Pascal/Camel/Snake/Kebab/Macro split words at an apostrophe: "don't stop" → "DonTStop" / "don_t_stop"

[matt-edmondson]

What's wrong

CaseConverter/CaseConverter.cs:35-56, ReplaceNonAlphaNumericWithSpace, turns an apostrophe into a space. It is used by the Pascal/Camel path (~line 327) and by the Snake/Kebab/Macro path (~line 399). The letters after an apostrophe therefore become a separate word:

Input	Method	Actual	Expected
"don't stop"	ToPascalCase	"DonTStop"	"DontStop"
"don't stop"	ToCamelCase	"donTStop"	"dontStop"
"don't stop"	ToSnakeCase	"don_t_stop"	"dont_stop"
"don't stop"	ToMacroCase	"DON_T_STOP"	"DONT_STOP"
"o'neil"	ToPascalCase / ToSnakeCase	"ONeil" / "o_neil"	"Oneil" / "oneil"

Reproduced with MSTest asserts.

Why it matters

Generating identifiers, slugs or constant names from human text such as UI labels or titles is a primary use case, and contractions are common in that text. ToTitleCase was fixed in d8bdab2 to keep "Don't" as one word, so the sibling methods now disagree with it. Lodash and similar libraries strip in-word apostrophes (lodash's reApos step) for this reason.

Suggested fix

In ReplaceNonAlphaNumericWithSpace, drop ' and U+2019 (right single quotation mark) when they sit between two letters, instead of replacing them with a space. Apostrophes elsewhere (leading or trailing quotes) can keep acting as separators.

Acceptance: every row above produces the expected output, with tests for both the ASCII ' and U+2019.

[matt-edmondson]

Triage

• Category: Bug
• Priority: Medium. Contractions are common in the human text that identifiers and slugs are generated from. The output is wrong rather than a crash, and it disagrees with ToTitleCase, which already keeps "Don't" as one word.
• Area / suggested assignee: ReplaceNonAlphaNumericWithSpace in CaseConverter.cs, which the Pascal/Camel and Snake/Kebab/Macro paths share. Maintainer: @matt-edmondson.
• Duplicates: None. Related: Pascal/Camel/Snake/Kebab/Macro delete combining marks and non-ASCII digits: "हिन्दी भाषा" → "ह_न_द_भ_ष", NFD "éclair" → "EClair" #80 (combining marks and non-ASCII digits are deleted by the same word-splitting paths). The two fixes touch the same character classification, so they are best done together.
• In progress: No. The only open PR in this repo, Trim before case-mapping the first char in ToUppercaseFirstChar/ToLowercaseFirstChar #81, fixes ToUppercaseFirstChar/ToLowercaseFirstChar do nothing when the input starts with whitespace (" hello" → "hello") #79 (ToUppercaseFirstChar) and does not touch this helper.

Next steps: Treat ' and U+2019 between two letters as part of the word, add the table from this issue as data-driven tests for both apostrophe forms, and check that the result agrees with ToTitleCase.

---

Generated by Claude Code