Background

PyStr::title() (crates/vm/src/builtins/str.rs:1040) walks each codepoint and chooses one of three actions per word boundary:

1. Lowercase character starting a new word → call to_titlecase()
2. Uppercase / titlecase character starting a new word → push unchanged
3. Cased character continuing a word → call to_lowercase()

Branch 2 is the gap. For ASCII letters and most Unicode characters, to_titlecase(c) == c, so pushing unchanged happens to produce the right result. But Latin Extended-B contains digraph triplets (LATIN LETTERS DZ / DŽ / LJ / NJ) where the uppercase and titlecase forms are distinct codepoints:

For these, c.to_titlecase() returns a different character than c itself, and the Branch 2 fall-through silently produced the wrong output.

Reproduction

'DZ'.title()        # CPython: 'Dz'  Pre-fix: 'DZ'
'DŽ'.title()        # CPython: 'Dž'  Pre-fix: 'DŽ'

12 cases probed against CPython 3.14.4 covering all 4 digraph families × 3 forms each (uppercase / titlecase / lowercase):

Fix

} else if c.is_uppercase() || c.is_titlecase() {
    if previous_is_cased {
        title.extend(c.to_lowercase());
    } else {
        title.extend(c.to_titlecase());   // was: title.push_char(c)
    }
    previous_is_cased = true;
}

The lowercase branch (line 1047) already uses c.to_titlecase() — Branch 2 is now symmetric. char::to_titlecase returns the same character when invoked on an ASCII letter or already-titlecase codepoint, so no existing case regresses.

Verification

CPython 3.14.4 byte-identical across:

• 12 single-codepoint digraph cases — all 4 families × 3 forms
• 8 mixed-string cases — digraphs at start, mid-word, as standalone word, consecutive digraphs, multi-word strings ('hello DZ world' → 'Hello Dz World')
• 14 sibling-method cases — upper, lower, swapcase, capitalize, casefold, istitle on digraphs unchanged

New Rust unit cases added to str_title (crates/vm/src/builtins/str.rs:2664) covering U+01F1 → U+01F2 and U+01C4 → U+01C5. cargo test -p rustpython-vm str_title and str_istitle both pass.

Regression sweep: test_str, test_codecs, test_format, test_pprint — 484 tests, 0 regressions.

No expectedFailure markers in Lib/test/test_str.py are tied to this root cause; this PR has no test unmask.

Issue scope note

Issue #7527 listed five examples. Probing on current main shows four already pass:

This PR closes the only remaining actionable case. The Ᲊ.istitle() expected value in the issue body appears to predate a Unicode database update in CPython 3.14 — current CPython behavior matches RustPython.

Closes #7527

Summary by CodeRabbit

• Bug Fixes
  • Corrected title-casing for Unicode characters whose titlecase mapping expands to multiple characters (including Latin Extended‑B digraphs), so title() now produces proper multi-character expansions and consistent title-case output across scripts.