Here I see three questions. Firstly, the comparison between the purchase order and the warehouse order. The data does not match exactly, but if you accept a one-to-one relationship between the purchase order and the delivery order, then this can be overcome. I assume that you can hardcode data such as warehouse ID, hazardous material codes, etc.
The second problem is whether you need a one-time or random conversion that you do manually, or a translation of the run time that you set up once, and integrate it into the translation system?
The third is based on which tool to use? Anyway, I would take a good look at Stylus Studio. It has both an IDE for local testing, as well as a manual conversion and a run-time component that you install on the server. My former company used it extensively, and it's cheaper than almost any other real-time translator (Mercator, Gentran, etc.).
To use Stylus, you import the source file (850) and create a reusable schema that can be used to parse the file in XML. Then import the sample output format (940) to create a diagram that can be used to serialize the document back to X12.
From there, you can use the matching tools to close you, and then manually edit XSLT for complex logic that you might need to configure (for example, loops are not processed perfectly). Stylus Studio even allows you to create "pipelines" that can be used to chain multiple XSLT cards in a row or pull data from external sources in the middle of a translation.
Then you can wrap it all up and export it to a real-time environment. Even better, if you do it manually, you can simply run your files on the pipeline as you receive them, and you're done.
source share