stringo 2.0.0

Zero-dependency Dart string toolkit - case conversion, word splitting, slugify, truncate, mask, and whitespace normalization.

CHANGELOG

2.0.0

A performance rewrite. Every operation is now built on a single tokenizer that reports word boundaries as index pairs and allocates nothing, and conversions stream those boundaries straight into one StringBuffer. Case mapping runs inline on code units for ASCII input and falls back to Dart's native Unicode casing otherwise, so accented text is never ASCII-folded.

The package still has zero dependencies and still runs everywhere Dart does. Native code through dart:ffi was measured and rejected: marshalling UTF-16 across the boundary costs more than the work itself, making it slower than staying in Dart.

Case-conversion output is unchanged from 1.0.0 apart from the corrections listed under Fixed. That equivalence is verified by running the 1.0.0 implementations against the new ones over generated input.

Performance

Measured on one machine with dart run benchmark/stringo_benchmark.dart:

Operation	1.0.0	2.0.0
isBlank, 100,000 chars	3,166,000 ns	3 ns	~1,000,000x
isBlank, 10 chars	516 ns	7 ns	79x
toWords	1,880 ns	148 ns	12.7x
toCamelCase	3,769 ns	288 ns	13.1x
toSnakeCase	2,111 ns	307 ns	6.9x
slugify, 54 chars	11,300 ns	362 ns	31x
200,000 identifiers to snake_case	341 ms	56 ms	6.1x
removeEmptyLines, 400 indented lines	14.6 ms	0.12 ms	121x
removeEmptyLines, 8 KB unbroken run	1,137 ms	0.05 ms	~22,700x

isBlank is the outlier because 1.0.0 answered it by allocating two whole-string copies and compiling a regex, which made it linear in the length of the input. It is now a scan that returns at the first non-whitespace character. If your code guarded isBlank with a length check or avoided it on large strings, you can drop that workaround.

slugify gained its final 3x late: it emitted a one-character separator with StringBuffer.write(String) while writing content with writeCharCode, and a buffer that receives both runs about 2.6x slower than one kept in a single mode. Case conversion had already been fixed for this; the transform layer had not.

removeEmptyLines was the other pathological case. Its pattern (?:[\t ]*(?:\r?\n|\r))+ backtracked catastrophically: inside a run of spaces not followed by a line break, the engine consumed the whole run then gave characters back one at a time. Cost was quadratic in the length of such a run, so a document indented 40 spaces cost roughly ten times an unindented one, and a 20 KB run of spaces took about seven seconds. It is now a linear scanner. There is no longer any RegExp in the transform layer at all.

Added

• Stringo, a namespace exposing every operation as a plain static function: Stringo.snakeCase('userProfileField'). Useful for passing an operation as a value, and as an escape hatch when an extension member name collides with one your project already defines. Every extension member is a one-line delegation to it.
• Precompiled patterns. Alongside the existing String sources there are now patternAlphanumeric, patternStartsWithNumber, patternContainsDigits, patternNumeric, patternAlphabet, and patternHasCapitalLetter. Prefer them: Dart does not cache compiled patterns, so RegExp(regexNumeric) recompiles on every call.

Removed

• isEmptyOrNull and isNotEmptyOrNull, which were exact synonyms of isBlank and isNotBlank. Rename; behavior is identical.

Changed

• mask and insert on a String? receiver now return String? and yield null for a null receiver, instead of '' and the inserted value. Null handling is now uniform across every transforming member on the nullable extensions. Append ?? '' to restore the old result exactly. orEmpty and toCharArray() keep their existing null behavior.
• mask validates its bounds before checking the receiver, so a negative visibleStart or visibleEnd throws ArgumentError even when the receiver is null.
• slugify treats -, _, and whitespace identically. Any run of them, in any mix, collapses to a single separator. Previously a literal hyphen passed through untouched, so 'a-b'.slugify(separator: '_') returned 'a-b'; it now returns 'a_b'. The default hyphen separator is unaffected.
• truncate counts its suffix against the requested length instead of appending it on top. 'Hello World'.truncate(5) was 'Hello...' (8 characters) and is now 'He...' (5). Ask for length + suffix.length to reproduce the old output. A suffix longer than length yields exactly the suffix.
• toCharArray() splits by Unicode code point rather than UTF-16 code unit, so a surrogate pair such as an emoji is no longer torn in half. It is still code points rather than grapheme clusters; use package:characters for those.

Fixed

• toWords no longer emits empty elements. ''.toWords was [''] and is now []; ' a '.toWords was ['', 'a', ''] and is now ['a'].
• A leading separator no longer capitalizes camelCase output. '_leading'.toCamelCase returned 'Leading' because the phantom empty first word shifted the real one; it now returns 'leading'. toPascalCase was never affected.
• The first character of supplementary-plane text is now cased. capitalizeFirstLetter, lowercaseFirstLetter, and capitalizeFirstLowerRest operated on s[0], which is a lone surrogate for any character outside the Basic Multilingual Plane, and casing a lone surrogate does nothing. The first letter of Deseret, Osage, or Adlam text was silently left alone, which also affected toPascalCase, toCamelCase, toTitleCase, toTitle, and the four Pascal/camel separator variants. Characters without a case mapping, such as emoji, are unchanged as before.

1.0.0

Initial release.

stringo is the string toolkit extracted from dart_helper_utils into a standalone, zero-dependency package. dart_helper_utils 6.1.0 and later re-export it, so existing code keeps working without an import change.

Included

• Case conversion: toWords plus 15 case styles (toCamelCase, toPascalCase, toSnakeCase, toKebabCase, toDotCase, toTitleCase, toScreamingSnakeCase, toScreamingKebabCase, toPascalSnakeCase, toPascalKebabCase, toTrainCase, toCamelSnakeCase, toCamelKebabCase, toFlatCase, toScreamingCase), toTitle, capitalizeFirstLetter, lowercaseFirstLetter, capitalizeFirstLowerRest, shouldIgnoreCapitalization, tryToLowerCase, tryToUpperCase.
• Transforms: slugify, truncate, mask, normalizeWhitespace, clean, toOneLine, removeWhiteSpaces, removeEmptyLines, words, lines, nullIfEmpty, nullIfBlank, orEmpty, insert, toCharArray, equalsIgnoreCase, removeSurrounding, replaceBefore, replaceAfter.
• Checks: isBlank / isEmptyOrNull (and negations), isNumeric, isAlphabet, isAlphanumeric, startsWithNumber, containsDigits, hasCapitalLetter, hasMatch, and the six backing regex constants.

Changes from the dart_helper_utils originals

• toTitleCase and toTitle now always capitalize the first word. Previously a leading stop word was lowercased, so 'the lord of the rings'.toTitleCase produced 'the Lord of the Rings'. It now produces 'The Lord of the Rings', matching conventional title casing and the behavior the original documentation already described.
• titleCaseExceptions is now public and backed by a Set for constant-time lookup instead of a private List.
• mask rejects negative visibleStart / visibleEnd with an ArgumentError instead of failing later with a RangeError.
• insert is implemented with substring rather than a List<String> round-trip. Behavior is unchanged; out-of-range indices still throw a RangeError.
• Extension type names dropped the DHU prefix and are grouped by concern: StringCaseExtensions, NullableStringCaseExtensions, StringTransformExtensions, NullableStringTransformExtensions, and StringChecksExtensions.

Deliberately not included

Domain validation (isValidEmail, isValidPhoneNumber, isValidUrl, isValidUsername, isValidCurrency, isUuid, isValidIp4), MIME and file type checks, parseDuration, and base64 helpers all remain in dart_helper_utils. Those judge real-world formats rather than transform text, and their correct behavior varies by project.