When a filter sends pushes an entry, it may provides the translation for that entry (e.g. in XLIFF, TXML, etc.). If the project has the segmentation option set, that entry may get segmented and in this case the translation is lost.
Maybe the segmenter could check for the presence of a translation before trying to segment, and not segment if there is a translation available.
See also this email thread for examples:
https://groups.yahoo.com/neo/groups/OmegaT/conversations/topics/36035
It is useful when translating PO files. +1
As an ordinary work flow, downloading or copying partial translated PO file from transifex, or applicaiton source and copy it into OmegaT project source.
Same issue is happen for also PO files.
We can use translate-toolkit to help OmegaT recognizing current translation result, using posegment and po2tmx. Unfortunately translate-toolkit cannot handle PO file when number of source sentence and target sentence is different.