Keyboard layout fixes

Eram\Abzar\Text\KeyboardFixer swaps between English QWERTY and the standard Iranian Persian keyboard (fa-IR) layout. Use it when a user typed with the wrong layout active — e.g. pressed the keys for سلام while English was engaged and produced sghl.

Minimal example

<?php
require 'vendor/autoload.php';

use Eram\Abzar\Text\KeyboardFixer;

echo KeyboardFixer::enToFa('sghl'), "\n";
echo KeyboardFixer::faToEn('سلام'), "\n";
echo KeyboardFixer::enToFa('ldBv,l'), "\n";
var_dump(KeyboardFixer::detect('hello'));

سلام
sghl
می‌روم
bool(false)

More examples

use Eram\Abzar\Text\KeyboardFixer;

KeyboardFixer::enToFa('sghl');   // سلام
KeyboardFixer::faToEn('سلام');   // sghl
KeyboardFixer::enToFa('Hfhn');   // آباد — Shift+H is آ (ISIRI 9147 Shift layer)
KeyboardFixer::enToFa('ldBv,l'); // می‌روم — Shift+B is ZWNJ

// Optional heuristic: was this input typed with the wrong layout?
KeyboardFixer::detect('sghl');     // true
KeyboardFixer::detect('hello');    // false (normal English vowel ratio)
KeyboardFixer::detect('سلام');     // false (already Persian)

Behaviour

• enToFa() maps Latin letters and a small set of punctuation. Digits, whitespace, Persian/Arabic letters, kashida and existing ZWNJ pass through unchanged in that direction. faToEn() reverses mapped Persian characters, including the mapped ZWNJ.
• Upper-case letters follow the Shift layer of the ISIRI 9147 standard layout: H → آ, C → ژ, M → ء, B → ZWNJ, K / L → » / «, A → ؤ, S → ئ, the top row → tashkeel, and so on. faToEn() reverses it, so faToEn(enToFa('Hfhn')) === 'Hfhn'. ASCII [ / ] (Shift+P / Shift+O) are left alone by faToEn().
• Text typed with Caps Lock on is not the Shift layer. Lower-case it first if that’s your case: enToFa(strtolower($typed)). Before 0.7, enToFa() lower-cased every input.
• The layout mapping is the standard Iranian Persian keyboard. Dari, Pashto, and other regional variants are out of scope.
• detect() is a coarse character-script entropy heuristic, not grammar-aware. It rejects input containing the Persian/Arabic characters detected by Script, then counts only ASCII letters after lowercasing. At least two letters and a vowel ratio strictly below 25% yield true; digits, punctuation and other scripts are ignored in that count. It is not a general mixed-script detector. Treat it as a signal for suggesting a layout swap to the user, not for auto-applying one: consonant-heavy English tokens (brand names like chatgpt, acronyms, or short technical jargon) will false-positive. Do not call enToFa() unconditionally on detect() === true without giving users an opt-out.

Related guides

See Persian text for character normalization, digits for digit conversion, and formatting for numeric display. Layout conversion is not transliteration or spelling correction.