String length calculator (characters and bytes)

How is the length of a string counted?

A string has several lengths, depending on the unit. The characters you see are grapheme clusters, defined by Unicode's text segmentation rules (UAX #29) and counted here with the browser's Intl.Segmenter. Underneath, each one is one or more Unicode code points: an accented letter can be one code point or two, and the three-person family emoji is five code points (three people and two zero-width joiners) shown as one picture.

Code points are then stored as code units. JavaScript strings are UTF-16, so .length counts 16-bit units, and a code point above U+FFFF, such as the waving hand emoji U+1F44B, takes two of them, a surrogate pair. UTF-8, the encoding TextEncoder produces and MySQL's utf8mb4 stores, uses 1 to 4 bytes per code point (RFC 3629): 1 for ASCII, 2 for é, 3 for ₹ and for Hindi and Tamil letters, and 4 for anything above U+FFFF.

Words come from the same segmenter's word boundaries, which also work for languages written without spaces; punctuation and emoji don't count as words. Lines are counted from line breaks, with a break at the very end not starting a new line.

SMS uses its own count. 3GPP TS 23.038 says a message in the GSM 7-bit default alphabet can hold up to 160 characters, and one in UCS2 up to 70; a message with even one character outside that alphabet, such as ₹, an emoji or a Hindi letter, needs UCS2 instead. UCS2 stops at U+FFFF, so this page counts a character above it, such as an emoji, as two 16-bit units, the way UTF-16 writes it; check with your SMS provider for those. The extension table characters € [ ] { } \ ^ ~ | each take two of the 160. TS 23.040 sets 153 and 67 per part when a message is split, because each part carries a header for reassembly.

Segmentation follows the Unicode version of the browser. Unicode 15.1 added a rule (GB9c) that keeps Devanagari conjuncts such as स्ते together, so नमस्ते is 3 characters on current systems and 4 with older Unicode data. TS 23.038 also defines national language tables for Hindi, Tamil and other languages, but says locking-shift encoding isn't intended to be used until a national regulator asks for it, so this page counts with the default alphabet.

characters you see  →  Intl.Segmenter(granularity: "grapheme")
code points         →  [...text].length
UTF-16 code units   →  text.length
UTF-8 bytes         →  new TextEncoder().encode(text).length
SMS: GSM 7-bit 160, or 153 per part · UCS-2 70, or 67 per part