::Ref(1))));
    }

    #[test]
    fn test_parse_not_with_and() {
        // NOT [1] ∧ [2] should parse as (NOT [1]) ∧ [2] because NOT has highest precedence
        let result = ConditionParser::parse("NOT [1] ∧ [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![
                ConditionExpr::Not(Box::new(ConditionExpr::Ref(1))),
                ConditionExpr::Ref(2),
            ])
        );
    }

    // === Real-world AHB expressions ===

    #[test]
    fn test_real_world_orders_expression() {
        // From ORDERS AHB: "X (([939] [147]) ∨ ([940] [148])) ∧ [567]"
        let result = ConditionParser::parse("X (([939] [147]) ∨ ([940] [148])) ∧ [567]")
            .unwrap()
            .unwrap();
        assert_eq!(result.condition_ids(), [147, 148, 567, 939, 940].into());
    }

    #[test]
    fn test_real_world_xor_expression() {
        // "Muss ([102] ∧ [2006]) ⊻ ([103] ∧ [2005])"
        let result = ConditionParser::parse("Muss ([102] ∧ [2006]) ⊻ ([103] ∧ [2005])")
            .unwrap()
            .unwrap();
        assert!(matches!(result, ConditionExpr::Xor(_, _)));
        assert_eq!(result.condition_ids(), [102, 103, 2005, 2006].into());
    }

    #[test]
    fn test_real_world_complex_nested_with_implicit_and() {
        // "([939][14]) ∨ ([940][15])"
        let result = ConditionParser::parse("([939][14]) ∨ ([940][15])")
            .unwrap()
            .unwrap();
        assert!(matches!(result, ConditionExpr::Or(_)));
        assert_eq!(result.condition_ids(), [14, 15, 939, 940].into());
    }

    // === Edge cases ===

    #[test]
    fn test_parse_empty_string() {
        assert!(ConditionParser::parse("").unwrap().is_none());
    }

    #[test]
    fn test_parse_whitespace_only() {
        assert!(ConditionParser::parse("   \t  ").unwrap().is_none());
    }

    #[test]
    fn test_parse_bare_muss() {
        assert!(ConditionParser::parse("Muss").unwrap().is_none());
    }

    #[test]
    fn test_parse_bare_x() {
        // "X" alone has no conditions after it
        assert!(ConditionParser::parse("X").unwrap().is_none());
    }

    #[test]
    fn test_parse_unmatched_open_paren_graceful() {
        // ([1] ∧ [2] — missing closing paren
        let result = ConditionParser::parse("([1] ∧ [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)])
        );
    }

    #[test]
    fn test_parse_text_and_operator() {
        let result = ConditionParser::parse("[1] AND [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)])
        );
    }

    #[test]
    fn test_parse_text_or_operator() {
        let result = ConditionParser::parse("[1] OR [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)])
        );
    }

    #[test]
    fn test_parse_text_xor_operator() {
        let result = ConditionParser::parse("[1] XOR [2]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Xor(
                Box::new(ConditionExpr::Ref(1)),
                Box::new(ConditionExpr::Ref(2)),
            )
        );
    }

    #[test]
    fn test_parse_mixed_unicode_and_text_operators() {
        let result = ConditionParser::parse("[1] ∧ [2] OR [3]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![
                ConditionExpr::And(vec![ConditionExpr::Ref(1), ConditionExpr::Ref(2)]),
                ConditionExpr::Ref(3),
            ])
        );
    }

    #[test]
    fn test_parse_deeply_nested() {
        // ((([1])))
        let result = ConditionParser::parse("((([1])))").unwrap().unwrap();
        assert_eq!(result, ConditionExpr::Ref(1));
    }

    // === Package conditions ===

    #[test]
    fn test_parse_package_condition_0_1() {
        let result = ConditionParser::parse("[4P0..1]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Package {
                id: 4,
                min: 0,
                max: 1
            }
        );
    }

    #[test]
    fn test_parse_package_condition_1_5() {
        let result = ConditionParser::parse("[10P1..5]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Package {
                id: 10,
                min: 1,
                max: 5
            }
        );
    }

    #[test]
    fn test_parse_package_in_expression() {
        let result = ConditionParser::parse("X [4P0..1] ⊻ [5P0..1]")
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::Xor(
                Box::new(ConditionExpr::Package {
                    id: 4,
                    min: 0,
                    max: 1
                }),
                Box::new(ConditionExpr::Package {
                    id: 5,
                    min: 0,
                    max: 1
                }),
            )
        );
    }

    #[test]
    fn test_parse_package_bare_p() {
        let result = ConditionParser::parse("[1P]").unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Package {
                id: 1,
                min: 0,
                max: u32::MAX
            }
        );
    }

    #[test]
    fn test_condition_ids_extraction_full() {
        let result = ConditionParser::parse("Muss ([102] ∧ [2006]) ⊻ ([103] ∧ [2005])")
            .unwrap()
            .unwrap();
        let ids = result.condition_ids();
        assert!(ids.contains(&102));
        assert!(ids.contains(&103));
        assert!(ids.contains(&2005));
        assert!(ids.contains(&2006));
        assert_eq!(ids.len(), 4);
    }

    // === UB inline expansion ===

    #[test]
    fn test_parse_ub_inline_expansion() {
        let ub1_expr = ConditionParser::parse("[931] ∧ [932]").unwrap().unwrap();
        let mut ub_map = BTreeMap::new();
        ub_map.insert("UB1".to_string(), ub1_expr.clone());

        let result = ConditionParser::parse_with_ub("X [UB1]", &ub_map)
            .unwrap()
            .unwrap();
        assert_eq!(result, ub1_expr);
    }

    /// Statuses are kept per thread once parsed; one naming a UB condition
    /// depends on the UB definitions it is parsed with and must follow them.
    #[test]
    fn a_kept_status_follows_its_ub_definitions() {
        let with = |expr: &str| {
            let mut ub_map = BTreeMap::new();
            ub_map.insert(
                "UB1".to_string(),
                ConditionParser::parse(expr).unwrap().unwrap(),
            );
            ConditionParser::parse_with_ub("X [931] ∧ [UB1]", &ub_map)
                .unwrap()
                .unwrap()
        };
        let first = with("[1]");
        assert_eq!(first, with("[1]"));
        assert_ne!(first, with("[2]"));
        // A status without one is the same however often it is asked for.
        let plain = ConditionParser::parse("Muss [92] ∧ [126]").unwrap();
        assert_eq!(plain, ConditionParser::parse("Muss [92] ∧ [126]").unwrap());
    }

    #[test]
    fn test_parse_ub_unknown_falls_back() {
        let ub_map = BTreeMap::new();
        let result = ConditionParser::parse_with_ub("[UB99]", &ub_map)
            .unwrap()
            .unwrap();
        assert_eq!(result, ConditionExpr::Ref(0));
    }

    #[test]
    fn test_parse_with_ub_empty_map_same_as_parse() {
        let ub_map = BTreeMap::new();
        // Normal conditions should work unchanged
        let result = ConditionParser::parse_with_ub("X [931] ∧ [932]", &ub_map)
            .unwrap()
            .unwrap();
        let expected = ConditionParser::parse("X [931] ∧ [932]").unwrap().unwrap();
        assert_eq!(result, expected);
    }

    #[test]
    fn test_parse_ub_in_complex_expression() {
        // UB expands inside a larger expression
        let ub1_expr = ConditionParser::parse("[931] ∧ [932]").unwrap().unwrap();
        let mut ub_map = BTreeMap::new();
        ub_map.insert("UB1".to_string(), ub1_expr);

        let result = ConditionParser::parse_with_ub("X [UB1] ∨ [100]", &ub_map)
            .unwrap()
            .unwrap();
        // Should be: ([931] ∧ [932]) ∨ [100]
        assert_eq!(
            result,
            ConditionExpr::Or(vec![
                ConditionExpr::And(vec![ConditionExpr::Ref(931), ConditionExpr::Ref(932)]),
                ConditionExpr::Ref(100),
            ])
        );
    }

    #[test]
    fn test_parse_ub_multiple_references() {
        // Two different UB references in one expression
        let ub1_expr = ConditionParser::parse("[931]").unwrap().unwrap();
        let ub2_expr = ConditionParser::parse("[932]").unwrap().unwrap();
        let mut ub_map = BTreeMap::new();
        ub_map.insert("UB1".to_string(), ub1_expr);
        ub_map.insert("UB2".to_string(), ub2_expr);

        let result = ConditionParser::parse_with_ub("X [UB1] ∧ [UB2]", &ub_map)
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(931), ConditionExpr::Ref(932)])
        );
    }

    #[test]
    fn test_parse_multi_line_alternatives_are_or() {
        // AHB_Status with alternative branches on separate lines, each with its
        // own status prefix. Branches form a disjunction (at least one must hold).
        let status = "Muss [315] ∧ [707]\nSoll [8] ∧ [301] ∧ [707]";
        let result = ConditionParser::parse(status).unwrap().unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![
                ConditionExpr::And(vec![ConditionExpr::Ref(315), ConditionExpr::Ref(707)]),
                ConditionExpr::And(vec![
                    ConditionExpr::Ref(8),
                    ConditionExpr::Ref(301),
                    ConditionExpr::Ref(707),
                ]),
            ])
        );
    }

    #[test]
    fn test_parse_multi_line_with_crlf() {
        // XML parsing produces \r\n line endings for multi-line AHB_Status values.
        let result = ConditionParser::parse("Muss [10]\r\nSoll [20]")
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![ConditionExpr::Ref(10), ConditionExpr::Ref(20)])
        );
    }

    #[test]
    fn test_parse_single_line_unchanged() {
        // Single-line inputs continue to produce a bare expression (not wrapped in Or).
        let result = ConditionParser::parse("Muss [315] ∧ [707]")
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::And(vec![ConditionExpr::Ref(315), ConditionExpr::Ref(707)])
        );
    }

    // --- parse_status_branches: one branch per status line ---

    fn branches(status: &str) -> Vec<StatusBranch> {
        ConditionParser::parse_status_branches(status, &BTreeMap::new()).unwrap()
    }

    #[test]
    fn test_status_branches_keep_each_lines_status() {
        assert_eq!(
            branches("Muss [2119] \r\nSoll [130]"),
            vec![
                StatusBranch {
                    kind: Some(StatusKind::Muss),
                    condition: Some(ConditionExpr::Ref(2119)),
                },
                StatusBranch {
                    kind: Some(StatusKind::Soll),
                    condition: Some(ConditionExpr::Ref(130)),
                },
            ]
        );
    }

    #[test]
    fn test_status_branches_bare_line_has_no_condition() {
        assert_eq!(
            branches("Muss [48]\r\nKann"),
            vec![
                StatusBranch {
                    kind: Some(StatusKind::Muss),
                    condition: Some(ConditionExpr::Ref(48)),
                },
                StatusBranch {
                    kind: Some(StatusKind::Kann),
                    condition: None,
                },
            ]
        );
    }

    #[test]
    fn test_status_branches_x_is_muss() {
        assert_eq!(
            branches("X [931]"),
            vec![StatusBranch {
                kind: Some(StatusKind::Muss),
                condition: Some(ConditionExpr::Ref(931)),
            }]
        );
    }

    #[test]
    fn test_status_branches_split_a_second_status_on_the_same_line() {
        // IFTSTA writes some two-status cells without a line break.
        assert_eq!(
            branches("Muss [56] ∧ [59] Soll [70]"),
            vec![
                StatusBranch {
                    kind: Some(StatusKind::Muss),
                    condition: Some(ConditionExpr::And(vec![
                        ConditionExpr::Ref(56),
                        ConditionExpr::Ref(59),
                    ])),
                },
                StatusBranch {
                    kind: Some(StatusKind::Soll),
                    condition: Some(ConditionExpr::Ref(70)),
                },
            ]
        );
    }

    #[test]
    fn test_status_branches_do_not_split_inside_parentheses() {
        // The PID schema joins nested group statuses into one parent status.
        // IFTSTA 21038's SG18 carries a one-line two-status cell, which ends up
        // parenthesised: splitting there would cut the parentheses apart.
        let status = "Muss ([525]) ∧ ([56] ∧ [59] Soll [70]) ∧ ([54])";
        let found = branches(status);
        assert_eq!(found.len(), 1, "{found:?}");
        assert_eq!(found[0].kind, Some(StatusKind::Muss));
        assert_eq!(
            found[0].condition,
            ConditionParser::parse_raw("([525]) ∧ ([56] ∧ [59] [70]) ∧ ([54])").unwrap()
        );
    }

    #[test]
    fn test_parse_same_line_second_status_is_or() {
        let result = ConditionParser::parse("Muss [56] ∧ [59] Soll [70]")
            .unwrap()
            .unwrap();
        assert_eq!(
            result,
            ConditionExpr::Or(vec![
                ConditionExpr::And(vec![ConditionExpr::Ref(56), ConditionExpr::Ref(59)]),
                ConditionExpr::Ref(70),
            ])
        );
    }
}