Question

I have:

• Raw Object: contains raw data before importing to the system
• Module #1: Validator which will validate Raw Objects
• Module #2: Updater which will use validated raw data (invalid data were removed) to generate Real Objects

Module #1 has to determine whether ItemCode (just an example, the actual number of keys are dozens) exists in the database or not. Module #2 has to get the Item object by querying ItemCode in the database.

A developer wants to add an alien property .TempItem to the raw object so that the Module #1 will assign the Item object to .TempItem property (because it is super easy to do that) and Module #2 just needs to use .TempItem without querying the database second time.

My technical expert advises that it is not good.

I want to follow programming principles and best practices but have no idea how.

Any help is appreciated.

Answer 1

What you have with Module #1 and #2 is a pipeline process. The output of #1 passes into #2. Any information that #2 needs which #1 has already ascertained can (arguably should) be cached.

I would create an additional object to pass this state information through the pipeline (@astander suggests a dictionary, but in many cases a hard-coded object that stores precisely the data you need will be simple to write and much more usable/maintainable)

Using a “tempitem” is essentially the same mechanism, but a poor implementation of it – the idea is the same though (don’t read the same information from the database multiple times)

By passing a state object through your pipeline you keep the pipeline processors loosely coupled (#2 doesn’t need to know about #1, it simply acts on the state information that is passed in), which makes it easier to add new pipeline stages or refactor your modules in future. By passing the data within a separate class, it also makes it really easy to add new state data to be passed through the system.