problem

In some use cases only a few segments are translated, and on export via Okapi the not translated segments are filled up with the source content.

But if the trans-unit contains multiple segments (<mrk mtype="seg">) separated with whitespace, Okapi recognizes that whitespace as not empty content. Therefore the original source content is not used, although no content was translated.

solution

A second XLF with source text copied into the empty targets is created and used for the Okapi communication and then deleted. The original XLF with empty targets is still delivered in the export zip.