Convert Bijoy to Unicode

What is converted and what is left alone

Only text that is in a Bijoy font is converted — the run's own font, the font of its character or paragraph style, or the document default, in that order of precedence. That is how a document that mixes an English heading in Arial with Bangla paragraphs in SutonnyMJ comes back with the heading untouched and the paragraphs in Unicode. The converted runs are set in Nirmala UI, which covers Bangla and Latin and ships with Windows, so the file opens as Bangla without installing anything; change it to any Unicode Bangla font afterwards if you prefer another. Word keeps a separate size and weight for complex-script text, and a Bijoy run has only the ordinary ones, so those are copied across — a 14-point document stays 14-point rather than falling back to the complex-script default.

Word breaks a run wherever the spelling checker left a mark or the formatting changed, which often puts the two halves of one word in two runs. Converting the halves separately would leave a vowel sign on the wrong side of the break, so neighbouring runs of the same Bijoy font are converted as one stretch and the result goes into the first of them. The only visible consequence is that a word that changed from bold to regular halfway through comes out in the first run's formatting. In Excel the same text is shared between every cell that shows it, so a string shown through SutonnyMJ in one cell and Calibri in another is split in two, and only the SutonnyMJ cells get the converted copy.

A plain text file is read as UTF-8 and, when it is not valid UTF-8 — a Bijoy .txt saved by Notepad is Windows-1252 — as Windows-1252, which is what gives "‡" its byte. Every line is converted and the result is saved as UTF-8. Nothing else in a Word or Excel file is rewritten: images, styles, numbering, comments, custom XML and macros are copied byte for byte, and a document that was digitally signed will show its signature as broken, because the content has changed.

When you want something else

A file that mixes Bijoy and English inside the same font is the one case this cannot judge. If someone typed an English sentence while SutonnyMJ was selected, the file holds Latin letters either way, and "Dhaka" in SutonnyMJ really does draw as Bangla — the converter has to treat every letter in that font as Bijoy, so the English word becomes a Bangla one too. Convert, then read for English that came out as nonsense; there is no automatic test that separates the two.

For hundreds of documents, a script is the right shape of tool. The same open-source conversion tables this page uses are available as libraries, and a loop over a folder with one of them, writing a converted copy beside each original, is twenty lines in Python or Node and runs unattended. This page is for the document in front of you.