I ended yesterday’s post asking whether or not there were any real-life examples that would bring my filter brainteaser to life. After pondering this question for a while, I came up with a clear example from a real customer.

The customer is a transportation agency that is building a mass-transit system. They maintain in TrackVia a mailing list of citizens to whom they communicate about the project timeline, the design of certain train stations, and so on. If the plans for a particular train station change, they’ll send an email update to an appropriate subset of their mailing list. Here, the “appropriate subset” is defined as follows:

1) The recipient must still be “opted in,” that is, they haven’t clicked “unsubscribe” on any previous newsletter or update.
2) If they have previously expressed interest in this particular station, they should get the update regardless of where they live.
3) Otherwise, if they live in one of the three zip codes served by this station, and they live close to the train tracks, they should get the update.

So, expressed as a TrackVia filter, the mailing list for this update would be defined like this:

opted-in AND (interested-in-this-station OR (close-to-tracks AND (zip=X OR zip=Y OR zip=Z)))

This would seem to be impossible to build in TrackVia’s online database, because it requires three layers of nesting: the zip code condition is a group inside a group inside a group.

While pondering this, and trying to decide if we should complicate our filter tool by allowing more layers of groupings, I came across a coworker running a report of recently added TrackVia customers. He appeared to be using a complex filter, so I asked to see it. I was surprised to see that he’d repeated one of his conditions inside more than one group. He explained that he’d done it to circumvent the single-nesting restriction. His technique, applied to our mass transit problem, would produce this:

opt-in AND (interested-in-this-station OR close-to-tracks) AND (interested-in-station OR zip=X OR zip=Y OR zip=Z)

Or, visually:

You can see how this will work: being interested in a station does double duty, making both groups above true. If someone is opted in and interested in the station, that’s sufficient to satisfy the filter and be in the mailing list. If they’re opted in but not interested in the station, then to get the two groups to be true, they need to be close to the tracks and have one of the three zip codes. My coworker had discovered this trick of repeating the interested-in-this-station condition inside two different groups.

Mathematically speaking, the trick relies on the fact that a filter like this:

A and (B or (C and (D or E or F)))

is equivalent to this:

A and (B or C) and (B or D or E or F).

The fact that this is possible is a good argument in favor of keeping our filter tool simple. (And, I suppose, for keeping our free customer support just a phone call away – so we can help customers with tips like this.).

This may only be one real-world example, but I can assure you that any condition can be represented in TrackVia’s filter tool. There is no filter too complex for TrackVia’s single-layer mechanism to represent. Still have doubts? Read a continuation of this post in the TrackVia Community.