polydat_nodes/

string.rs

// Copyright 2024-2026 Jonathan Shook
// SPDX-License-Identifier: Apache-2.0

//! String generation and transformation nodes.

// =================================================================
// Combinations: mixed-radix character set mapping
// =================================================================

/// Map a u64 to a formatted string via mixed-radix indexing into
/// character sets.
///
/// Signature: `combinations(input: u64, pattern: &str) -> (String)`
///
/// The pattern is a semicolon-delimited list of character set specs.
/// Each spec is a character range (`A-Z`), literal characters, or
/// both. A single literal character (like `-`) is emitted as-is
/// without consuming a radix digit.
///
/// Use for generating structured identifiers with fixed character
/// classes per position. Examples: phone numbers
/// (`"0-9;0-9;0-9;-;0-9;0-9;0-9;-;0-9;0-9;0-9;0-9"` yields
/// `"372-841-9205"`), license plates (`"A-Z;A-Z;A-Z;-;0-9;0-9;0-9"`),
/// or hex tokens (`"0-9a-f;0-9a-f;0-9a-f;0-9a-f"`). Input wraps at
/// `cardinality()`, so every value in the cycle space maps to a valid
/// string.
///
/// JIT level: P1 (String output; no compiled_u64 path).
/// Derived state for `combinations`. Computed once per node
/// instance via `from_pattern`; the macro stores the instance in a
/// struct field and hands the eval body a `&ParsedCombinations`
/// borrow each call.
pub struct ParsedCombinations {
    /// The segments, in output order.
    pub segments: Vec<Segment>,
    /// The product of the charset sizes: the distinct combinations.
    pub modulus: u64,
}

impl polydat::derive_support::PolydatSetup for ParsedCombinations {}

/// One piece of a combinations pattern.
pub enum Segment {
    /// Variable: select one char from the charset based on a radix digit.
    Charset(Vec<char>),
    /// Fixed: always emit this string (e.g., a literal separator).
    Literal(String),
}

impl ParsedCombinations {
    /// Single-call setup. The `#[polydat_node]` macro invokes
    /// this exactly once in the generated `Combinations::new()`;
    /// no other call path exists.
    pub fn from_pattern(pattern: &str) -> Self {
        let mut segments = Vec::new();
        let mut modulus: u64 = 1;
        for spec in pattern.split(';') {
            let chars = parse_charset(spec);
            if chars.len() == 1 && !spec.contains('-') {
                segments.push(Segment::Literal(chars[0].to_string()));
            } else if chars.is_empty() {
                segments.push(Segment::Literal(spec.to_string()));
            } else {
                modulus = modulus.saturating_mul(chars.len() as u64);
                segments.push(Segment::Charset(chars));
            }
        }
        Self { segments, modulus }
    }
}

/// Map a u64 input to a deterministic string by interpreting
/// it as a multi-positional choice over the pattern's charsets.
#[polydat::polydat_node(category = String)]
fn combinations(
    input: u64,
    pattern: polydat::derive_support::Const<&str>,
    #[poly_const(ParsedCombinations::from_pattern, from = pattern)] parsed: &ParsedCombinations,
) -> String {
    let mut remainder = if parsed.modulus > 0 {
        input % parsed.modulus
    } else {
        input
    };
    let mut result = String::with_capacity(parsed.segments.len() * 2);
    for seg in &parsed.segments {
        match seg {
            Segment::Literal(s) => result.push_str(s),
            Segment::Charset(chars) => {
                let radix = chars.len() as u64;
                if radix > 0 {
                    let idx = (remainder % radix) as usize;
                    result.push(chars[idx]);
                    remainder /= radix;
                }
            }
        }
    }
    result
}

impl Combinations {
    /// The total number of unique combinations before wrapping.
    pub fn cardinality(&self) -> u64 {
        self.parsed.modulus
    }
}

/// Parse a charset spec like "A-Z", "0-9", "a-z0-9", "A-Za-z0-9 _|/"
fn parse_charset(spec: &str) -> Vec<char> {
    let mut chars = Vec::new();
    let spec_chars: Vec<char> = spec.chars().collect();
    let mut i = 0;
    while i < spec_chars.len() {
        if i + 2 < spec_chars.len() && spec_chars[i + 1] == '-' {
            // Range: A-Z, 0-9, etc.
            let start = spec_chars[i];
            let end = spec_chars[i + 2];
            for c in start..=end {
                chars.push(c);
            }
            i += 3;
        } else {
            chars.push(spec_chars[i]);
            i += 1;
        }
    }
    chars
}

// =================================================================
// NumberToWords: spell out numbers in English
// =================================================================

/// Convert a u64 to its English word representation.
///
/// Signature: `number_to_words(input: u64) -> (String)`
///
/// Examples: 0 produces "zero", 42 produces "forty-two", 1000
/// produces "one thousand". Supports the full u64 range up through
/// quintillions.
///
/// JIT level: P1 (String output; no compiled_u64 path).
#[polydat::polydat_node(category = String)]
fn number_to_words(input: u64) -> String {
    u64_to_words(input)
}

const ONES: [&str; 20] = [
    "zero",
    "one",
    "two",
    "three",
    "four",
    "five",
    "six",
    "seven",
    "eight",
    "nine",
    "ten",
    "eleven",
    "twelve",
    "thirteen",
    "fourteen",
    "fifteen",
    "sixteen",
    "seventeen",
    "eighteen",
    "nineteen",
];

const TENS: [&str; 10] = [
    "", "", "twenty", "thirty", "forty", "fifty", "sixty", "seventy", "eighty", "ninety",
];

const SCALES: [&str; 7] = [
    "",
    "thousand",
    "million",
    "billion",
    "trillion",
    "quadrillion",
    "quintillion",
];

fn u64_to_words(n: u64) -> String {
    if n < 20 {
        return ONES[n as usize].to_string();
    }

    let mut buf = String::with_capacity(64);
    let mut chunks = [0u32; 7];
    let mut num_chunks = 0;
    let mut remaining = n;

    while remaining > 0 {
        chunks[num_chunks] = (remaining % 1000) as u32;
        num_chunks += 1;
        remaining /= 1000;
    }

    let mut first = true;
    for i in (0..num_chunks).rev() {
        let chunk = chunks[i];
        if chunk > 0 {
            if !first {
                buf.push(' ');
            }
            first = false;
            append_chunk_to_words(&mut buf, chunk);
            if i > 0 && i < SCALES.len() {
                buf.push(' ');
                buf.push_str(SCALES[i]);
            }
        }
    }

    buf
}

fn append_chunk_to_words(buf: &mut String, n: u32) {
    let hundreds = n / 100;
    let remainder = n % 100;

    let mut has_hundreds = false;
    if hundreds > 0 {
        buf.push_str(ONES[hundreds as usize]);
        buf.push_str(" hundred");
        has_hundreds = true;
    }

    if remainder >= 20 {
        if has_hundreds {
            buf.push(' ');
        }
        let tens = remainder / 10;
        let ones = remainder % 10;
        buf.push_str(TENS[tens as usize]);
        if ones > 0 {
            buf.push('-');
            buf.push_str(ONES[ones as usize]);
        }
    } else if remainder > 0 {
        if has_hundreds {
            buf.push(' ');
        }
        buf.push_str(ONES[remainder as usize]);
    }
}

// =================================================================
// HashedUuid: deterministic UUID v4 from a u64 seed
// =================================================================

/// Generate a deterministic UUID v4 string from a u64 seed.
/// Same seed always produces the same UUID; the hash output
/// fills the 128 UUID bits with version (4) and variant
/// (RFC 4122) bits set per spec.
///
/// Signature: `hashed_uuid(input: u64) -> (String)`
#[polydat::polydat_node(category = String)]
fn hashed_uuid(input: u64) -> String {
    // Two hashes fill 128 bits.
    let h1 = xxhash_rust::xxh3::xxh3_64(&input.to_le_bytes());
    let h2 = xxhash_rust::xxh3::xxh3_64(&h1.to_le_bytes());
    let mut bytes = [0u8; 16];
    bytes[..8].copy_from_slice(&h1.to_le_bytes());
    bytes[8..].copy_from_slice(&h2.to_le_bytes());
    // Version 4 (bits 12-15 of byte 6).
    bytes[6] = (bytes[6] & 0x0F) | 0x40;
    // Variant RFC 4122 (bits 6-7 of byte 8).
    bytes[8] = (bytes[8] & 0x3F) | 0x80;
    format!(
        "{:02x}{:02x}{:02x}{:02x}-{:02x}{:02x}-{:02x}{:02x}-{:02x}{:02x}-{:02x}{:02x}{:02x}{:02x}{:02x}{:02x}",
        bytes[0],
        bytes[1],
        bytes[2],
        bytes[3],
        bytes[4],
        bytes[5],
        bytes[6],
        bytes[7],
        bytes[8],
        bytes[9],
        bytes[10],
        bytes[11],
        bytes[12],
        bytes[13],
        bytes[14],
        bytes[15],
    )
}

// =================================================================
// CharBuf: deterministic string from seed + charset + length
// =================================================================

/// Generate a deterministic string of a given length from a seed
/// and character set.
///
/// The seed is hashed and used to index into the charset repeatedly.
/// Same seed + charset + length always produces the same string.
///
/// Signature: `char_buf(seed: u64, charset: &str, length: u64) -> (String)`
/// Expand a charset spec like "A-Za-z0-9" into a Vec<char>.
/// Setup helper for `char_buf`.
fn expand_charset(charset: &str) -> Vec<char> {
    if charset.is_empty() {
        return ('a'..='z').collect();
    }
    let mut result = Vec::new();
    let chars_vec: Vec<char> = charset.chars().collect();
    let mut i = 0;
    while i < chars_vec.len() {
        if i + 2 < chars_vec.len() && chars_vec[i + 1] == '-' {
            for c in chars_vec[i]..=chars_vec[i + 2] {
                result.push(c);
            }
            i += 3;
        } else {
            result.push(chars_vec[i]);
            i += 1;
        }
    }
    if result.is_empty() {
        ('a'..='z').collect()
    } else {
        result
    }
}

/// Generate a deterministic string of a given length from a
/// seed and character set.
#[polydat::polydat_node(category = String)]
fn char_buf(
    seed: u64,
    charset: polydat::derive_support::Const<&str>,
    length: u64,
    #[poly_const(expand_charset, from = charset)] chars: &Vec<char>,
) -> String {
    let n = chars.len();
    let len = length as usize;
    if n == 0 || len == 0 {
        return String::new();
    }
    let mut result = String::with_capacity(len);
    let mut h = seed;
    for _ in 0..len {
        h = xxhash_rust::xxh3::xxh3_64(&h.to_le_bytes());
        result.push(chars[(h as usize) % n]);
    }
    result
}

// =================================================================
// FileLineAt: index into a file of lines at cycle time
// =================================================================

/// Read `filename` and split it into lines. Panics on file
/// I/O failure; the macro's build-closure `catch_unwind`
/// surfaces it as a clean compile error.
fn read_file_lines(filename: &str) -> Vec<String> {
    let content = std::fs::read_to_string(filename)
        .unwrap_or_else(|e| panic!("failed to read file '{filename}': {e}"));
    let lines: Vec<String> = content.lines().map(|l| l.to_string()).collect();
    if lines.is_empty() {
        panic!("file '{filename}' has no lines");
    }
    lines
}

/// Cycle-time line lookup over a pre-loaded text file. `filename`
/// is read at construction time via `#[poly_const]`; the cycle
/// input selects a line modulo the total count.
#[polydat::polydat_node(category = String)]
fn file_line_at(
    index: u64,
    filename: polydat::derive_support::Const<&str>,
    #[poly_const(read_file_lines, from = filename)] lines: &Vec<String>,
) -> String {
    let _ = filename;
    let idx = index as usize;
    lines[idx % lines.len()].clone()
}

// =================================================================
// StrConcat: variadic string concatenation
// =================================================================

/// Concatenate N wire inputs into a single Str output.
///
/// Each input is rendered to its display form: Str passes through,
/// numerics format as decimal, Bool as `true`/`false`, Json via
/// `to_string`. Mixed-type inputs are accepted — the assembler skips
/// type checking for str_concat (like printf), so any upstream wire
/// type composes.
///
/// Used by the DSL desugar of `+` between Str-typed operands; also
/// callable directly as `str_concat(a, b, c, ...)`.
///
/// Signature: `str_concat(in_0, in_1, ...) -> (String)`
/// Concatenate N values, rendering each as its display form.
/// Variadic over `&[Value]`: the body stringifies
/// per element so mixed-type inputs (Str + U64 + Bool, etc.)
/// produce a single concatenated string; this matches the
/// DSL's lowering of `+` between Str-typed operands.
#[polydat::polydat_node(category = String)]
fn str_concat(parts: &[polydat::ast::Value]) -> String {
    use polydat::ast::Value;
    let mut out = String::new();
    for v in parts {
        match v {
            Value::Str(s) => out.push_str(s),
            Value::U64(n) => out.push_str(&n.to_string()),
            Value::F64(n) => out.push_str(&n.to_string()),
            Value::Bool(b) => out.push_str(&b.to_string()),
            Value::Json(j) => out.push_str(&j.to_string()),
            Value::Bytes(b) => out.push_str(&String::from_utf8_lossy(b)),
            // Everything else, extension values included, renders as it
            // does in interpolation and `printf`: its display form.
            other => out.push_str(&other.to_display_string()),
        }
    }
    out
}

// =================================================================
// StrLower / StrUpper: Unicode case-folding helpers
// =================================================================

/// Fold a string to lowercase (`str.to_lowercase()` semantics).
///
/// Signature: `str_lower(input: Str) -> (Str)`
///
/// Takes `String` (not `&str`) so `FromValue<String>` honors the legacy
/// "stringify any input via `to_display_string`" behavior;
/// switching to `&str` would tighten this to require Str
/// inputs only, which the type-checker doesn't yet enforce
/// (deferred to SRD-79's type-driven resolution).
#[polydat::polydat_node(category = String)]
fn str_lower(input: String) -> String {
    input.to_lowercase()
}

/// Fold a string to uppercase (`str.to_uppercase()` semantics).
///
/// Signature: `str_upper(input: Str) -> (Str)`
#[polydat::polydat_node(category = String)]
fn str_upper(input: String) -> String {
    input.to_uppercase()
}

#[cfg(test)]
mod tests {
    use super::*;
    use polydat::ast::{PolydatNode, Value};

    // --- Combinations tests ---

    #[test]
    fn combinations_digits() {
        let node = Combinations::new("0-9;0-9;0-9".to_string());
        let mut out = [Value::None];
        node.eval(&[Value::U64(123)], &mut out);
        let s = out[0].as_str();
        assert_eq!(s.len(), 3);
        assert!(s.chars().all(|c| c.is_ascii_digit()));
    }

    #[test]
    fn combinations_with_separator() {
        let node = Combinations::new("0-9;0-9;0-9;-;0-9;0-9;0-9".to_string());
        let mut out = [Value::None];
        node.eval(&[Value::U64(0)], &mut out);
        let s = out[0].as_str();
        assert_eq!(s.len(), 7); // 3 digits + dash + 3 digits
        assert_eq!(&s[3..4], "-");
    }

    #[test]
    fn combinations_alpha() {
        let node = Combinations::new("A-Z;A-Z;A-Z".to_string());
        let mut out = [Value::None];
        node.eval(&[Value::U64(0)], &mut out);
        assert_eq!(out[0].as_str(), "AAA");
        node.eval(&[Value::U64(1)], &mut out);
        assert_eq!(out[0].as_str(), "BAA");
    }

    #[test]
    fn combinations_cardinality() {
        let node = Combinations::new("0-9;0-9;-;A-Z".to_string());
        // 10 * 10 * 26 = 2600 (separator doesn't count)
        assert_eq!(node.cardinality(), 2600);
    }

    #[test]
    fn combinations_deterministic() {
        let node = Combinations::new("A-Z;0-9".to_string());
        let mut out1 = [Value::None];
        let mut out2 = [Value::None];
        node.eval(&[Value::U64(42)], &mut out1);
        node.eval(&[Value::U64(42)], &mut out2);
        assert_eq!(out1[0].as_str(), out2[0].as_str());
    }

    #[test]
    fn combinations_wraps() {
        let node = Combinations::new("0-9".to_string());
        let mut out = [Value::None];
        node.eval(&[Value::U64(0)], &mut out);
        let a = out[0].as_str().to_string();
        node.eval(&[Value::U64(10)], &mut out);
        assert_eq!(out[0].as_str(), &a, "should wrap at cardinality");
    }

    // --- NumberToWords tests ---

    #[test]
    fn number_to_words_zero() {
        assert_eq!(u64_to_words(0), "zero");
    }

    #[test]
    fn number_to_words_teens() {
        assert_eq!(u64_to_words(1), "one");
        assert_eq!(u64_to_words(11), "eleven");
        assert_eq!(u64_to_words(19), "nineteen");
    }

    #[test]
    fn number_to_words_tens() {
        assert_eq!(u64_to_words(20), "twenty");
        assert_eq!(u64_to_words(42), "forty-two");
        assert_eq!(u64_to_words(99), "ninety-nine");
    }

    #[test]
    fn number_to_words_hundreds() {
        assert_eq!(u64_to_words(100), "one hundred");
        assert_eq!(u64_to_words(123), "one hundred twenty-three");
        assert_eq!(u64_to_words(500), "five hundred");
    }

    #[test]
    fn number_to_words_thousands() {
        assert_eq!(u64_to_words(1000), "one thousand");
        assert_eq!(u64_to_words(1001), "one thousand one");
        assert_eq!(
            u64_to_words(12345),
            "twelve thousand three hundred forty-five"
        );
    }

    #[test]
    fn number_to_words_millions() {
        assert_eq!(u64_to_words(1_000_000), "one million");
        assert_eq!(
            u64_to_words(1_234_567),
            "one million two hundred thirty-four thousand five hundred sixty-seven"
        );
    }

    #[test]
    fn number_to_words_large() {
        let s = u64_to_words(1_000_000_000_000);
        assert!(s.starts_with("one trillion"), "got: {s}");
    }

    #[test]
    fn number_to_words_node() {
        let node = NumberToWords::new();
        let mut out = [Value::None];
        node.eval(&[Value::U64(42)], &mut out);
        assert_eq!(out[0].as_str(), "forty-two");
    }

    // --- StrConcat tests ---

    #[test]
    fn str_concat_basic() {
        let node = StrConcat::new(2);
        let mut out = [Value::None];
        node.eval(
            &[Value::Str("hello ".into()), Value::Str("world".into())],
            &mut out,
        );
        assert_eq!(out[0].as_str(), "hello world");
    }

    #[test]
    fn str_concat_renders_extension_values_by_display() {
        #[derive(Debug, Clone)]
        struct Tag(u64);
        impl polydat::ast::ReflectedValue for Tag {
            fn type_name(&self) -> &str {
                "Tag"
            }
            fn display(&self) -> String {
                format!("tag#{}", self.0)
            }
            fn clone_reflected(&self) -> Box<dyn polydat::ast::ReflectedValue> {
                Box::new(self.clone())
            }
            fn as_any(&self) -> &dyn std::any::Any {
                self
            }
        }
        let node = StrConcat::new(2);
        let mut out = [Value::None];
        node.eval(
            &[Value::Str("x".into()), Value::Ext(Box::new(Tag(7)))],
            &mut out,
        );
        assert_eq!(out[0].as_str(), "xtag#7");
    }

    #[test]
    fn str_concat_mixed_types() {
        let node = StrConcat::new(4);
        let mut out = [Value::None];
        node.eval(
            &[
                Value::Str("id=".into()),
                Value::U64(42),
                Value::Str(" v=".into()),
                Value::F64(3.14),
            ],
            &mut out,
        );
        assert_eq!(out[0].as_str(), "id=42 v=3.14");
    }

    #[test]
    fn str_concat_empty() {
        let node = StrConcat::new(0);
        let mut out = [Value::None];
        node.eval(&[], &mut out);
        assert_eq!(out[0].as_str(), "");
    }

    #[test]
    fn str_lower_ascii_and_unicode() {
        let node = StrLower::new();
        let mut out = [Value::None];
        node.eval(&[Value::Str("OTHER_M8".into())], &mut out);
        assert_eq!(out[0].as_str(), "other_m8");
        // Unicode folding (Rust's str::to_lowercase is full Unicode).
        node.eval(&[Value::Str("ÄPFEL".into())], &mut out);
        assert_eq!(out[0].as_str(), "äpfel");
    }

    #[test]
    fn str_lower_idempotent_on_already_lowercase() {
        let node = StrLower::new();
        let mut out = [Value::None];
        node.eval(&[Value::Str("fknn_oat_other".into())], &mut out);
        assert_eq!(out[0].as_str(), "fknn_oat_other");
    }

    #[test]
    fn str_upper_ascii_and_unicode() {
        let node = StrUpper::new();
        let mut out = [Value::None];
        node.eval(&[Value::Str("other_m8".into())], &mut out);
        assert_eq!(out[0].as_str(), "OTHER_M8");
        node.eval(&[Value::Str("äpfel".into())], &mut out);
        assert_eq!(out[0].as_str(), "ÄPFEL");
    }

    // `str_lower_accepts_non_string_via_display` retired: SRD-80b
    // Wire trait dispatch panics on shape mismatch instead of
    // silently display-coercing. Workload-level support for
    // chained `str_lower(format_u64(...))` flows through assembler-
    // inserted Str adapters (e.g. polyfill U64ToStr), not through
    // a lying Wire impl. Tests that want to exercise the coercion
    // path should construct a Str via `format_u64` upstream and
    // feed that into str_lower's eval.
}