Two separate faults, both visible on opening a DOCX with Hebrew or Arabic.
1. Alignment lands on the wrong side
DocxSemanticBackend writes the resolved alignment physically:
<w:pPr><w:bidi/><w:jc w:val="right"/></w:pPr>
Word reads w:jc's left/right as start/end relative to the paragraph direction, so in a w:bidi paragraph right means the end of the flow — the left edge. A right-to-left paragraph that the page laid out flush right is drawn flush left.
Fix: map alignment through the paragraph direction before writing it — a right-to-left paragraph aligned to its natural start must be written left, and an explicitly left-aligned one right.
2. Complex-script size and weight are never written
A 15pt Hebrew paragraph reaches Word as:
<w:rFonts w:ascii="David Libre" w:hAnsi="David Libre" w:cs="David Libre" w:eastAsia="David Libre"/>
<w:sz w:val="30"/>
The font slots are right, including w:cs. But w:szCs, w:bCs and w:iCs are never written — 0 occurrences in the output and no code for them in the backend — and the export ships no styles.xml, so there are no docDefaults to fall back to either. Word takes size, bold and italic for complex-script characters from w:szCs / w:bCs / w:iCs, so Hebrew and Arabic are drawn at Word's built-in default size and weight rather than the requested style, while Latin in the same run obeys w:sz.
Fix: write the complex-script twin alongside each of w:sz, w:b, w:i.
Both recorded in the capability matrix by #544.
Two separate faults, both visible on opening a DOCX with Hebrew or Arabic.
1. Alignment lands on the wrong side
DocxSemanticBackendwrites the resolved alignment physically:Word reads
w:jc'sleft/rightas start/end relative to the paragraph direction, so in aw:bidiparagraphrightmeans the end of the flow — the left edge. A right-to-left paragraph that the page laid out flush right is drawn flush left.Fix: map alignment through the paragraph direction before writing it — a right-to-left paragraph aligned to its natural start must be written
left, and an explicitly left-aligned oneright.2. Complex-script size and weight are never written
A 15pt Hebrew paragraph reaches Word as:
The font slots are right, including
w:cs. Butw:szCs,w:bCsandw:iCsare never written — 0 occurrences in the output and no code for them in the backend — and the export ships nostyles.xml, so there are nodocDefaultsto fall back to either. Word takes size, bold and italic for complex-script characters fromw:szCs/w:bCs/w:iCs, so Hebrew and Arabic are drawn at Word's built-in default size and weight rather than the requested style, while Latin in the same run obeysw:sz.Fix: write the complex-script twin alongside each of
w:sz,w:b,w:i.Both recorded in the capability matrix by #544.