I can not find my code, but I remember using spark.read().text("file_name") and then manipulated the file (explode etc.) to get lines in the correct order. Of course, it will be slower, and as the whole file will go to one cell, it has memory limits as it will go through a single worker. So files have to be smaller than the RAM on the worker.

There is no spark function showing a row in the source (as it splits everything per partition and works on chunks), so other solutions will not 100% guaranteed.

If the file is really big or as alternative, you need to add ID inside the file.