Text to number reads Arabic, Persian, Devanagari and Thai digits, but not fullwidth digits

FN-VALUE-001: Text to number reads Arabic, Persian, Devanagari and Thai digits, but not fullwidth digits

Quirk · measured · recorded 2026-11-04

In the recorded en-ZA settings, converting text to a number accepts Arabic-Indic (U+0660 to U+0669), extended Arabic-Indic (U+06F0 to U+06F9), Devanagari (U+0966 to U+096F) and Thai (U+0E50 to U+0E59) digits. They work alone, in runs, with a decimal point, in the exponent, with a sign or percent, and mixed with ASCII digits or with each other: VALUE("1٢3") is 123, and SQRT of a four-digit string using four scripts is the root of 1234. Fullwidth digits (U+FF10 to U+FF19) give #VALUE! alone or mixed. DATEVALUE reads the four accepted scripts in a date written year-month-day.

Basis: 66 recorded VALUE readings (all ten digits of each of the five scripts, then composed forms), 10 SQRT and 4 DATEVALUE readings.

Inferred: That other text-to-number routes, such as arithmetic operators, treat digits the same way is inferred from VALUE and SQRT agreeing.

Open: Other digit blocks (Bengali, Tibetan, Khmer and so on) have not been read.

Open: Whether the set depends on the Windows or workbook locale has not been read.