src/site/php/mb ascii.php

cfw_mb_ascii()

cfw_mb_ascii(mixed  $s) : bool

True when no byte is >= 0x80, so the single-byte string functions are exact.

Parameters

Returns

cfw_mb_final_sigma()

cfw_mb_final_sigma(string  $lo) : string|null

Applies Unicode SpecialCasing final-sigma to an ALREADY-lowercased string.

Not applied to mb_strtoupper: there is no uppercase counterpart and running it there would corrupt correct output. The strpos() guard is the common case -- a string with no sigma in it cannot need this.

Parameters

Returns

The text with final sigmas applied, or NULL when the pattern failed.

cfw_mb_utf8()

cfw_mb_utf8(string|null  $encoding) : bool

The case tables are keyed by utf-8, so a caller naming another encoding must not reach them.

Parameters

Returns

cfw_mb_title_regexp()

cfw_mb_title_regexp() : string|null

The word-boundary pattern for titlecasing, built from the polyfill's Case_Ignorable data.

The 6,201-byte character class is READ from the package rather than copied, so upstream keeps owning the Unicode half; the two-line grammar around it is replaced, because upstream's lookbehind-then-two-letter-groups is wrong in both directions. It blocks a match on the letter after an ignorable and then RE-ENTERS on the letter after that, so "abc<SHY>def" comes out "Abc<SHY>dEf" where mbstring gives "Abc<SHY>def". Consuming ignorables in the word TAIL fixes it: a character inside a match is never offered a second start.

null means the package moved and every caller falls back to the polyfill's own title path.

Returns

The pattern, or NULL when the polyfill's data file is not where it was.

cfw_mb_title_word()

cfw_mb_title_word(array  $m) : string

One word: its first character titlecased, the rest lowercased.

The polyfill UPPERCASES the first character, which is a different operation for the 31 Lt digraphs and for every ligature that expands (native titlecases U+FB01 to "Fi", the polyfill uppercases it to "FI").

Parameters

Returns

cfw_mb_encode_subst()

cfw_mb_encode_subst(string  $s, string  $to) : string

Replaces every character the TARGET encoding cannot represent with "?", which is what native mb_convert_encoding does and what the polyfill does not.

The polyfill converts with //IGNORE, so an unmappable character is DROPPED: a Cyrillic sentence converted to ISO-8859-1 comes back as its spaces. One whole-string probe decides -- iconv without //IGNORE answers false only when something is unmappable -- so the per-character loop is paid by the strings that actually need it.

Parameters

Returns

cfw_mb_iconv_label()

cfw_mb_iconv_label(mixed  $enc) : mixed

Spells a charset the way symfony/polyfill-iconv's alias table spells it.

THE CHARMAPS ARE PRESENT AND THE NAMES ARE NOT. Measured against the polyfill rather than against the real extension: SJIS, GBK, BIG5 and EUC-KR are all REFUSED outright, while Shift_JIS, CP936, CP950 and CP949 decode -- out of the same 55 from.*.php files. So the gap that looked like missing capability is six alias entries.

CP950 and CP949 are SUPERSETS of Big5 and EUC-KR, so those two decode every assigned byte correctly and additionally decode bytes the narrower charset leaves unassigned, where mbstring substitutes. That is a real difference and it is why they are named here rather than presented as exact. EUC-JP and ISO-2022-JP ship no charmap at all and stay refused.

Parameters

Returns

The polyfill's spelling, or the input when it is not a string or needs no alias.

cfw_mb_decode_subst()

cfw_mb_decode_subst(string  $s, string  $from) : ?string

The same substitution in the DECODE direction: a source byte the charmap does not know becomes "?" instead of vanishing.

Walks the source the way the polyfill's own mapToUtf8 does -- a two-byte key first, then a one-byte key -- so it is right for SJIS and Big5 as well as for the single-byte charsets. The pair was used exactly when its answer is not the two singles concatenated; where those agree, either choice produces the same bytes.

Returns null when it does not apply, which leaves the polyfill's answer alone: an unknown charset (no charmap ships for EUC-JP) and a source with nothing unmappable both take that path, and the second is the common one.

Parameters

Returns

cfw_mb_isubpart()

cfw_mb_isubpart(string  $haystack, int|false  $pos, bool  $before, string|null  $encoding) : string|false

Case-insensitive search on the ORIGINAL haystack.

The polyfill returns a slice of its own lowercased copy, so a haystack containing a character that lowercases to two comes back with different bytes than went in.

Parameters

Returns