chipi_core/

codegen_cpp.rs

//! C++ code generation from validated definitions and decision trees.
//!
//! Generates a single-header C++ file with instruction enum, decode function,
//! and disassembly formatting.

use std::collections::HashMap;
use std::fmt::Write;

use crate::config::{CppGuardStyle as GuardStyle, CppOptions};
use crate::tree::DecodeNode;
use crate::types::*;

/// Generate a complete C++ header file.
pub fn generate_cpp_code(
    def: &ValidatedDef,
    tree: &DecodeNode,
    opts: &CppOptions,
    type_maps: &HashMap<String, String>,
) -> String {
    let mut out = String::new();

    let default_ns = to_snake_case(&def.config.name);
    let ns = opts.namespace.as_deref().unwrap_or(&default_ns);
    let guard_name = format!("CHIPI_{}_HPP", ns.to_ascii_uppercase());
    let unit_bytes = def.config.width / 8;
    let word_type = cpp_word_type(def.config.width);
    let endian = &def.config.endian;
    let variable_length = def.instructions.iter().any(|i| i.unit_count() > 1);

    // Header guard
    match opts.guard_style {
        GuardStyle::Pragma => writeln!(out, "#pragma once").unwrap(),
        GuardStyle::Ifndef => {
            writeln!(out, "#ifndef {}", guard_name).unwrap();
            writeln!(out, "#define {}", guard_name).unwrap();
        }
    }
    writeln!(out).unwrap();
    writeln!(
        out,
        "// Auto-generated by https://github.com/ioncodes/chipi"
    )
    .unwrap();
    writeln!(out, "// Do not edit.").unwrap();
    writeln!(out).unwrap();

    // Includes
    writeln!(out, "#include <cstdint>").unwrap();
    writeln!(out, "#include <cstddef>").unwrap();
    writeln!(out, "#include <cstring>").unwrap();
    writeln!(out, "#include <string>").unwrap();
    writeln!(out, "#include <optional>").unwrap();
    writeln!(out, "#include <format>").unwrap();
    for inc in &opts.includes {
        writeln!(out, "#include \"{}\"", inc).unwrap();
    }
    writeln!(out).unwrap();

    writeln!(out, "namespace {} {{", ns).unwrap();
    writeln!(out).unwrap();

    // Built-in display wrapper types for type aliases with display hints
    emit_display_types(&mut out, def, type_maps);

    // Opcode enum
    emit_opcode_enum(&mut out, def);

    // Instruction struct
    emit_instruction_struct(&mut out, def, type_maps);

    // Sub-decoder types and functions
    for sd in &def.sub_decoders {
        emit_subdecoder(&mut out, sd, def.config.width);
    }

    // Map functions
    emit_map_functions(&mut out, def);

    // Decode function
    emit_decode_function(
        &mut out,
        def,
        tree,
        word_type,
        unit_bytes,
        endian,
        variable_length,
        type_maps,
    );

    // Format function
    emit_format_function(&mut out, def, type_maps, opts);

    writeln!(out, "}} // namespace {}", ns).unwrap();

    if opts.guard_style == GuardStyle::Ifndef {
        writeln!(out).unwrap();
        writeln!(out, "#endif // {}", guard_name).unwrap();
    }

    out
}

fn cpp_word_type(width: u32) -> &'static str {
    match width {
        8 => "uint8_t",
        16 => "uint16_t",
        32 => "uint32_t",
        _ => "uint32_t",
    }
}

fn cpp_signed_type(base: &str) -> &'static str {
    match base {
        "u8" | "i8" => "int8_t",
        "u16" | "i16" => "int16_t",
        "u32" | "i32" => "int32_t",
        _ => "int32_t",
    }
}

fn cpp_type_for(base: &str) -> &'static str {
    match base {
        "bool" => "bool",
        "u1" | "u2" | "u3" | "u4" | "u5" | "u6" | "u7" | "u8" => "uint8_t",
        "i8" => "int8_t",
        "u16" => "uint16_t",
        "i16" => "int16_t",
        "u32" => "uint32_t",
        "i32" => "int32_t",
        _ => "uint32_t",
    }
}

fn type_bits(base: &str) -> u32 {
    match base {
        "u8" | "i8" => 8,
        "u16" | "i16" => 16,
        "u32" | "i32" => 32,
        _ => 32,
    }
}

fn to_snake_case(name: &str) -> String {
    let mut result = String::new();
    for (i, ch) in name.chars().enumerate() {
        if ch.is_ascii_uppercase() && i > 0 {
            result.push('_');
        }
        result.push(ch.to_ascii_lowercase());
    }
    result
}

fn to_pascal_case(name: &str) -> String {
    let mut result = String::new();
    let mut cap_next = true;
    for ch in name.chars() {
        if ch == '_' {
            cap_next = true;
        } else if cap_next {
            result.push(ch.to_ascii_uppercase());
            cap_next = false;
        } else {
            result.push(ch.to_ascii_lowercase());
        }
    }
    result
}

/// Emit built-in wrapper types for type aliases with `display(hex)` or `display(signed_hex)`.
/// Only emits types that aren't already overridden via `type_map`.
fn emit_display_types(out: &mut String, def: &ValidatedDef, type_maps: &HashMap<String, String>) {
    let mut need_signed_hex = false;
    let mut need_hex = false;

    for alias in &def.type_aliases {
        // Skip if the user provided a type_map override
        if type_maps.contains_key(&alias.name) {
            continue;
        }
        match alias.display_format {
            Some(DisplayFormat::SignedHex) => need_signed_hex = true,
            Some(DisplayFormat::Hex) => need_hex = true,
            None => {}
        }
    }

    if need_signed_hex {
        writeln!(out, "struct SignedHex {{").unwrap();
        writeln!(out, "    int32_t value;").unwrap();
        writeln!(out, "    SignedHex() = default;").unwrap();
        writeln!(out, "    constexpr SignedHex(int32_t v) : value(v) {{}}").unwrap();
        writeln!(
            out,
            "    bool operator==(const SignedHex&) const = default;"
        )
        .unwrap();
        writeln!(
            out,
            "    bool operator==(int other) const {{ return value == other; }}"
        )
        .unwrap();
        writeln!(
            out,
            "    bool operator!=(int other) const {{ return value != other; }}"
        )
        .unwrap();
        writeln!(
            out,
            "    bool operator<(int other) const {{ return value < other; }}"
        )
        .unwrap();
        writeln!(
            out,
            "    bool operator<=(int other) const {{ return value <= other; }}"
        )
        .unwrap();
        writeln!(
            out,
            "    bool operator>(int other) const {{ return value > other; }}"
        )
        .unwrap();
        writeln!(
            out,
            "    bool operator>=(int other) const {{ return value >= other; }}"
        )
        .unwrap();
        writeln!(
            out,
            "    SignedHex operator-() const {{ return SignedHex(-value); }}"
        )
        .unwrap();
        writeln!(out, "    friend SignedHex operator-(int lhs, SignedHex rhs) {{ return SignedHex(lhs - rhs.value); }}").unwrap();
        writeln!(out, "    friend SignedHex operator+(int lhs, SignedHex rhs) {{ return SignedHex(lhs + rhs.value); }}").unwrap();
        writeln!(out, "    friend SignedHex operator+(SignedHex lhs, int rhs) {{ return SignedHex(lhs.value + rhs); }}").unwrap();
        writeln!(out, "    friend SignedHex operator-(SignedHex lhs, int rhs) {{ return SignedHex(lhs.value - rhs); }}").unwrap();
        writeln!(out, "    friend SignedHex operator*(SignedHex lhs, int rhs) {{ return SignedHex(lhs.value * rhs); }}").unwrap();
        writeln!(out, "}};").unwrap();
        writeln!(out).unwrap();
    }

    if need_hex {
        writeln!(out, "struct Hex {{").unwrap();
        writeln!(out, "    uint32_t value;").unwrap();
        writeln!(out, "    Hex() = default;").unwrap();
        writeln!(out, "    constexpr Hex(uint32_t v) : value(v) {{}}").unwrap();
        writeln!(out, "    bool operator==(const Hex&) const = default;").unwrap();
        writeln!(
            out,
            "    bool operator==(unsigned other) const {{ return value == other; }}"
        )
        .unwrap();
        writeln!(out, "}};").unwrap();
        writeln!(out).unwrap();
    }

    writeln!(out, "}} // namespace {}", to_snake_case(&def.config.name)).unwrap();
    writeln!(out).unwrap();

    // std::formatter specializations must be outside the namespace
    if need_signed_hex {
        let ns = to_snake_case(&def.config.name);
        writeln!(
            out,
            "template <> struct std::formatter<{}::SignedHex> : std::formatter<std::string> {{",
            ns
        )
        .unwrap();
        writeln!(
            out,
            "    auto format({}::SignedHex v, auto& ctx) const {{",
            ns
        )
        .unwrap();
        writeln!(out, "        if (v.value < 0)").unwrap();
        writeln!(out, "            return std::formatter<std::string>::format(std::format(\"-0x{{:x}}\", static_cast<unsigned>(-v.value)), ctx);").unwrap();
        writeln!(out, "        return std::formatter<std::string>::format(std::format(\"0x{{:x}}\", static_cast<unsigned>(v.value)), ctx);").unwrap();
        writeln!(out, "    }}").unwrap();
        writeln!(out, "}};").unwrap();
        writeln!(out).unwrap();
    }

    if need_hex {
        let ns = to_snake_case(&def.config.name);
        writeln!(
            out,
            "template <> struct std::formatter<{}::Hex> : std::formatter<std::string> {{",
            ns
        )
        .unwrap();
        writeln!(out, "    auto format({}::Hex v, auto& ctx) const {{", ns).unwrap();
        writeln!(out, "        return std::formatter<std::string>::format(std::format(\"0x{{:x}}\", v.value), ctx);").unwrap();
        writeln!(out, "    }}").unwrap();
        writeln!(out, "}};").unwrap();
        writeln!(out).unwrap();
    }

    // Re-open the namespace
    if need_signed_hex || need_hex {
        writeln!(out, "namespace {} {{", to_snake_case(&def.config.name)).unwrap();
        writeln!(out).unwrap();
    }
}

/// Emit the Opcode enum.
fn emit_opcode_enum(out: &mut String, def: &ValidatedDef) {
    writeln!(out, "enum class Opcode : uint32_t {{").unwrap();
    for (i, instr) in def.instructions.iter().enumerate() {
        writeln!(out, "    {} = {},", to_pascal_case(&instr.name), i).unwrap();
    }
    writeln!(out, "}};").unwrap();
    writeln!(out).unwrap();
}

/// Emit the Instruction struct with a tagged union for fields.
fn emit_instruction_struct(
    out: &mut String,
    def: &ValidatedDef,
    type_maps: &HashMap<String, String>,
) {
    writeln!(out, "struct Instruction {{").unwrap();
    writeln!(out, "    Opcode opcode;").unwrap();
    writeln!(out, "    uint32_t size; // bytes consumed").unwrap();
    writeln!(out).unwrap();

    // Generate a union with per-instruction field structs
    let has_fields = def
        .instructions
        .iter()
        .any(|i| !i.resolved_fields.is_empty());
    if has_fields {
        writeln!(out, "    union {{").unwrap();
        for instr in &def.instructions {
            if instr.resolved_fields.is_empty() {
                continue;
            }
            writeln!(out, "        struct {{").unwrap();
            for field in &instr.resolved_fields {
                let cpp_type = field_cpp_type(field, type_maps);
                writeln!(out, "            {} {};", cpp_type, field.name).unwrap();
            }
            writeln!(out, "        }} {};", instr.name).unwrap();
        }
        writeln!(out, "    }};").unwrap();
    }

    writeln!(out, "}};").unwrap();
    writeln!(out).unwrap();
}

/// Get the C++ type for a field.
fn field_cpp_type(field: &ResolvedField, type_maps: &HashMap<String, String>) -> String {
    // Check type map first (user override)
    if let Some(alias) = &field.resolved_type.alias_name {
        if let Some(mapped) = type_maps.get(alias) {
            return mapped.clone();
        }
    }

    // Check sub-decoder
    if let Some(ref sd_name) = field.resolved_type.sub_decoder {
        return format!("{}Insn", to_pascal_case(sd_name));
    }

    // Check display format -> built-in wrapper types
    match field.resolved_type.display_format {
        Some(DisplayFormat::SignedHex) => return "SignedHex".to_string(),
        Some(DisplayFormat::Hex) => return "Hex".to_string(),
        None => {}
    }

    cpp_type_for(&field.resolved_type.base_type).to_string()
}

/// Emit sub-decoder struct and dispatch function.
fn emit_subdecoder(out: &mut String, sd: &ValidatedSubDecoder, _parent_width: u32) {
    let type_name = format!("{}Insn", to_pascal_case(&sd.name));
    let word_type = cpp_word_type(sd.width);

    // Fragment struct
    writeln!(out, "struct {} {{", type_name).unwrap();
    for frag_name in &sd.fragment_names {
        writeln!(out, "    const char* {};", frag_name).unwrap();
    }
    writeln!(out, "}};").unwrap();
    writeln!(out).unwrap();

    // Pre-baked fragment strings for each instruction
    // Then dispatch function
    let fn_name = format!("decode_{}", to_snake_case(&sd.name));
    writeln!(
        out,
        "inline std::optional<{}> {}({} val) {{",
        type_name, fn_name, word_type
    )
    .unwrap();

    for (i, instr) in sd.instructions.iter().enumerate() {
        let (mask, value) = compute_instruction_mask_value_sub(instr);
        let keyword = if i == 0 { "if" } else { "} else if" };
        writeln!(
            out,
            "    {} ((val & {:#x}) == {:#x}) {{",
            keyword, mask, value
        )
        .unwrap();

        // Build fragment values
        // For simplicity, use string literals where possible; for field-dependent
        // fragments, generate inline formatting
        for frag in &instr.fragments {
            let all_literal = frag
                .pieces
                .iter()
                .all(|p| matches!(p, FormatPiece::Literal(_)));
            if all_literal {
                let s: String = frag
                    .pieces
                    .iter()
                    .map(|p| {
                        if let FormatPiece::Literal(lit) = p {
                            lit.as_str()
                        } else {
                            ""
                        }
                    })
                    .collect();
                writeln!(out, "        // {}.{} = \"{}\"", instr.name, frag.name, s).unwrap();
            }
        }

        // Return struct with fragment values
        let frag_values: Vec<String> = instr
            .fragments
            .iter()
            .map(|frag| {
                let all_literal = frag
                    .pieces
                    .iter()
                    .all(|p| matches!(p, FormatPiece::Literal(_)));
                if all_literal {
                    let s: String = frag
                        .pieces
                        .iter()
                        .map(|p| {
                            if let FormatPiece::Literal(lit) = p {
                                lit.as_str()
                            } else {
                                ""
                            }
                        })
                        .collect();
                    format!("\"{}\"", s)
                } else {
                    // For dynamic fragments, we'd need snprintf; use empty for now
                    "\"\"".to_string()
                }
            })
            .collect();

        writeln!(
            out,
            "        return {} {{ {} }};",
            type_name,
            frag_values.join(", ")
        )
        .unwrap();
    }

    if !sd.instructions.is_empty() {
        writeln!(out, "    }}").unwrap();
    }
    writeln!(out, "    return std::nullopt;").unwrap();
    writeln!(out, "}}").unwrap();
    writeln!(out).unwrap();
}

fn compute_instruction_mask_value_sub(instr: &ValidatedSubInstruction) -> (u64, u64) {
    let mut mask: u64 = 0;
    let mut value: u64 = 0;
    for seg in &instr.segments {
        if let Segment::Fixed {
            ranges, pattern, ..
        } = seg
        {
            let mut bit_idx = 0;
            for range in ranges {
                for i in 0..range.width() {
                    if bit_idx < pattern.len() {
                        let bit = pattern[bit_idx];
                        if bit != Bit::Wildcard {
                            let hw_bit = range.start - i;
                            mask |= 1u64 << hw_bit;
                            if bit == Bit::One {
                                value |= 1u64 << hw_bit;
                            }
                        }
                        bit_idx += 1;
                    }
                }
            }
        }
    }
    (mask, value)
}

/// Emit map (lookup) functions.
fn emit_map_functions(out: &mut String, def: &ValidatedDef) {
    for map_def in &def.maps {
        let params: Vec<String> = map_def
            .params
            .iter()
            .map(|p| format!("int {}", p))
            .collect();
        writeln!(
            out,
            "inline const char* {}({}) {{",
            map_def.name,
            params.join(", ")
        )
        .unwrap();

        let key_var = if map_def.params.len() == 1 {
            map_def.params[0].clone()
        } else {
            // Multi-param: won't use switch
            String::new()
        };

        if map_def.params.len() == 1 {
            writeln!(out, "    switch ({}) {{", key_var).unwrap();
            for entry in &map_def.entries {
                if entry.keys.len() == 1 && entry.keys[0] == MapKey::Wildcard {
                    continue;
                }
                if let Some(MapKey::Value(v)) = entry.keys.first() {
                    let s = pieces_to_str(&entry.output);
                    writeln!(out, "    case {}: return \"{}\";", v, s).unwrap();
                }
            }

            let default_str = map_def
                .entries
                .iter()
                .find(|e| e.keys.len() == 1 && e.keys[0] == MapKey::Wildcard)
                .map(|e| pieces_to_str(&e.output))
                .unwrap_or_else(|| "???".to_string());
            writeln!(out, "    default: return \"{}\";", default_str).unwrap();
            writeln!(out, "    }}").unwrap();
        } else {
            // Fallback for multi-param maps: if/else chain
            for entry in &map_def.entries {
                if entry.keys.iter().any(|k| *k == MapKey::Wildcard) {
                    continue;
                }
                let conds: Vec<String> = entry
                    .keys
                    .iter()
                    .zip(map_def.params.iter())
                    .map(|(k, p)| {
                        if let MapKey::Value(v) = k {
                            format!("{} == {}", p, v)
                        } else {
                            "true".to_string()
                        }
                    })
                    .collect();
                let s = pieces_to_str(&entry.output);
                writeln!(out, "    if ({}) return \"{}\";", conds.join(" && "), s).unwrap();
            }
            writeln!(out, "    return \"???\";").unwrap();
        }

        writeln!(out, "}}").unwrap();
        writeln!(out).unwrap();
    }

    // Also emit sub-decoder maps
    for sd in &def.sub_decoders {
        for map_def in &sd.maps {
            let params: Vec<String> = map_def
                .params
                .iter()
                .map(|p| format!("int {}", p))
                .collect();
            writeln!(
                out,
                "inline const char* {}({}) {{",
                map_def.name,
                params.join(", ")
            )
            .unwrap();
            writeln!(out, "    switch ({}) {{", map_def.params[0]).unwrap();
            for entry in &map_def.entries {
                if entry.keys.len() == 1 && entry.keys[0] == MapKey::Wildcard {
                    continue;
                }
                if let Some(MapKey::Value(v)) = entry.keys.first() {
                    let s = pieces_to_str(&entry.output);
                    writeln!(out, "    case {}: return \"{}\";", v, s).unwrap();
                }
            }
            let default_str = map_def
                .entries
                .iter()
                .find(|e| e.keys.len() == 1 && e.keys[0] == MapKey::Wildcard)
                .map(|e| pieces_to_str(&e.output))
                .unwrap_or_else(|| "???".to_string());
            writeln!(out, "    default: return \"{}\";", default_str).unwrap();
            writeln!(out, "    }}").unwrap();
            writeln!(out, "}}").unwrap();
            writeln!(out).unwrap();
        }
    }
}

fn pieces_to_str(pieces: &[FormatPiece]) -> String {
    let mut s = String::new();
    for piece in pieces {
        if let FormatPiece::Literal(lit) = piece {
            s.push_str(lit);
        }
    }
    s
}

/// Emit the decode function.
fn emit_decode_function(
    out: &mut String,
    def: &ValidatedDef,
    tree: &DecodeNode,
    word_type: &str,
    unit_bytes: u32,
    endian: &ByteEndian,
    variable_length: bool,
    type_maps: &HashMap<String, String>,
) {
    writeln!(
        out,
        "inline std::optional<Instruction> decode(const uint8_t* data, size_t len) {{"
    )
    .unwrap();
    writeln!(out, "    if (len < {}) return std::nullopt;", unit_bytes).unwrap();

    // Read first unit
    emit_word_read(out, "opcode", word_type, 0, unit_bytes, endian, 1);

    emit_tree_cpp(
        out,
        tree,
        def,
        1,
        word_type,
        unit_bytes,
        endian,
        variable_length,
        type_maps,
    );

    writeln!(out, "}}").unwrap();
    writeln!(out).unwrap();
}

/// Emit a word read from the data buffer.
fn emit_word_read(
    out: &mut String,
    var_name: &str,
    word_type: &str,
    offset: u32,
    unit_bytes: u32,
    endian: &ByteEndian,
    indent: usize,
) {
    let pad = "    ".repeat(indent);
    match (unit_bytes, endian) {
        (1, _) => {
            writeln!(out, "{}{} {} = data[{}];", pad, word_type, var_name, offset).unwrap();
        }
        (2, ByteEndian::Big) => {
            writeln!(
                out,
                "{}{} {} = (static_cast<uint16_t>(data[{}]) << 8) | data[{}];",
                pad,
                word_type,
                var_name,
                offset,
                offset + 1
            )
            .unwrap();
        }
        (2, ByteEndian::Little) => {
            writeln!(
                out,
                "{}{} {} = data[{}] | (static_cast<uint16_t>(data[{}]) << 8);",
                pad,
                word_type,
                var_name,
                offset,
                offset + 1
            )
            .unwrap();
        }
        (4, ByteEndian::Big) => {
            writeln!(
                out,
                "{}{} {} = (static_cast<uint32_t>(data[{}]) << 24) | (static_cast<uint32_t>(data[{}]) << 16) | (static_cast<uint32_t>(data[{}]) << 8) | data[{}];",
                pad, word_type, var_name, offset, offset + 1, offset + 2, offset + 3
            ).unwrap();
        }
        (4, ByteEndian::Little) => {
            writeln!(
                out,
                "{}{} {} = data[{}] | (static_cast<uint32_t>(data[{}]) << 8) | (static_cast<uint32_t>(data[{}]) << 16) | (static_cast<uint32_t>(data[{}]) << 24);",
                pad, word_type, var_name, offset, offset + 1, offset + 2, offset + 3
            ).unwrap();
        }
        _ => {
            writeln!(out, "{}// unsupported width", pad).unwrap();
        }
    }
}

/// Unit read expression (inline).
fn unit_read_expr(unit: u32, _word_type: &str, unit_bytes: u32, endian: &ByteEndian) -> String {
    if unit == 0 {
        return "opcode".to_string();
    }
    let offset = unit * unit_bytes;
    match (unit_bytes, endian) {
        (1, _) => format!("data[{}]", offset),
        (2, ByteEndian::Big) => format!(
            "(static_cast<uint16_t>(data[{}]) << 8 | data[{}])",
            offset,
            offset + 1
        ),
        (2, ByteEndian::Little) => format!(
            "(data[{}] | static_cast<uint16_t>(data[{}]) << 8)",
            offset,
            offset + 1
        ),
        (4, ByteEndian::Big) => format!(
            "(static_cast<uint32_t>(data[{}]) << 24 | static_cast<uint32_t>(data[{}]) << 16 | static_cast<uint32_t>(data[{}]) << 8 | data[{}])",
            offset,
            offset + 1,
            offset + 2,
            offset + 3
        ),
        (4, ByteEndian::Little) => format!(
            "(data[{}] | static_cast<uint32_t>(data[{}]) << 8 | static_cast<uint32_t>(data[{}]) << 16 | static_cast<uint32_t>(data[{}]) << 24)",
            offset,
            offset + 1,
            offset + 2,
            offset + 3
        ),
        _ => "0".to_string(),
    }
}

/// Extract bits expression.
fn extract_expr(
    var: &str,
    ranges: &[BitRange],
    word_type: &str,
    unit_bytes: u32,
    endian: &ByteEndian,
) -> String {
    if ranges.is_empty() {
        return "0".to_string();
    }

    if ranges.len() == 1 {
        let range = ranges[0];
        let source = if range.unit == 0 {
            var.to_string()
        } else {
            unit_read_expr(range.unit, word_type, unit_bytes, endian)
        };
        let width = range.width();
        let shift = range.end;
        let mask = (1u64 << width) - 1;
        if shift == 0 {
            format!("({} & {:#x})", source, mask)
        } else {
            format!("(({} >> {}) & {:#x})", source, shift, mask)
        }
    } else {
        let mut parts = Vec::new();
        let mut accumulated = 0u32;
        for range in ranges {
            let source = if range.unit == 0 {
                var.to_string()
            } else {
                unit_read_expr(range.unit, word_type, unit_bytes, endian)
            };
            let width = range.width();
            let shift = range.end;
            let mask = (1u64 << width) - 1;
            let extracted = if shift == 0 {
                format!("({} & {:#x})", source, mask)
            } else {
                format!("(({} >> {}) & {:#x})", source, shift, mask)
            };
            if accumulated > 0 {
                parts.push(format!("({} << {})", extracted, accumulated));
            } else {
                parts.push(extracted);
            }
            accumulated += width;
        }
        parts.join(" | ")
    }
}

/// Compute mask/value for leaf guard.
fn leaf_guard(
    instr: &ValidatedInstruction,
    word_type: &str,
    unit_bytes: u32,
    endian: &ByteEndian,
) -> Option<String> {
    let fixed_bits = instr.fixed_bits();
    if fixed_bits.is_empty() {
        return None;
    }

    let mut units_map: HashMap<u32, Vec<(u32, Bit)>> = HashMap::new();
    for (unit, hw_bit, bit) in fixed_bits {
        units_map.entry(unit).or_default().push((hw_bit, bit));
    }

    let mut conditions = Vec::new();
    for (unit, bits) in &units_map {
        let (mask, value) = compute_mask_value(bits);
        if mask != 0 {
            let source = if *unit == 0 {
                "opcode".to_string()
            } else {
                unit_read_expr(*unit, word_type, unit_bytes, endian)
            };
            conditions.push(format!("({} & {:#x}) == {:#x}", source, mask, value));
        }
    }

    if conditions.is_empty() {
        None
    } else {
        Some(conditions.join(" && "))
    }
}

fn compute_mask_value(fixed_bits: &[(u32, Bit)]) -> (u64, u64) {
    let mut mask: u64 = 0;
    let mut value: u64 = 0;
    for &(bit_pos, bit_val) in fixed_bits {
        if bit_val == Bit::Wildcard {
            continue;
        }
        mask |= 1u64 << bit_pos;
        if bit_val == Bit::One {
            value |= 1u64 << bit_pos;
        }
    }
    (mask, value)
}

/// Apply transforms to an extraction expression.
fn apply_transforms(
    extract: &str,
    resolved: &ResolvedFieldType,
    type_maps: &HashMap<String, String>,
) -> String {
    let mut expr = extract.to_string();

    for transform in &resolved.transforms {
        match transform {
            Transform::SignExtend(n) => {
                let signed = cpp_signed_type(&resolved.base_type);
                let bits = type_bits(&resolved.base_type);
                expr = format!(
                    "static_cast<{}>(static_cast<{}>(({}) << ({} - {})) >> ({} - {}))",
                    cpp_type_for(&resolved.base_type),
                    signed,
                    expr,
                    bits,
                    n,
                    bits,
                    n
                );
            }
            Transform::ZeroExtend(_) => {}
            Transform::ShiftLeft(n) => {
                expr = format!("(({}) << {})", expr, n);
            }
        }
    }

    // Sub-decoder dispatch
    if let Some(ref sd_name) = resolved.sub_decoder {
        let decode_fn = format!("decode_{}", to_snake_case(sd_name));
        return format!(
            "{}(static_cast<{}>({})).value()",
            decode_fn,
            cpp_word_type(type_bits(&resolved.base_type).min(32)),
            expr
        );
    }

    // Type map wrapper (user override)
    if let Some(alias) = &resolved.alias_name {
        if let Some(mapped) = type_maps.get(alias) {
            return format!("static_cast<{}>({})", mapped, expr);
        }
    }

    // Display format -> built-in wrapper types
    match resolved.display_format {
        Some(DisplayFormat::SignedHex) => {
            return format!("SignedHex(static_cast<int32_t>({}))", expr);
        }
        Some(DisplayFormat::Hex) => {
            return format!("Hex(static_cast<uint32_t>({}))", expr);
        }
        None => {}
    }

    if resolved.base_type == "bool" {
        format!("({}) != 0", expr)
    } else {
        format!(
            "static_cast<{}>({})",
            cpp_type_for(&resolved.base_type),
            expr
        )
    }
}

/// Emit the decision tree as C++ switch/if-else.
fn emit_tree_cpp(
    out: &mut String,
    node: &DecodeNode,
    def: &ValidatedDef,
    indent: usize,
    word_type: &str,
    unit_bytes: u32,
    endian: &ByteEndian,
    variable_length: bool,
    type_maps: &HashMap<String, String>,
) {
    let pad = "    ".repeat(indent);
    match node {
        DecodeNode::Leaf { instruction_index } => {
            let instr = &def.instructions[*instruction_index];
            if let Some(guard) = leaf_guard(instr, word_type, unit_bytes, endian) {
                writeln!(out, "{}if ({}) {{", pad, guard).unwrap();
                emit_return_instruction(
                    out,
                    instr,
                    indent + 1,
                    word_type,
                    unit_bytes,
                    endian,
                    variable_length,
                    type_maps,
                );
                writeln!(out, "{}}} else {{", pad).unwrap();
                writeln!(out, "{}    return std::nullopt;", pad).unwrap();
                writeln!(out, "{}}}", pad).unwrap();
            } else {
                emit_return_instruction(
                    out,
                    instr,
                    indent,
                    word_type,
                    unit_bytes,
                    endian,
                    variable_length,
                    type_maps,
                );
            }
        }
        DecodeNode::PriorityLeaves { candidates } => {
            for (i, &idx) in candidates.iter().enumerate() {
                let instr = &def.instructions[idx];
                let guard = leaf_guard(instr, word_type, unit_bytes, endian);
                if i == 0 {
                    if let Some(g) = guard {
                        writeln!(out, "{}if ({}) {{", pad, g).unwrap();
                        emit_return_instruction(
                            out,
                            instr,
                            indent + 1,
                            word_type,
                            unit_bytes,
                            endian,
                            variable_length,
                            type_maps,
                        );
                    } else {
                        emit_return_instruction(
                            out,
                            instr,
                            indent,
                            word_type,
                            unit_bytes,
                            endian,
                            variable_length,
                            type_maps,
                        );
                        break;
                    }
                } else if i == candidates.len() - 1 {
                    if let Some(g) = guard {
                        writeln!(out, "{}}} else if ({}) {{", pad, g).unwrap();
                        emit_return_instruction(
                            out,
                            instr,
                            indent + 1,
                            word_type,
                            unit_bytes,
                            endian,
                            variable_length,
                            type_maps,
                        );
                        writeln!(out, "{}}} else {{", pad).unwrap();
                        writeln!(out, "{}    return std::nullopt;", pad).unwrap();
                        writeln!(out, "{}}}", pad).unwrap();
                    } else {
                        writeln!(out, "{}}} else {{", pad).unwrap();
                        emit_return_instruction(
                            out,
                            instr,
                            indent + 1,
                            word_type,
                            unit_bytes,
                            endian,
                            variable_length,
                            type_maps,
                        );
                        writeln!(out, "{}}}", pad).unwrap();
                    }
                } else {
                    let g = guard.unwrap_or_else(|| "true".to_string());
                    writeln!(out, "{}}} else if ({}) {{", pad, g).unwrap();
                    emit_return_instruction(
                        out,
                        instr,
                        indent + 1,
                        word_type,
                        unit_bytes,
                        endian,
                        variable_length,
                        type_maps,
                    );
                }
            }
        }
        DecodeNode::Fail => {
            writeln!(out, "{}return std::nullopt;", pad).unwrap();
        }
        DecodeNode::Branch {
            range,
            arms,
            default,
        } => {
            let ext = extract_expr("opcode", &[*range], word_type, unit_bytes, endian);
            writeln!(out, "{}switch ({}) {{", pad, ext).unwrap();
            for (value, child) in arms {
                writeln!(out, "{}case {:#x}: {{", pad, value).unwrap();
                emit_tree_cpp(
                    out,
                    child,
                    def,
                    indent + 1,
                    word_type,
                    unit_bytes,
                    endian,
                    variable_length,
                    type_maps,
                );
                writeln!(out, "{}    break;", pad).unwrap();
                writeln!(out, "{}}}", pad).unwrap();
            }
            writeln!(out, "{}default: {{", pad).unwrap();
            emit_tree_cpp(
                out,
                default,
                def,
                indent + 1,
                word_type,
                unit_bytes,
                endian,
                variable_length,
                type_maps,
            );
            writeln!(out, "{}    break;", pad).unwrap();
            writeln!(out, "{}}}", pad).unwrap();
            writeln!(out, "{}}}", pad).unwrap();
        }
    }
}

/// Emit code to return a decoded Instruction.
fn emit_return_instruction(
    out: &mut String,
    instr: &ValidatedInstruction,
    indent: usize,
    word_type: &str,
    unit_bytes: u32,
    endian: &ByteEndian,
    variable_length: bool,
    type_maps: &HashMap<String, String>,
) {
    let pad = "    ".repeat(indent);
    let unit_count = instr.unit_count();
    let bytes_consumed = unit_count * unit_bytes;
    let variant = to_pascal_case(&instr.name);

    if variable_length && unit_count > 1 {
        writeln!(
            out,
            "{}if (len < {}) return std::nullopt;",
            pad, bytes_consumed
        )
        .unwrap();
    }

    if instr.resolved_fields.is_empty() {
        writeln!(
            out,
            "{}return Instruction {{ Opcode::{}, {} }};",
            pad, variant, bytes_consumed
        )
        .unwrap();
    } else {
        writeln!(out, "{}{{", pad).unwrap();
        writeln!(out, "{}    Instruction insn{{}};", pad).unwrap();
        writeln!(out, "{}    insn.opcode = Opcode::{};", pad, variant).unwrap();
        writeln!(out, "{}    insn.size = {};", pad, bytes_consumed).unwrap();
        for field in &instr.resolved_fields {
            let ext = extract_expr("opcode", &field.ranges, word_type, unit_bytes, endian);
            let expr = apply_transforms(&ext, &field.resolved_type, type_maps);
            writeln!(
                out,
                "{}    insn.{}.{} = {};",
                pad, instr.name, field.name, expr
            )
            .unwrap();
        }
        writeln!(out, "{}    return insn;", pad).unwrap();
        writeln!(out, "{}}}", pad).unwrap();
    }
}

/// Emit the format/disassemble function using std::format.
fn emit_format_function(
    out: &mut String,
    def: &ValidatedDef,
    _type_maps: &HashMap<String, String>,
    _opts: &CppOptions,
) {
    writeln!(out, "inline std::string format(const Instruction& insn) {{").unwrap();
    writeln!(out, "    switch (insn.opcode) {{").unwrap();

    for instr in &def.instructions {
        let variant = to_pascal_case(&instr.name);
        writeln!(out, "    case Opcode::{}: {{", variant).unwrap();

        if instr.format_lines.is_empty() {
            writeln!(out, "        return \"{}\";", instr.name).unwrap();
        } else {
            emit_format_lines_cpp(out, instr, 2);
        }

        writeln!(out, "    }}").unwrap();
    }

    writeln!(out, "    default: return \"???\";").unwrap();
    writeln!(out, "    }}").unwrap();
    writeln!(out, "}}").unwrap();
    writeln!(out).unwrap();
}

/// Emit format lines for a single instruction using std::format.
fn emit_format_lines_cpp(out: &mut String, instr: &ValidatedInstruction, indent: usize) {
    let pad = "    ".repeat(indent);

    if instr.format_lines.len() == 1 && instr.format_lines[0].guard.is_none() {
        let fl = &instr.format_lines[0];
        emit_std_format_call(out, &fl.pieces, instr, &pad);
        return;
    }

    for (i, fl) in instr.format_lines.iter().enumerate() {
        if let Some(guard) = &fl.guard {
            let guard_code = guard_to_cpp(guard, instr);
            if i == 0 {
                writeln!(out, "{}if ({}) {{", pad, guard_code).unwrap();
            } else {
                writeln!(out, "{}}} else if ({}) {{", pad, guard_code).unwrap();
            }
            emit_std_format_call(out, &fl.pieces, instr, &format!("{}    ", pad));
        } else {
            if i > 0 {
                writeln!(out, "{}}} else {{", pad).unwrap();
            }
            emit_std_format_call(out, &fl.pieces, instr, &format!("{}    ", pad));
        }
    }

    if instr.format_lines.len() > 1
        || instr
            .format_lines
            .first()
            .map_or(false, |fl| fl.guard.is_some())
    {
        writeln!(out, "{}}}", pad).unwrap();
    }
}

/// Emit a return std::format(...) call for format pieces.
/// Fields are passed as arguments; user types participate via std::formatter specializations.
fn emit_std_format_call(
    out: &mut String,
    pieces: &[FormatPiece],
    instr: &ValidatedInstruction,
    pad: &str,
) {
    let mut fmt_str = String::new();
    let mut args: Vec<String> = Vec::new();

    for piece in pieces {
        match piece {
            FormatPiece::Literal(lit) => {
                // Escape { and } for std::format
                for ch in lit.chars() {
                    match ch {
                        '{' => fmt_str.push_str("{{"),
                        '}' => fmt_str.push_str("}}"),
                        _ => fmt_str.push(ch),
                    }
                }
            }
            FormatPiece::FieldRef { expr, spec } => {
                let cpp_expr = expr_to_cpp(expr, instr);
                // Build std::format placeholder
                if let Some(spec) = spec {
                    fmt_str.push_str(&format!("{{:{}}}", translate_std_format_spec(spec)));
                } else {
                    fmt_str.push_str("{}");
                }
                args.push(cpp_expr);
            }
        }
    }

    if args.is_empty() {
        writeln!(out, "{}return \"{}\";", pad, fmt_str).unwrap();
    } else {
        writeln!(
            out,
            "{}return std::format(\"{}\", {});",
            pad,
            fmt_str,
            args.join(", ")
        )
        .unwrap();
    }
}

/// Translate chipi/Rust format specs to std::format specs.
/// Most are identical since std::format uses Python-style specs.
fn translate_std_format_spec(spec: &str) -> String {
    // chipi specs like "04x", "#06x", "#x" are already valid std::format specs
    spec.to_string()
}

/// Convert a FormatExpr to a C++ expression string.
fn expr_to_cpp(expr: &FormatExpr, instr: &ValidatedInstruction) -> String {
    match expr {
        FormatExpr::Field(name) => {
            format!("insn.{}.{}", instr.name, name)
        }
        FormatExpr::Ternary {
            field,
            if_nonzero,
            if_zero,
        } => {
            let else_val = if_zero.as_deref().unwrap_or("");
            format!(
                "(insn.{}.{} ? \"{}\" : \"{}\")",
                instr.name, field, if_nonzero, else_val
            )
        }
        FormatExpr::Arithmetic { left, op, right } => {
            let l = expr_to_cpp(left, instr);
            let r = expr_to_cpp(right, instr);
            let op_str = match op {
                ArithOp::Add => "+",
                ArithOp::Sub => "-",
                ArithOp::Mul => "*",
                ArithOp::Div => "/",
                ArithOp::Mod => "%",
            };
            format!("({} {} {})", l, op_str, r)
        }
        FormatExpr::IntLiteral(val) => format!("{}", val),
        FormatExpr::MapCall { map_name, args } => {
            let arg_strs: Vec<String> = args.iter().map(|a| expr_to_cpp(a, instr)).collect();
            format!("{}({})", map_name, arg_strs.join(", "))
        }
        FormatExpr::BuiltinCall { func, args } => {
            let arg_strs: Vec<String> = args.iter().map(|a| expr_to_cpp(a, instr)).collect();
            match func {
                BuiltinFunc::RotateRight => {
                    format!(
                        "((static_cast<uint32_t>({}) >> {}) | (static_cast<uint32_t>({}) << (32 - {})))",
                        arg_strs.first().map(|s| s.as_str()).unwrap_or("0"),
                        arg_strs.get(1).map(|s| s.as_str()).unwrap_or("0"),
                        arg_strs.first().map(|s| s.as_str()).unwrap_or("0"),
                        arg_strs.get(1).map(|s| s.as_str()).unwrap_or("0"),
                    )
                }
                BuiltinFunc::RotateLeft => {
                    format!(
                        "((static_cast<uint32_t>({}) << {}) | (static_cast<uint32_t>({}) >> (32 - {})))",
                        arg_strs.first().map(|s| s.as_str()).unwrap_or("0"),
                        arg_strs.get(1).map(|s| s.as_str()).unwrap_or("0"),
                        arg_strs.first().map(|s| s.as_str()).unwrap_or("0"),
                        arg_strs.get(1).map(|s| s.as_str()).unwrap_or("0"),
                    )
                }
            }
        }
        FormatExpr::SubDecoderAccess { field, fragment } => {
            format!("insn.{}.{}.{}", instr.name, field, fragment)
        }
    }
}

/// Convert a guard condition to C++ code.
fn guard_to_cpp(guard: &Guard, instr: &ValidatedInstruction) -> String {
    let conditions: Vec<String> = guard
        .conditions
        .iter()
        .map(|cond| {
            let left = guard_operand_cpp(&cond.left, instr);
            let right = guard_operand_cpp(&cond.right, instr);
            let op = match cond.op {
                CompareOp::Eq => "==",
                CompareOp::Ne => "!=",
                CompareOp::Lt => "<",
                CompareOp::Le => "<=",
                CompareOp::Gt => ">",
                CompareOp::Ge => ">=",
            };
            format!("{} {} {}", left, op, right)
        })
        .collect();
    conditions.join(" && ")
}

fn guard_operand_cpp(operand: &GuardOperand, instr: &ValidatedInstruction) -> String {
    match operand {
        GuardOperand::Field(name) => format!("insn.{}.{}", instr.name, name),
        GuardOperand::Literal(val) => format!("{}", val),
        GuardOperand::Expr { left, op, right } => {
            let l = guard_operand_cpp(left, instr);
            let r = guard_operand_cpp(right, instr);
            let op_str = match op {
                ArithOp::Add => "+",
                ArithOp::Sub => "-",
                ArithOp::Mul => "*",
                ArithOp::Div => "/",
                ArithOp::Mod => "%",
            };
            format!("({} {} {})", l, op_str, r)
        }
    }
}