ToTitleCase keeps an all-caps word unchanged when it's joined by '-', '.' or a tab: "foo-BAR" → "Foo-BAR" but "FOO-BAR" → "Foo-Bar"

[matt-edmondson]

What's wrong

CaseConverter/CaseConverter.cs:214-231, LowercaseAllCapsWords, splits the input only on ' ':

string[] words = input.Split(' ');
...
builder.Append(IsAllCaps(word) ? word.ToLowerInvariant() : word);

TextInfo.ToTitleCase treats -, . and all whitespace, including tabs, as word boundaries, and it keeps any all-caps word it finds as an acronym. When a space-separated token such as foo-BAR contains lowercase letters, it isn't lowercased first, so TextInfo keeps BAR as is.

Reproduction

Input	Actual	Expected (per-word rule)
"foo-BAR"	"Foo-BAR"	"Foo-Bar"
"FOO-BAR"	"Foo-Bar"	same
"API-key"	"API-Key"	"Api-Key"
"foo.BAR"	"Foo.BAR"	"Foo.Bar"
"HELLO\tworld"	"HELLO\tWorld"	"Hello\tWorld"
"HELLO world"	"Hello World"	same

The ToTitleCase docs and commit 0b9154a ("Normalize an all-caps word per word, not per string") promise that a word converts the same way whatever else is in the string. These cases break that promise: the result for BAR depends on whether the text next to it contains lowercase letters, and on whether the separator is a space or a tab. Reproduced with MSTest asserts.

Suggested fix

In LowercaseAllCapsWords, split on the same separators TextInfo.ToTitleCase uses: any whitespace, -, ., and so on. Or lowercase each maximal run of letters and apostrophes that is entirely uppercase, instead of each space-delimited token.

Acceptance: every row above produces the expected output, and the existing ToTitleCase tests (including the punctuation cases from d8bdab2) still pass.

[matt-edmondson]

Triage

• Category: Bug
• Priority: Low. The output casing is inconsistent only for all-caps segments joined by -, . or a tab inside a token that also has lowercase letters. Nothing crashes, and the common space-separated case works.
• Area / suggested assignee: LowercaseAllCapsWords in CaseConverter.cs (:214-231), which is part of ToTitleCase. Maintainer: @matt-edmondson.
• Duplicates: None. Related: A letter after a digit starts a new word in Title/Pascal/Camel but not in Snake/Kebab: "1st place" → "1St Place", "abc123def" → "abc_123def" #78 (word-boundary rules differ between the Title/Pascal/Camel and Snake/Kebab paths) and Pascal/Camel/Snake/Kebab/Macro split words at an apostrophe: "don't stop" → "DonTStop" / "don_t_stop" #83 (apostrophe splitting). All three are about where a "word" starts, so a single shared word-boundary definition would settle them together.
• In progress: No. Open PR Trim before case-mapping the first char in ToUppercaseFirstChar/ToLowercaseFirstChar #81 targets ToUppercaseFirstChar/ToLowercaseFirstChar do nothing when the input starts with whitespace (" hello" → "hello") #79 and does not touch LowercaseAllCapsWords.

Next steps: Split on the same separators TextInfo.ToTitleCase uses, or lowercase each fully uppercase run of letters. Add the reproduction table as tests, and keep the existing punctuation tests from d8bdab2 passing.

---

Generated by Claude Code