powerliners 0.1.1

// vim:fileencoding=utf-8:noet
//! Port of `powerline/lib/unicode.py`.
//!
//! Upstream is a 283-line Python 2/3 unicode-compat layer. The vast
//! majority of it handles Py2 `unicode`/`str` distinctions, codec
//! error fallbacks, and `__builtin__.unichr` polyfills — all of which
//! are no-ops in Rust where every `String` is UTF-8 by construction
//! and `char` is a 4-byte Unicode scalar.
//!
//! Ported surface:
//!   - `u()` / `safe_unicode*` — string coercion
//!   - `tointiter()` — byte iter helper
//!   - `powerline_decode_error()` — `<XX>` hex error formatter
//!   - `register_strwidth_error()` — encode-error name generator
//!   - `out_u()` — bytes → string via preferred output encoding
//!   - `FailedUnicode` — marker newtype
//!   - `surrogate_pair_to_character()` — high/low surrogate join
//!   - `string()` — identity for `&str`
//!   - `strwidth_ucs_4()` / `strwidth_ucs_2()` — display-width helpers

// from __future__ import (unicode_literals, division, absolute_import, print_function)  // py:2
// import sys                                       // py:4
// import codecs                                    // py:5
// from unicodedata import east_asian_width, combining  // py:7
// from powerline.lib.encoding import get_preferred_output_encoding  // py:9

use std::collections::HashMap;
use std::sync::atomic::{AtomicUsize, Ordering};

/// Port of `u()` from `powerline/lib/unicode.py:35`.
///
/// Python:
/// ```python
/// def u(s):
///     '''Return unicode instance assuming UTF-8 encoded string.'''
///     if type(s) is unicode:
///         return s
///     else:
///         return unicode(s, 'utf-8')
/// ```
///
/// In Python the input could be `unicode` (Py2) / `str` (Py3) or
/// `bytes`. Rust callers pass `&str` (already valid UTF-8) or
/// `&[u8]` (byte slice). Both shapes are provided.
pub fn u(s: &str) -> String {
    // py:12  try:
    // py:13  from __builtin__ import unicode
    // py:14  except ImportError:
    // py:15  unicode = str
    // py:18  try:
    // py:19  from __builtin__ import unichr
    // py:20  except ImportError:
    // py:21  unichr = chr
    // py:24  if sys.maxunicode < 0x10FFFF:
    // py:25  _unichr = unichr
    // py:27  def unichr(ch):
    // py:28  if ch <= sys.maxunicode:
    // py:29  return _unichr(ch)
    // py:30  else:
    // py:31  ch -= 0x10000
    // py:32  return _unichr((ch >> 10) + 0xD800) + _unichr((ch & ((1 << 10) - 1)) + 0xDC00)
    // py:35  def u(s):
    // py:36  '''Return unicode instance assuming UTF-8 encoded string.
    // py:37  '''
    // py:38  if type(s) is unicode:
    // py:39  return s
    // py:40  else:
    // py:41  return unicode(s, 'utf-8')
    s.to_string()
}

/// Port of `tointiter()` from
/// `powerline/lib/unicode.py:44-53`.
///
/// Python: convert a byte string to the sequence of integers.
/// Py2 uses `ord(c) for c in s`, Py3 uses `iter(s)`. Rust's `&[u8]`
/// already iterates as `u8`, so the impl is just `.iter().copied()`.
pub fn tointiter(s: &[u8]) -> impl Iterator<Item = u8> + '_ {
    // py:44  if sys.version_info < (3,):
    // py:45  def tointiter(s):
    // py:46  '''Convert a byte string to the sequence of integers
    // py:47  '''
    // py:48  return (ord(c) for c in s)
    // py:49  else:
    // py:50  def tointiter(s):
    // py:51  '''Convert a byte string to the sequence of integers
    // py:52  '''
    // py:53  return iter(s)
    s.iter().copied()
}

/// Port of `powerline_decode_error()` from
/// `powerline/lib/unicode.py:56-62`.
///
/// Replaces an invalid byte range with `<XX>` hex notation. Python
/// receives a `UnicodeDecodeError` exception object; Rust port takes
/// the raw byte slice that triggered the error since Rust decoders
/// (e.g. `from_utf8_lossy`) don't surface the same exception shape.
/// Returns the replacement string plus the consumed byte count, the
/// same `(replacement, end)` tuple Python returns at py:62.
pub fn powerline_decode_error(bytes: &[u8]) -> (String, usize) {
    // py:56  def powerline_decode_error(e):
    // py:57  if not isinstance(e, UnicodeDecodeError):
    // py:58  raise NotImplementedError
    // py:59  return (''.join((
    // py:60  '<{0:02X}>'.format(c)
    // py:61  for c in tointiter(e.object[e.start:e.end])
    // py:62  )), e.end)
    // py:65  codecs.register_error('powerline_decode_error', powerline_decode_error)
    let mut out = String::with_capacity(bytes.len() * 4);
    for c in tointiter(bytes) {
        out.push_str(&format!("<{:02X}>", c));
    }
    (out, bytes.len())
}

/// Module-level counter mirroring Python's `last_swe_idx` global at
/// `powerline/lib/unicode.py:68`. Bumped by `register_strwidth_error`
/// each time a new error handler is registered.
static LAST_SWE_IDX: AtomicUsize = AtomicUsize::new(0);

/// Port of `register_strwidth_error()` from
/// `powerline/lib/unicode.py:71-103`.
///
/// Python registers a global codec error handler named
/// `powerline_encode_strwidth_error_<N>` that replaces unencodable
/// chars with question marks proportional to display width. The
/// returned string is the handler name, used later by `s.encode(enc,
/// errors=name)` callers.
///
/// Rust has no codec error registry. The port returns a `(name,
/// handler)` pair: `name` matches Python's generated identifier so
/// log/debug output stays consistent, `handler` is a closure callers
/// invoke directly when they hit unencodable runs. The internal
/// `LAST_SWE_IDX` counter mirrors Python's `global last_swe_idx`.
pub fn register_strwidth_error<F>(strwidth: F) -> (String, Box<dyn Fn(&str) -> (String, usize)>)
where
    F: Fn(&str) -> usize + Send + Sync + 'static,
{
    // py:68  last_swe_idx = 0
    // py:71  def register_strwidth_error(strwidth):
    // py:72-92  docstring
    // py:93  global last_swe_idx
    // py:94  last_swe_idx += 1
    let idx = LAST_SWE_IDX.fetch_add(1, Ordering::SeqCst) + 1;
    // py:96  def powerline_encode_strwidth_error(e):
    // py:97  if not isinstance(e, UnicodeEncodeError):
    // py:98  raise NotImplementedError
    // py:99  return ('?' * strwidth(e.object[e.start:e.end]), e.end)
    let handler: Box<dyn Fn(&str) -> (String, usize)> = Box::new(move |slice: &str| {
        let w = strwidth(slice);
        ("?".repeat(w), slice.len())
    });
    // py:101  ename = 'powerline_encode_strwidth_error_{0}'.format(last_swe_idx)
    // py:102  codecs.register_error(ename, powerline_encode_strwidth_error)
    // py:103  return ename
    let ename = format!("powerline_encode_strwidth_error_{}", idx);
    (ename, handler)
}

/// Port of `unichr()` from `powerline/lib/unicode.py:21-32`.
///
/// Convert a Unicode codepoint to a String. Python defines this as a
/// compat shim for the Py2/Py3 split (Py2 had `unichr`, Py3 has
/// `chr`); the UCS-2 build at py:27-32 emits a surrogate pair for
/// codepoints above U+FFFF.
///
/// Rust's `char` is a full Unicode scalar value (any valid codepoint
/// except surrogates), so the UCS-2 surrogate-pair fallback never
/// fires here — codepoints above U+FFFF map directly to a single
/// `char` via `char::from_u32`.
///
/// Returns `None` for surrogate codepoints (D800-DFFF) which Rust
/// can't represent as a `char`. Python's UCS-4 branch (py:21) also
/// errors on surrogates via `chr`, matching the Rust behaviour.
pub fn unichr(ch: u32) -> Option<String> {
    // py:21  unichr = chr  (UCS-4 path)
    // py:27-32  UCS-2 surrogate-pair encoding — unreachable in Rust
    //           since `char` is a 32-bit scalar value.
    char::from_u32(ch).map(|c| c.to_string())
}

/// Port of the inner `powerline_encode_strwidth_error()` closure from
/// `powerline/lib/unicode.py:96-99`.
///
/// codecs error handler that replaces unencodable codepoints with
/// `'?'` repeated by the codepoint's display width. The Rust port
/// surfaces it as a free fn taking the strwidth callable + the raw
/// substring + its (start, end) byte range so callers route through
/// any codec they like.
pub fn powerline_encode_strwidth_error<F>(
    strwidth: F,
    object: &str,
    start: usize,
    end: usize,
) -> (String, usize)
where
    F: Fn(&str) -> usize,
{
    // py:96  def powerline_encode_strwidth_error(e):
    // py:97  if not isinstance(e, UnicodeEncodeError):
    // py:98  raise NotImplementedError
    // py:99  return ('?' * strwidth(e.object[e.start:e.end]), e.end)
    let slice = &object[start..end];
    let width = strwidth(slice);
    ("?".repeat(width), end)
}

/// Input dispatch for [`out_u`] — Python's `out_u(s)` accepts either a
/// `unicode` (str) or `bytes`. Rust models the type-dispatch as a sum
/// type since runtime overloading isn't available.
pub enum OutUInput<'a> {
    /// Already-decoded unicode string (py:113-114 fast path).
    Str(&'a str),
    /// Bytes that need decoding (py:115-116 lossy fallback).
    Bytes(&'a [u8]),
}

/// Port of `out_u()` from `powerline/lib/unicode.py:106-118`.
///
/// Unified dispatcher that mirrors Python's runtime
/// `isinstance(s, unicode|bytes)` branch at py:113-118. Calls
/// [`out_u_str`] / [`out_u_bytes`] based on the input variant.
///
/// The `else` branch at py:117-118 raises `TypeError`; Rust returns
/// `Err` instead since the sum type encodes the valid inputs.
pub fn out_u(input: OutUInput<'_>) -> String {
    match input {
        OutUInput::Str(s) => out_u_str(s),
        OutUInput::Bytes(b) => out_u_bytes(b),
    }
}

/// Input dispatch for [`string`] — Python's `string(s)` is defined
/// twice (Py2 at py:163-167, Py3 at py:169-173). The Rust port
/// dispatches on the input variant.
pub enum StringInput<'a> {
    /// Py3 str input (`type(s) is str` at py:170 → identity).
    Str(&'a str),
    /// Bytes input → decode UTF-8 (py:171).
    Bytes(&'a [u8]),
}

/// Port of `string()` from `powerline/lib/unicode.py:163-167` (Py2)
/// or `py:169-173` (Py3).
///
/// The Py3 branch returns `s` for str input and `s.decode('utf-8')`
/// for non-str input. The Rust port mirrors the Py3 dispatch since
/// Python 2 is end-of-life.
pub fn string(input: StringInput<'_>) -> String {
    match input {
        StringInput::Str(s) => string_from_str(s),
        StringInput::Bytes(b) => string_from_bytes(b),
    }
}

/// Port of `out_u()` from `powerline/lib/unicode.py:106-118`.
///
/// Return unicode string suitable for displaying. Python decodes
/// bytes with `get_preferred_output_encoding()` falling through to
/// `powerline_decode_error`. Rust port:
///   - `&str` input → identity copy (py:113-114)
///   - `&[u8]` input → lossy UTF-8 decode (py:115-116)
pub fn out_u_str(s: &str) -> String {
    // py:106  def out_u(s):
    // py:107  '''Return unicode string suitable for displaying
    // py:108
    // py:109  Unlike other functions assumes get_preferred_output_encoding() first. Unlike
    // py:110  u() does not throw exceptions for invalid unicode strings. Unlike
    // py:111  safe_unicode() does throw an exception if object is not a string.
    // py:112  '''
    // py:113  if isinstance(s, unicode):
    // py:114  return s
    s.to_string()
}

/// Bytes overload of `out_u()`. Falls back to lossy UTF-8 decoding
/// (Python tries `get_preferred_output_encoding()` then
/// `powerline_decode_error`; Rust uses `from_utf8_lossy` which
/// substitutes U+FFFD for invalid sequences).
pub fn out_u_bytes(s: &[u8]) -> String {
    // py:115  elif isinstance(s, bytes):
    // py:116  return unicode(s, get_preferred_output_encoding(), 'powerline_decode_error')
    // py:117  else:
    // py:118  raise TypeError('Expected unicode or bytes instance, got {0}'.format(repr(type(s))))
    String::from_utf8_lossy(s).into_owned()
}

/// Port of `safe_unicode()` from `powerline/lib/unicode.py:121`.
///
/// Return unicode instance without raising an exception.
///
/// Python tries ASCII → UTF-8 → `__str__`/`__repr__` → preferred output
/// encoding → recursive fallback. In Rust, every input has a
/// well-defined `Display` or `Debug` impl, and `String::from_utf8_lossy`
/// covers the byte-slice fallback. The port collapses the cascade into
/// the one operation that survives all cases.
pub fn safe_unicode_str(s: &str) -> String {
    // py:121  def safe_unicode(s):
    // py:122  '''Return unicode instance without raising an exception.
    // py:123-132  docstring
    // py:133  try:
    // py:134  try:
    // py:135  if type(s) is bytes:
    // py:136  return unicode(s, 'ascii')
    // py:137  else:
    // py:138  return unicode(s)
    // py:139  except UnicodeDecodeError:
    // py:140  try:
    // py:141  return unicode(s, 'utf-8')
    // py:142  except TypeError:
    // py:143  return unicode(str(s), 'utf-8')
    // py:144  except UnicodeDecodeError:
    // py:145  return unicode(s, get_preferred_output_encoding())
    // py:146  except Exception as e:
    // py:147  return safe_unicode(e)
    s.to_string()
}

/// Bytes overload of `safe_unicode()`.
///
/// Falls back to lossy UTF-8 decoding for non-ASCII bytes
/// (Python tries UTF-8 then the preferred output encoding; Rust uses
/// `from_utf8_lossy` which substitutes U+FFFD for invalid sequences).
pub fn safe_unicode_bytes(s: &[u8]) -> String {
    // py:121
    String::from_utf8_lossy(s).into_owned()
}

/// `safe_unicode` accepting any `Display` value.
///
/// Mirrors Python's `unicode(s)` fallback at py:140 which calls
/// `__str__` / `__repr__` on the input. Rust's `Display` is the analog.
pub fn safe_unicode<T: std::fmt::Display>(s: T) -> String {
    format!("{}", s)
}

/// Port of `FailedUnicode` from `powerline/lib/unicode.py:150-159`.
///
/// Python: builtin `unicode` subclass indicating fatal error in
/// `.render()`. Callers check `isinstance(result, FailedUnicode)` to
/// detect failure without raising. Rust port is a newtype around
/// `String` so callers can pattern-match via `if let` instead of
/// `isinstance`.
#[derive(Debug, Clone, PartialEq, Eq)]
// py:150  class FailedUnicode(unicode):
// py:151  '''Builtin ``unicode`` subclass indicating fatal error
// py:152
// py:153  If your code for some reason wants to determine whether `.render()` method
// py:154  failed it should check returned string for being a FailedUnicode instance.
// py:155  Alternatively you could subclass Powerline and override `.render()` method
// py:156  to do what you like in place of catching the exception and returning
// py:157  FailedUnicode.
// py:158  '''
// py:159  pass
pub struct FailedUnicode(pub String);

impl FailedUnicode {
    /// Construct from any `Display` value.
    pub fn new<T: std::fmt::Display>(s: T) -> Self {
        FailedUnicode(format!("{}", s))
    }

    /// Borrow as `&str`.
    pub fn as_str(&self) -> &str {
        &self.0
    }

    /// Consume into owned `String`.
    pub fn into_string(self) -> String {
        self.0
    }
}

impl std::fmt::Display for FailedUnicode {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.write_str(&self.0)
    }
}

impl From<String> for FailedUnicode {
    fn from(s: String) -> Self {
        FailedUnicode(s)
    }
}

impl From<&str> for FailedUnicode {
    fn from(s: &str) -> Self {
        FailedUnicode(s.to_string())
    }
}

/// Port of `string()` from `powerline/lib/unicode.py:162-173`.
///
/// Py2: encode `unicode` to UTF-8 bytes. Py3: decode `bytes` to
/// `str` via UTF-8. Either way the result is the Py-native `str`.
/// Rust's `&str` is already UTF-8 so the str overload is identity;
/// the bytes overload uses `from_utf8_lossy`.
pub fn string_from_str(s: &str) -> String {
    // py:162  if sys.version_info < (3,):
    // py:163  def string(s):
    // py:164  if type(s) is not str:
    // py:165  return s.encode('utf-8')
    // py:166  else:
    // py:167  return s
    // py:168  else:
    // py:169  def string(s):
    // py:170  if type(s) is not str:
    // py:171  return s.decode('utf-8')
    // py:172  else:
    // py:173  return s
    s.to_string()
}

/// Bytes overload of `string()`. Py3 decodes UTF-8 at py:171.
pub fn string_from_bytes(s: &[u8]) -> String {
    // py:176  string.__doc__ = (
    // py:177  '''Transform ``unicode`` or ``bytes`` object into ``str`` object
    // py:178-186  docstring
    // py:187  )
    String::from_utf8_lossy(s).into_owned()
}

/// Port of `surrogate_pair_to_character()` from
/// `powerline/lib/unicode.py:190-193`.
///
/// Transform a pair of surrogate codepoints to one codepoint.
pub fn surrogate_pair_to_character(high: u32, low: u32) -> u32 {
    // py:190  def surrogate_pair_to_character(high, low):
    // py:191  '''Transform a pair of surrogate codepoints to one codepoint
    // py:192  '''
    // py:193  return 0x10000 + ((high - 0xD800) << 10) + (low - 0xDC00)
    // py:196  _strwidth_documentation = (
    // py:197  '''Compute string width in display cells
    0x10000 + ((high - 0xD800) << 10) + (low - 0xDC00)
}

/// Port of `strwidth_ucs_4()` from
/// `powerline/lib/unicode.py:247-254`.
///
/// Compute string width in display cells. Rust `char` is already
/// UCS-4 so the UCS-4 path is the default.
///
/// `width_data` maps east_asian_width category codes ('F', 'H',
/// 'W', 'Na', 'A', 'N') to display widths. Combining marks count
/// as 0 (py:251).
///
/// Note: Rust stdlib has no `east_asian_width` / `combining` lookup.
/// Until a foundational unicode-properties crate is wired in, this
/// function falls back to "Narrow" (`'Na' or 'N'`) for every char.
/// The structure of the port (lookup + sum) matches Python so
/// callers wire up identically once the table is available.
pub fn strwidth_ucs_4(width_data: &HashMap<String, usize>, string: &str) -> usize {
    // py:248-254  sum(0 if combining(c) else width_data[east_asian_width(c)] for c in string)
    let fallback = width_data
        .get("N")
        .or_else(|| width_data.get("Na"))
        .copied()
        .unwrap_or(1);
    string.chars().map(|_c| fallback).sum()
}

/// Port of `strwidth_ucs_2()` from
/// `powerline/lib/unicode.py:267-276`.
///
/// UCS-2 variant — handles surrogate pairs. Rust's `&str` is always
/// UTF-8 and `chars()` yields decoded `char` values (no surrogate
/// pairs ever surface), so this path collapses to the UCS-4 impl.
pub fn strwidth_ucs_2(width_data: &HashMap<String, usize>, string: &str) -> usize {
    // py:267-276  surrogate-pair aware; collapses to UCS-4 in Rust
    strwidth_ucs_4(width_data, string)
}

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

    #[test]
    fn u_passes_through_str() {
        assert_eq!(u("hello"), "hello");
    }

    #[test]
    fn u_handles_utf8() {
        assert_eq!(u("héllo →"), "héllo →");
    }

    #[test]
    fn tointiter_yields_byte_ints() {
        let v: Vec<u8> = tointiter(b"abc").collect();
        assert_eq!(v, vec![b'a', b'b', b'c']);
    }

    #[test]
    fn tointiter_handles_empty() {
        let v: Vec<u8> = tointiter(b"").collect();
        assert!(v.is_empty());
    }

    #[test]
    fn powerline_decode_error_formats_hex() {
        // py:60  '<{0:02X}>'.format(c)
        let (s, end) = powerline_decode_error(&[0xff, 0xfe]);
        assert_eq!(s, "<FF><FE>");
        assert_eq!(end, 2);
    }

    #[test]
    fn powerline_decode_error_handles_empty_range() {
        let (s, end) = powerline_decode_error(&[]);
        assert!(s.is_empty());
        assert_eq!(end, 0);
    }

    #[test]
    fn register_strwidth_error_generates_unique_name() {
        let (n1, _) = register_strwidth_error(|s| s.chars().count());
        let (n2, _) = register_strwidth_error(|s| s.chars().count());
        assert!(n1.starts_with("powerline_encode_strwidth_error_"));
        assert!(n2.starts_with("powerline_encode_strwidth_error_"));
        assert_ne!(n1, n2);
    }

    #[test]
    fn register_strwidth_error_handler_emits_question_marks() {
        let (_, handler) = register_strwidth_error(|s| s.chars().count());
        let (out, end) = handler("…");
        // 1 character wide → 1 question mark
        assert_eq!(out, "?");
        assert_eq!(end, "…".len());
    }

    #[test]
    fn register_strwidth_error_handler_respects_width() {
        // Simulate fullwidth: 2 cells per char
        let (_, handler) = register_strwidth_error(|s| s.chars().count() * 2);
        let (out, _) = handler("Ａ");
        assert_eq!(out, "??");
    }

    #[test]
    fn out_u_str_passes_through() {
        assert_eq!(out_u_str("hello"), "hello");
    }

    #[test]
    fn out_u_bytes_decodes_valid_utf8() {
        assert_eq!(out_u_bytes(b"hello"), "hello");
    }

    #[test]
    fn out_u_bytes_lossy_on_invalid() {
        let bad = &[0xff, b'a'];
        let out = out_u_bytes(bad);
        assert!(out.contains('\u{FFFD}'));
    }

    #[test]
    fn safe_unicode_str_round_trips() {
        assert_eq!(safe_unicode_str("hello"), "hello");
    }

    #[test]
    fn safe_unicode_bytes_handles_valid_utf8() {
        assert_eq!(safe_unicode_bytes(b"hello"), "hello");
    }

    #[test]
    fn safe_unicode_bytes_lossy_on_invalid() {
        let bad = &[0xff, 0xfe, b'a'];
        let result = safe_unicode_bytes(bad);
        assert!(result.contains('\u{FFFD}') || result.contains('a'));
    }

    #[test]
    fn safe_unicode_display_works_on_int() {
        assert_eq!(safe_unicode(42), "42");
    }

    #[test]
    fn string_from_str_identity() {
        assert_eq!(string_from_str("hello"), "hello");
    }

    #[test]
    fn string_from_bytes_decodes_utf8() {
        assert_eq!(string_from_bytes(b"hello"), "hello");
    }

    #[test]
    fn string_from_bytes_lossy_on_invalid() {
        assert!(string_from_bytes(&[0xff, b'x']).contains('\u{FFFD}'));
    }

    #[test]
    fn surrogate_pair_to_character_round_trips_emoji_range() {
        // U+1F600 GRINNING FACE = surrogate pair D83D DE00
        let cp = surrogate_pair_to_character(0xD83D, 0xDE00);
        assert_eq!(cp, 0x1F600);
    }

    #[test]
    fn surrogate_pair_to_character_low_surrogate_boundary() {
        // U+10000 = D800 DC00 (smallest surrogate pair)
        let cp = surrogate_pair_to_character(0xD800, 0xDC00);
        assert_eq!(cp, 0x10000);
    }

    #[test]
    fn failed_unicode_display_returns_inner_string() {
        let fu = FailedUnicode::new("No window 5");
        assert_eq!(fu.to_string(), "No window 5");
    }

    #[test]
    fn failed_unicode_eq_compares_by_value() {
        let a = FailedUnicode::from("err");
        let b = FailedUnicode::from(String::from("err"));
        assert_eq!(a, b);
    }

    #[test]
    fn failed_unicode_into_string_yields_inner() {
        let fu = FailedUnicode::new("oops");
        assert_eq!(fu.into_string(), "oops");
    }

    #[test]
    fn strwidth_ucs_4_sums_with_default_table() {
        let mut width_data = HashMap::new();
        width_data.insert("N".to_string(), 1);
        width_data.insert("Na".to_string(), 1);
        assert_eq!(strwidth_ucs_4(&width_data, "hello"), 5);
        assert_eq!(strwidth_ucs_4(&width_data, ""), 0);
    }

    #[test]
    fn strwidth_ucs_4_uses_table_value() {
        let mut width_data = HashMap::new();
        // Pretend everything is fullwidth
        width_data.insert("N".to_string(), 2);
        width_data.insert("Na".to_string(), 2);
        assert_eq!(strwidth_ucs_4(&width_data, "hi"), 4);
    }

    #[test]
    fn strwidth_ucs_4_empty_table_falls_back_to_one() {
        let width_data = HashMap::new();
        // Falls back to 1 per char when no key matches
        assert_eq!(strwidth_ucs_4(&width_data, "abc"), 3);
    }

    #[test]
    fn strwidth_ucs_2_matches_ucs_4_for_basic_strs() {
        let mut width_data = HashMap::new();
        width_data.insert("N".to_string(), 1);
        width_data.insert("Na".to_string(), 1);
        assert_eq!(
            strwidth_ucs_2(&width_data, "hello"),
            strwidth_ucs_4(&width_data, "hello")
        );
    }

    #[test]
    fn unichr_returns_basic_ascii() {
        // py:21  unichr = chr
        assert_eq!(unichr(b'A' as u32), Some("A".to_string()));
        assert_eq!(unichr(0x20), Some(" ".to_string()));
    }

    #[test]
    fn unichr_returns_high_codepoint_as_single_char() {
        // py:32  UCS-2 surrogate-pair encoding — Rust char covers
        // the codepoint directly, no surrogates emitted.
        let r = unichr(0x1F600).unwrap();
        assert_eq!(r.chars().count(), 1);
        assert_eq!(r, "\u{1F600}");
    }

    #[test]
    fn unichr_returns_none_for_surrogate_codepoint() {
        // Surrogates (D800-DFFF) are not valid Rust chars.
        assert_eq!(unichr(0xD800), None);
        assert_eq!(unichr(0xDFFF), None);
    }

    #[test]
    fn powerline_encode_strwidth_error_replaces_with_question_marks() {
        // py:99  '?' * strwidth(object[start:end])
        let object = "abc";
        let strwidth = |s: &str| s.chars().count();
        let (replacement, new_end) = powerline_encode_strwidth_error(strwidth, object, 0, 3);
        assert_eq!(replacement, "???");
        assert_eq!(new_end, 3);
    }

    #[test]
    fn out_u_dispatches_str_to_str_branch() {
        // py:113-114
        let r = out_u(OutUInput::Str("hello"));
        assert_eq!(r, "hello");
    }

    #[test]
    fn out_u_dispatches_bytes_to_bytes_branch() {
        // py:115-116
        let r = out_u(OutUInput::Bytes(b"world"));
        assert_eq!(r, "world");
    }

    #[test]
    fn string_dispatches_str_to_identity() {
        // py:170  type(s) is str → return s
        let r = string(StringInput::Str("hello"));
        assert_eq!(r, "hello");
    }

    #[test]
    fn string_dispatches_bytes_to_utf8_decode() {
        // py:171  s.decode('utf-8')
        let r = string(StringInput::Bytes(b"hello"));
        assert_eq!(r, "hello");
    }
}