StringiFor

Appearance

Upgrading to 1.10.0 ​

  • partition follows Python str.partition. If the separator is not found the parts are the string and two null strings: the second part was the separator, so a caller could not tell whether it was found. A separator as long as the string, or longer, is searched too: 'ab'%partition('ab') gives '', 'ab', '' (it gave 'ab', 'ab', '').
  • Assigning a part of a string to the string itself (astring = astring%raw(2:)) read freed memory and could give wrong data, depending on the platform; it is now safe.

Upgrading to 1.9.0 ​

Release 1.9.0 reads UTF-8 text as characters (Unicode and UTF-8). On ASCII text nothing changes. On non-ASCII text the methods that worked on single bytes now work on units, a unit being a well-formed UTF-8 sequence or a byte that is not part of one:

  • reverse keeps the bytes of each character in order: 'ab' followed by é reverses to é then 'ba', it gave invalid UTF-8.
  • match: ? and a class [...] match one character, 'caf?' matches café; a range compares codepoints.
  • scan, verify, strip/lstrip/rstrip with remove=, squeeze with a set take a set of characters: ã no longer matches the set é (they share their first byte), and a backward verify gives the first byte of a character, not its last one.
  • squeeze reduces runs of a repeated non-ASCII character, which it did not see.
  • transliterate maps characters to characters, so old_set and new_set can hold non-ASCII ones and the result can be longer than the string.

Bytes that are not well-formed UTF-8 (Latin-1 text, binary data) are units on their own and keep their old behaviour. The only difference for them is Latin-1 or binary content that happens to contain a well-formed sequence (the Latin-1 é is the bytes C3 A9, é in UTF-8): that pair is now one unit.

Two defects are fixed too: split_chunked crashed on a string made only of separators, it now gives no tokens as split does; end_with read one byte before a null string.

Upgrading to 1.4.0 ​

Release 1.4.0 makes the string methods linear in time and memory and fixes their edge cases. Three changes can alter the results of existing programs:

  • replace is not recursive. The replaced text is not searched again, as in Python str.replace: 'aaaa' with 'aa' replaced by 'a' gives 'aa' (it gave 'a'), and a new containing old no longer loops forever. count=0 replaces nothing (it replaced one occurrence).
  • Unformatted I/O of a string stores its length. write(unit) astring writes the length (an I8P integer) then the characters, and read reads them back exactly; strings were truncated at 100 characters. Unformatted files written by an earlier version are not readable.
  • Defined results on edge cases. to_number on a string that is not a number gives 0 (integer kinds) or a quiet NaN (real kinds), it was undefined; split of a string made only of separators gives no tokens, it gave the string itself; count (of characters) no longer undercounts; count('') and unique('') no longer loop forever.

Changelog ​

All notable changes to this project are documented here. Versions follow Semantic Versioning. Format follows Keep a Changelog.

[1.11.0] — 2026-10-31 ​

Added ​

  • regex: Regular expressions, a Pike VM linear in the subject

  • regex: Finditer, findall, split, sub and the re_* procedures

Documentation ​

  • regex: Regular expressions guide, tutorial chapter and recipes

  • readme: Regular expressions and UTF-8 in the features

[1.10.0] — 2026-10-30 ​

Added ​

  • utf8: Positions and lengths in characters with by=SF_CHARS

  • utf8: Slice, insert and len_last_word in characters with by=SF_CHARS

  • utf8: Widths in characters with by=SF_CHARS

  • utf8: Display width in terminal columns

  • xml: Escape_xml, unescape_xml and xml_error

  • utf8: Unicode whitespace with unicode=.true.

  • sort: Strsort and the orders strless, strless_nocase, strless_natural

Documentation ​

  • utf8: Positions and widths in characters

  • utf8: Display width, and its differences from wcwidth

  • xml: XML escaping

  • utf8: Unicode whitespace

  • sort: Sorting arrays of strings

Fixed ​

  • string_t: Assigning a part of a string to itself

  • string_t: Partition as Python str.partition ⚠ BREAKING CHANGE

[1.9.1] — 2026-10-30 ​

Fixed ​

  • io: Use-after-free in the module read_lines with skip_bom

[1.9.0] — 2026-10-30 ​

Added ​

  • unicode: Generate the Unicode property tables from the UCD

  • utf8: Add the UTF-8 core: validation, codepoints and repair

  • escape: Add the C escape codecs escape_c, unescape_c and escapes_error

  • utf8: Scan, verify and strip work on units

  • utf8: Reverse, squeeze and transliterate work on units

  • utf8: Match wildcard patterns by units

  • io: Read and write the UTF-8 byte order mark

Documentation ​

  • Add searching to the tutorial, unit, binary and stdin recipes

  • utf8: Document the methods that work on units, add the 1.9.0 upgrade notes ⚠ BREAKING CHANGE

Fixed ​

  • ci: Run install smoke test from the release workflow

  • string_t: Split_chunked without tokens, end_with on a null string

Performance ​

  • utf8: ASCII fast paths for the methods that work on units

[1.8.0] — 2026-10-04 ​

Added ​

  • string_t: Fast exact number casting, count/index/strip options

Documentation ​

  • Cover the new methods in tutorial, cookbook and home pages

[1.7.0] — 2026-10-04 ​

Added ​

  • string_t: Add character classes, tr-like and padding methods

[1.6.0] — 2026-10-03 ​

Added ​

  • string_t: Add read_number with an error status, keep empty fields in split

[1.5.0] — 2026-10-03 ​

Added ​

  • string_t: Add match of wildcard patterns, make glob safe

[1.4.0] — 2026-10-03 ​

Documentation ​

  • Add the 1.4.0 upgrade notes, sync run_tests.sh and release.sh

Performance ​

  • string_t: Make split, replace, unique, escape, join and read_* linear

[1.3.1] — 2026-10-03 ​

Changed ​

  • cmake: Find PENF and FACE by their standard config names

Fixed ​

  • makefile: Repair the make build and keep the doctests in sync

  • cmake: Make the installed package usable by find_package

  • string_t: Return a not allocated string for an unknown codec

  • release: Sync fpm.toml version on release, unblock docs build

[1.3.0] — 2026-10-02 ​

Added ​

  • string_t: Add six string methods from the issue #3 wishlist

Documentation ​

  • Rebuild README and site on compiled examples, fix I/O bugs

Fixed ​

  • string_t: Prevent out-of-bounds on blank and null strings

  • fobos: Repair makecoverage rule and adopt double-dash fobis CLI

  • string_t: Correct is_real and base64 decode, extend slice/strip

[1.2.0] — 2026-10-02 ​

Fixed ​

  • docs: Untrack package-lock and pin esbuild for lock-free vite build

  • string_t: Guard R16P generic bindings with PENF_R16P macro

[1.1.6] — 2026-02-18 ​

Fixed ​

  • docs: Correct project name casing to StringiFor throughout

[1.1.5] — 2026-02-18 ​

Documentation ​

  • Add VitePress site, rework CI/CD, and refactor README

[1.1.4] — 2026-02-18 ​

Fixed ​

  • Fix GNU gfortran issue with recursive procedure