|
Hi Folks, Suppose you are tasked with defining a mapping between one set of data items (source data items, S) to another set of data items (destination data items, D).
Here’s what you need to do: You need to create 6 documents that show complete mappings, incomplete mappings, and so forth. You need to create a tool that auto-generates an instance of D from an instance of S. And you need to devise a method
to verify that your mappings are correct (using your eyeballs to verify the correctness of your mappings is not acceptable). Create 6 documents
1. List the elements of D that have a straightforward direct mapping to them. 2. List the elements of D that have an imperfect mapping to them. This gives a list of wonky but maybe close enough mappings. Be picky and precise about this because these will be problematic going forward. 3. List the elements of D that have no mapping to them. Create these two lists: 3.1. List the elements of D that are required and have no mapping to them. If it is required in D but not available in S, that’s a big problem.
This yields a list of data that is needed to make a valid D document but cannot be built from S data alone. Hopefully this is a very small list and hopefully an empty list. 3.2 List the elements of D that are optional and have no mapping to them. This is a smaller problem. This list is also useful, just not a big problem. 4. List the fields of S that don’t map to any D element. This should be mostly stuff that S users require to be in S but has no home in D. It may be important to know this. 5. List the elements that are not mapped because they are used solely to reference other data and has no other purpose. Include a justification for why they are not mapped. This is also important to identify. Autogenerate D from S
Record the mappings using a machine-processable document format. Why? So a tool can be implemented which uses the mapping document to automatically map S to D. The mapping document must not only state “This S data item maps to this D data
item” but it must also inform the tool on how the mapping is to be done. For example, “the first 4 characters of this S date maps to the D year, the next 2 characters maps to month, and the last 2 characters maps to day.” This might involve creating
a little mapping language that the tool interprets to carry out the mapping actions. Verify the mapping
It is important you get the mapping right, as lives and/or money may depend on it. Eyeballing the mapping and announcing, "Yea, that looks right" is not acceptable. You need a higher level of assurance that your mapping is correct. How
to verify your mapping? That depends on what you’re mapping. If you’re mapping air navigation data, then you could run a flight simulator on the S data, run it on the mapped D data, and then check that the flight results are within a certain tolerance. I welcome your comments. /Roger |