//! Lexer: zerlegt Quelltext in Tokens.
//!
//! Besonderheiten des Dialekts:
//! - Typ-Suffixe an Bezeichnern und Literalen (`%  &  !  #  $  @`)
//! - Zeilennummern und Labels (löst der Parser auf)
//! - Keywords case-insensitiv; nur echte Sprach-Keywords sind reserviert,
//!   Bibliotheksnamen (`CLS`, `LEFT$` …) bleiben Bezeichner und werden in
//!   der Semantik als Builtins aufgelöst
//! - `:` trennt Anweisungen, `'` und `REM` leiten Kommentare ein
//! - Zeilenfortsetzung: `_` als letztes Zeichen nach Leerraum
//! - Literal-Typisierung nach Sprachreferenz (docs/sprachreferenz.md §1)

use crate::{Diagnostic, SourcePos};

/// Typ-Suffix eines Bezeichners oder Literals.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Suffix {
    Integer,  // %
    Long,     // &
    Single,   // !
    Double,   // #
    Str,      // $
    Currency, // @
}

impl Suffix {
    pub fn from_char(c: char) -> Option<Suffix> {
        match c {
            '%' => Some(Suffix::Integer),
            '&' => Some(Suffix::Long),
            '!' => Some(Suffix::Single),
            '#' => Some(Suffix::Double),
            '$' => Some(Suffix::Str),
            '@' => Some(Suffix::Currency),
            _ => None,
        }
    }
    pub fn as_char(self) -> char {
        match self {
            Suffix::Integer => '%',
            Suffix::Long => '&',
            Suffix::Single => '!',
            Suffix::Double => '#',
            Suffix::Str => '$',
            Suffix::Currency => '@',
        }
    }
}

/// Wert eines numerischen Literals, bereits typisiert.
#[derive(Debug, Clone, Copy, PartialEq)]
pub enum NumValue {
    Int(i16),
    Long(i32),
    Single(f32),
    Double(f64),
    /// Festkomma ×10 000
    Currency(i64),
}

/// Reservierte Sprach-Keywords (bewusst schlank: Bibliotheksfunktionen
/// und -anweisungen sind KEINE Keywords, sondern Builtins der Semantik).
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Kw {
    And,
    As,
    Call,
    Case,
    Common,
    Const,
    Data,
    Declare,
    Def,
    DefCur,
    DefDbl,
    DefInt,
    DefLng,
    DefSng,
    DefStr,
    Dim,
    Do,
    Double,
    Else,
    ElseIf,
    End,
    Eqv,
    Erase,
    Error,
    Exit,
    For,
    Function,
    Gosub,
    Goto,
    If,
    Imp,
    Input,
    Integer,
    Is,
    Let,
    Line,
    Local,
    Long,
    Loop,
    Mod,
    Next,
    Not,
    On,
    Option,
    Or,
    Print,
    Read,
    ReDim,
    Rem,
    Restore,
    Resume,
    Return,
    Select,
    Shared,
    Single,
    Static,
    Step,
    Stop,
    String,
    Sub,
    System,
    Then,
    To,
    Type,
    Until,
    Using,
    Wend,
    While,
    Xor,
    Currency,
    // Datei-E/A-Keywords: werden geparst, aber erst in Phase 3 implementiert
    Open,
    Close,
    Write,
    Field,
    Get,
    Put,
    Seek,
    Lset,
    Rset,
}

fn keyword(upper: &str) -> Option<Kw> {
    use Kw::*;
    Some(match upper {
        "AND" => And,
        "AS" => As,
        "CALL" => Call,
        "CASE" => Case,
        "COMMON" => Common,
        "CONST" => Const,
        "CURRENCY" => Currency,
        "DATA" => Data,
        "DECLARE" => Declare,
        "DEF" => Def,
        "DEFCUR" => DefCur,
        "DEFDBL" => DefDbl,
        "DEFINT" => DefInt,
        "DEFLNG" => DefLng,
        "DEFSNG" => DefSng,
        "DEFSTR" => DefStr,
        "DIM" => Dim,
        "DO" => Do,
        "DOUBLE" => Double,
        "ELSE" => Else,
        "ELSEIF" => ElseIf,
        "END" => End,
        "EQV" => Eqv,
        "ERASE" => Erase,
        "ERROR" => Error,
        "EXIT" => Exit,
        "FIELD" => Field,
        "FOR" => For,
        "FUNCTION" => Function,
        "GET" => Get,
        "GOSUB" => Gosub,
        "GOTO" => Goto,
        "IF" => If,
        "IMP" => Imp,
        "INPUT" => Input,
        "INTEGER" => Integer,
        "IS" => Is,
        "LET" => Let,
        "LINE" => Line,
        "LOCAL" => Local,
        "LONG" => Long,
        "LOOP" => Loop,
        "LSET" => Lset,
        "MOD" => Mod,
        "NEXT" => Next,
        "NOT" => Not,
        "ON" => On,
        "OPEN" => Open,
        "OPTION" => Option,
        "OR" => Or,
        "PRINT" => Print,
        "PUT" => Put,
        "READ" => Read,
        "REDIM" => ReDim,
        "REM" => Rem,
        "RESTORE" => Restore,
        "RESUME" => Resume,
        "RETURN" => Return,
        "RSET" => Rset,
        "SEEK" => Seek,
        "SELECT" => Select,
        "SHARED" => Shared,
        "SINGLE" => Single,
        "STATIC" => Static,
        "STEP" => Step,
        "STOP" => Stop,
        "STRING" => String,
        "SUB" => Sub,
        "SYSTEM" => System,
        "THEN" => Then,
        "TO" => To,
        "TYPE" => Type,
        "UNTIL" => Until,
        "USING" => Using,
        "WEND" => Wend,
        "WHILE" => While,
        "WRITE" => Write,
        "XOR" => Xor,
        "CLOSE" => Close,
        _ => return None,
    })
}

#[derive(Debug, Clone, PartialEq)]
pub enum TokenKind {
    /// Bezeichner; `name` ist bereits in Großschreibung normalisiert.
    Ident {
        name: String,
        suffix: Option<Suffix>,
    },
    Kw(Kw),
    Num(NumValue),
    Str(String),
    Plus,
    Minus,
    Star,
    Slash,
    Backslash,
    Caret,
    Eq,
    Ne,
    Lt,
    Le,
    Gt,
    Ge,
    LParen,
    RParen,
    Comma,
    Semicolon,
    Colon,
    Hash,
    Dot,
    /// Metabefehl `'$INCLUDE: 'datei''` (Pfad; leer = fehlerhafte Syntax).
    MetaInclude(String),
    /// Metabefehle `'$STATIC` / `'$DYNAMIC`.
    MetaStatic,
    MetaDynamic,
    /// Metabefehl `'$FORM`: das Modul ist ein Formularmodul.
    MetaForm,
    /// Rohtext einer `DATA`-Anweisung bis zum Anweisungsende. Er wird
    /// bewusst **nicht** zerlegt oder normalisiert: unquotierte Elemente
    /// behalten ihre Groß- und Kleinschreibung und ihren inneren Leerraum.
    DataRaw(String),
    /// Ende einer logischen Zeile.
    Eol,
    Eof,
}

/// Prüft einen Kommentar auf Metabefehle (`$INCLUDE`, `$STATIC`, `$DYNAMIC`).
fn meta_token(rest: &[char]) -> Option<TokenKind> {
    let s: String = rest.iter().collect();
    let t = s.trim_start();
    let up = t.to_uppercase();
    if up.starts_with("$STATIC") {
        return Some(TokenKind::MetaStatic);
    }
    if up.starts_with("$DYNAMIC") {
        return Some(TokenKind::MetaDynamic);
    }
    if up.starts_with("$FORM") {
        return Some(TokenKind::MetaForm);
    }
    if up.starts_with("$INCLUDE") {
        if let Some(colon) = t.find(':') {
            let after = t[colon + 1..].trim();
            if let Some(stripped) = after.strip_prefix('\'') {
                if let Some(end) = stripped.find('\'') {
                    return Some(TokenKind::MetaInclude(stripped[..end].to_string()));
                }
            }
        }
        return Some(TokenKind::MetaInclude(String::new()));
    }
    None
}

#[derive(Debug, Clone, PartialEq)]
pub struct Token {
    pub kind: TokenKind,
    pub pos: SourcePos,
}

pub struct LexOutput {
    pub tokens: Vec<Token>,
    pub diagnostics: Vec<Diagnostic>,
}

/// Signifikante Stellen einer Ziffernfolge (führende Nullen zählen nicht).
fn significant_digits(int_part: &str, frac_part: &str) -> usize {
    let all: String = int_part.chars().chain(frac_part.chars()).collect();
    let trimmed = all.trim_start_matches('0');
    trimmed.len()
}

pub fn lex(source: &str) -> LexOutput {
    let mut tokens: Vec<Token> = Vec::new();
    let mut diagnostics: Vec<Diagnostic> = Vec::new();
    let mut continuation = false;

    for (line_idx, raw_line) in source.lines().enumerate() {
        let line_no = (line_idx + 1) as u32;
        let chars: Vec<char> = raw_line.chars().collect();
        let mut i = 0usize;
        let mut line_continued = false;

        'line: while i < chars.len() {
            // Leerraum überspringen
            while i < chars.len() && (chars[i] == ' ' || chars[i] == '\t') {
                i += 1;
            }
            if i >= chars.len() {
                break;
            }
            let start = i;
            let pos = SourcePos {
                source: 0,
                line: line_no,
                column: (start + 1) as u32,
            };
            let c = chars[i];

            // Zeilenfortsetzung: `_` nach Leerraum, danach nur noch Leerraum
            if c == '_'
                && (start == 0 || chars[start - 1] == ' ' || chars[start - 1] == '\t')
                && chars[start + 1..].iter().all(|&ch| ch == ' ' || ch == '\t')
            {
                line_continued = true;
                break 'line;
            }

            match c {
                '\'' => {
                    // Kommentar bis Zeilenende; ggf. Metabefehl
                    if let Some(tok) = meta_token(&chars[i + 1..]) {
                        tokens.push(Token { kind: tok, pos });
                    }
                    break 'line;
                }
                '"' => {
                    i += 1;
                    let mut s = String::new();
                    let mut closed = false;
                    while i < chars.len() {
                        if chars[i] == '"' {
                            if i + 1 < chars.len() && chars[i + 1] == '"' {
                                s.push('"');
                                i += 2;
                            } else {
                                i += 1;
                                closed = true;
                                break;
                            }
                        } else {
                            s.push(chars[i]);
                            i += 1;
                        }
                    }
                    if !closed {
                        // Das Vorbild toleriert fehlende schließende
                        // Anführungszeichen am Zeilenende.
                    }
                    tokens.push(Token {
                        kind: TokenKind::Str(s),
                        pos,
                    });
                }
                '&' if i + 1 < chars.len() && matches!(chars[i + 1], 'h' | 'H' | 'o' | 'O') => {
                    let hex = matches!(chars[i + 1], 'h' | 'H');
                    i += 2;
                    let digit_start = i;
                    while i < chars.len() && chars[i].is_ascii_alphanumeric() {
                        i += 1;
                    }
                    let digits: String = chars[digit_start..i].iter().collect();
                    let long_suffix = i < chars.len() && chars[i] == '&';
                    if long_suffix {
                        i += 1;
                    }
                    let radix = if hex { 16 } else { 8 };
                    match u32::from_str_radix(&digits, radix) {
                        Ok(v) => {
                            let kind = if long_suffix {
                                TokenKind::Num(NumValue::Long(v as i32))
                            } else if v <= 0xFFFF {
                                TokenKind::Num(NumValue::Int(v as u16 as i16))
                            } else {
                                diagnostics.push(Diagnostic {
                                    file: None,
                                    pos,
                                    message: "Overflow".into(),
                                });
                                TokenKind::Num(NumValue::Long(v as i32))
                            };
                            tokens.push(Token { kind, pos });
                        }
                        Err(_) => diagnostics.push(Diagnostic {
                            file: None,
                            pos,
                            message: "Syntax error".into(),
                        }),
                    }
                }
                '0'..='9' | '.'
                    if c != '.' || (i + 1 < chars.len() && chars[i + 1].is_ascii_digit()) =>
                {
                    let int_start = i;
                    while i < chars.len() && chars[i].is_ascii_digit() {
                        i += 1;
                    }
                    let int_part: String = chars[int_start..i].iter().collect();
                    let mut frac_part = String::new();
                    let mut has_point = false;
                    if i < chars.len() && chars[i] == '.' {
                        has_point = true;
                        i += 1;
                        let fs = i;
                        while i < chars.len() && chars[i].is_ascii_digit() {
                            i += 1;
                        }
                        frac_part = chars[fs..i].iter().collect();
                    }
                    // Exponent E/D
                    let mut exp_kind: Option<char> = None;
                    let mut exp_str = String::new();
                    if i < chars.len() && matches!(chars[i], 'e' | 'E' | 'd' | 'D') {
                        let save = i;
                        let k = chars[i].to_ascii_uppercase();
                        let mut j = i + 1;
                        let mut e = String::new();
                        if j < chars.len() && (chars[j] == '+' || chars[j] == '-') {
                            e.push(chars[j]);
                            j += 1;
                        }
                        let ds = j;
                        while j < chars.len() && chars[j].is_ascii_digit() {
                            j += 1;
                        }
                        if j > ds {
                            for ch in &chars[ds..j] {
                                e.push(*ch);
                            }
                            exp_kind = Some(k);
                            exp_str = e;
                            i = j;
                        } else {
                            i = save; // kein Exponent (z. B. Variable `e`)
                        }
                    }
                    // Suffix
                    let suffix = if i < chars.len() {
                        Suffix::from_char(chars[i])
                    } else {
                        None
                    };
                    if suffix.is_some() {
                        i += 1;
                    }
                    let text = format!(
                        "{}{}{}{}",
                        int_part,
                        if has_point { "." } else { "" },
                        frac_part,
                        match exp_kind {
                            Some(_) => format!("e{exp_str}"),
                            None => String::new(),
                        }
                    );
                    let dval: f64 = text.parse().unwrap_or(0.0);
                    let value = match (suffix, exp_kind) {
                        (Some(Suffix::Integer), _) => {
                            if dval > i16::MAX as f64 || dval < i16::MIN as f64 {
                                diagnostics.push(Diagnostic {
                                    file: None,
                                    pos,
                                    message: "Overflow".into(),
                                });
                            }
                            NumValue::Int(dval as i16)
                        }
                        (Some(Suffix::Long), _) => {
                            if dval > i32::MAX as f64 || dval < i32::MIN as f64 {
                                diagnostics.push(Diagnostic {
                                    file: None,
                                    pos,
                                    message: "Overflow".into(),
                                });
                            }
                            NumValue::Long(dval as i32)
                        }
                        (Some(Suffix::Single), _) => NumValue::Single(dval as f32),
                        (Some(Suffix::Double), _) => NumValue::Double(dval),
                        (Some(Suffix::Currency), _) => {
                            NumValue::Currency((dval * 10_000.0).round() as i64)
                        }
                        (Some(Suffix::Str), _) => {
                            diagnostics.push(Diagnostic {
                                file: None,
                                pos,
                                message: "Syntax error".into(),
                            });
                            NumValue::Double(dval)
                        }
                        (None, Some('D')) => NumValue::Double(dval),
                        (None, Some(_)) => NumValue::Single(dval as f32),
                        (None, None) => {
                            if !has_point {
                                // Ganzzahl: INTEGER → LONG → Gleitkomma
                                if let Ok(v) = text.parse::<i64>() {
                                    if let Ok(v16) = i16::try_from(v) {
                                        NumValue::Int(v16)
                                    } else if let Ok(v32) = i32::try_from(v) {
                                        NumValue::Long(v32)
                                    } else {
                                        NumValue::Double(dval)
                                    }
                                } else {
                                    NumValue::Double(dval)
                                }
                            } else {
                                // Entscheidung (2026-09-02, siehe
                                // Sprachreferenz §1): > 7 signifikante
                                // Stellen → DOUBLE, sonst SINGLE.
                                if significant_digits(&int_part, &frac_part) > 7 {
                                    NumValue::Double(dval)
                                } else {
                                    NumValue::Single(dval as f32)
                                }
                            }
                        }
                    };
                    tokens.push(Token {
                        kind: TokenKind::Num(value),
                        pos,
                    });
                }
                c if c.is_alphabetic() => {
                    i += 1;
                    while i < chars.len()
                        && (chars[i].is_alphanumeric() || chars[i] == '.' || chars[i] == '_')
                    {
                        i += 1;
                    }
                    let mut name: String =
                        chars[start..i].iter().collect::<String>().to_uppercase();
                    let container_bang = i < chars.len()
                        && chars[i] == '!'
                        && i + 1 < chars.len()
                        && chars[i + 1].is_alphabetic();
                    let suffix = if i < chars.len() && !container_bang {
                        Suffix::from_char(chars[i])
                    } else {
                        None
                    };
                    if suffix.is_some() {
                        i += 1;
                    }
                    if container_bang {
                        name.push('!');
                        i += 1;
                        while i < chars.len()
                            && (chars[i].is_alphanumeric() || chars[i] == '.' || chars[i] == '_')
                        {
                            name.push(chars[i].to_ascii_uppercase());
                            i += 1;
                        }
                    }
                    if suffix.is_none() {
                        if let Some(kw) = keyword(&name) {
                            if kw == Kw::Rem {
                                // REM: Rest ist Kommentar; ggf. Metabefehl
                                if let Some(tok) = meta_token(&chars[i..]) {
                                    tokens.push(Token { kind: tok, pos });
                                }
                                break 'line;
                            }
                            tokens.push(Token {
                                kind: TokenKind::Kw(kw),
                                pos,
                            });
                            if kw == Kw::Data {
                                // Der Rest der Anweisung ist Rohtext: bis zum
                                // `:` außerhalb von Anführungszeichen oder bis
                                // zum Zeilenende.
                                let start = i;
                                let mut in_quote = false;
                                while i < chars.len() {
                                    match chars[i] {
                                        '"' => in_quote = !in_quote,
                                        ':' if !in_quote => break,
                                        '\'' if !in_quote => break,
                                        _ => {}
                                    }
                                    i += 1;
                                }
                                let roh: String = chars[start..i].iter().collect();
                                tokens.push(Token {
                                    kind: TokenKind::DataRaw(roh),
                                    pos,
                                });
                            }
                            continue;
                        }
                    }
                    // Bezeichner dürfen nicht mit '.' enden (a.b. → a.b + .)
                    while name.ends_with('.') {
                        name.pop();
                        i -= 1;
                    }
                    tokens.push(Token {
                        kind: TokenKind::Ident { name, suffix },
                        pos,
                    });
                }
                _ => {
                    i += 1;
                    let kind = match c {
                        '+' => TokenKind::Plus,
                        '-' => TokenKind::Minus,
                        '*' => TokenKind::Star,
                        '/' => TokenKind::Slash,
                        '\\' => TokenKind::Backslash,
                        '^' => TokenKind::Caret,
                        '=' => TokenKind::Eq,
                        '(' => TokenKind::LParen,
                        ')' => TokenKind::RParen,
                        ',' => TokenKind::Comma,
                        ';' => TokenKind::Semicolon,
                        ':' => TokenKind::Colon,
                        '#' => TokenKind::Hash,
                        '.' => TokenKind::Dot,
                        '?' => TokenKind::Kw(Kw::Print), // Editor-Kurzform
                        '&' => TokenKind::Kw(Kw::Long),  // isoliertes & (selten)
                        '<' => {
                            if i < chars.len() && chars[i] == '=' {
                                i += 1;
                                TokenKind::Le
                            } else if i < chars.len() && chars[i] == '>' {
                                i += 1;
                                TokenKind::Ne
                            } else {
                                TokenKind::Lt
                            }
                        }
                        '>' => {
                            if i < chars.len() && chars[i] == '=' {
                                i += 1;
                                TokenKind::Ge
                            } else {
                                TokenKind::Gt
                            }
                        }
                        other => {
                            diagnostics.push(Diagnostic {
                                file: None,
                                pos,
                                message: format!("Syntax error ('{other}')"),
                            });
                            continue;
                        }
                    };
                    tokens.push(Token { kind, pos });
                }
            }
        }

        if line_continued {
            continuation = true;
        } else {
            if !continuation || !tokens.is_empty() {
                tokens.push(Token {
                    kind: TokenKind::Eol,
                    pos: SourcePos {
                        source: 0,
                        line: line_no,
                        column: (chars.len() + 1) as u32,
                    },
                });
            }
            continuation = false;
        }
    }

    tokens.push(Token {
        kind: TokenKind::Eof,
        pos: SourcePos {
            source: 0,
            line: (source.lines().count() + 1) as u32,
            column: 1,
        },
    });
    LexOutput {
        tokens,
        diagnostics,
    }
}

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

    fn kinds(src: &str) -> Vec<TokenKind> {
        lex(src).tokens.into_iter().map(|t| t.kind).collect()
    }

    #[test]
    fn keywords_case_insensitiv() {
        let k = kinds("print If tHeN");
        assert_eq!(k[0], TokenKind::Kw(Kw::Print));
        assert_eq!(k[1], TokenKind::Kw(Kw::If));
        assert_eq!(k[2], TokenKind::Kw(Kw::Then));
    }

    #[test]
    fn suffix_macht_keyword_zum_bezeichner() {
        // STRING$ ist die Bibliotheksfunktion, STRING das Typ-Keyword
        let k = kinds("STRING$ STRING");
        assert_eq!(
            k[0],
            TokenKind::Ident {
                name: "STRING".into(),
                suffix: Some(Suffix::Str)
            }
        );
        assert_eq!(k[1], TokenKind::Kw(Kw::String));
    }

    #[test]
    fn literal_typisierung() {
        assert_eq!(kinds("42")[0], TokenKind::Num(NumValue::Int(42)));
        assert_eq!(kinds("40000")[0], TokenKind::Num(NumValue::Long(40000)));
        assert_eq!(kinds("1.5")[0], TokenKind::Num(NumValue::Single(1.5)));
        assert_eq!(
            kinds("1.23456789")[0],
            TokenKind::Num(NumValue::Double(1.23456789))
        );
        assert_eq!(kinds("1E3")[0], TokenKind::Num(NumValue::Single(1000.0)));
        assert_eq!(kinds("1D3")[0], TokenKind::Num(NumValue::Double(1000.0)));
        assert_eq!(kinds("2.5@")[0], TokenKind::Num(NumValue::Currency(25000)));
        assert_eq!(kinds("&HFF")[0], TokenKind::Num(NumValue::Int(255)));
        assert_eq!(kinds("&HFFFF")[0], TokenKind::Num(NumValue::Int(-1)));
        assert_eq!(kinds("&HFFFF&")[0], TokenKind::Num(NumValue::Long(0xFFFF)));
        assert_eq!(kinds("&O777")[0], TokenKind::Num(NumValue::Int(511)));
    }

    #[test]
    fn strings_mit_doppelten_anfuehrungszeichen() {
        assert_eq!(kinds("\"a\"\"b\"")[0], TokenKind::Str("a\"b".into()));
    }

    #[test]
    fn kommentare_und_rem() {
        let k = kinds("PRINT 1 ' Kommentar\nREM ganze Zeile\nPRINT 2");
        // PRINT 1 EOL EOL PRINT 2 EOL EOF
        assert_eq!(k.len(), 8);
        assert_eq!(k[2], TokenKind::Eol);
        assert_eq!(k[3], TokenKind::Eol);
    }

    #[test]
    fn zeilenfortsetzung() {
        let k = kinds("PRINT 1, _\n  2");
        // Kein Eol zwischen 1, und 2
        assert!(matches!(k[3], TokenKind::Num(NumValue::Int(2))));
    }

    #[test]
    fn bezeichner_mit_punkt_und_unterstrich() {
        let k = kinds("kunde.name_2$");
        assert_eq!(
            k[0],
            TokenKind::Ident {
                name: "KUNDE.NAME_2".into(),
                suffix: Some(Suffix::Str)
            }
        );
    }

    #[test]
    fn bang_ist_suffix_oder_containeroperator_nach_folgetoken() {
        let k = kinds("Wert! = 1.5\nForm1!Text1.Text = \"a\"");
        assert_eq!(
            k[0],
            TokenKind::Ident {
                name: "WERT".into(),
                suffix: Some(Suffix::Single)
            }
        );
        assert_eq!(
            k[4],
            TokenKind::Ident {
                name: "FORM1!TEXT1.TEXT".into(),
                suffix: None
            }
        );
    }
}