Is there a design pattern for dynamically applying transformations or filters to a model?

Question

For example, there is a model of an event with certain attributes that can be adjusted by existing processes: maybe one that looks at a geo-location stamp and determines a nearby city; another that looks at the number of registrations and applies rounding; another that calculates local popularity based on attendees vs. nearby events.

The important piece is having a way to dynamically order the operations--the filters/transformations/processing tasks--so that you can pass the events to a processing object that arranges the tasks in the correct order for the configuration/situation, and then passes the models through to apply them.

I can think of some ugly ways to make this happen, but want to know if there is a SOLID pattern that would allow for better, more-testable code.

Answer 1

Pipe & Filter is what you're looking for.

http://msdn.microsoft.com/en-us/library/dn568100.aspx

Basically, you define a standard input object and a standard output object. In object oriented programmation, you can easily define those as interfaces. The filters take the standard input object, apply some transformations and returns a standard output object. Because every filters takes the same input and output, you can chain them together however you like. You can easily produce parallel pipes to take advantage of multithreaded processors.

You could also create a configuration object to define flows of operations; your solution would then be reusable.

Answer 2

I have seen similar things done using the Chain of Responsibility pattern. Basically, each transformation/filter is a single command, and your configuration specifies the order in which they execute. The result of each transformation command can be passed to the next along with a command context to contain other metadata that should not be in the data being operated on.