The relation analyzer (RA) tries to assemble the relation data to see if it is in one piece (either a line with just two ends, or Y-shaped, or more complicated). The pieces are assembled using rules which include looking at roles such as "forward" and "backward". If a relation cannot be connected into one single piece, the RA will create more than one.
You can see these below, including the end node IDs. The RA also calculates the distances between those end nodes and sorts the pieces by the shortest distance within the pieces. Pieces consisting of a circular path will have only one end node; pieces consisting just of nodes in a line will have two end nodes; Y-shaped pieces 3 end nodes; and so on.
If you move the mouse over the node IDs below it will highlight them so that you can see which other nodes an end node is close to.
If you see an end node in one piece very close to an end node in another, you might have detected a gap between two pieces - you can check the actual data by using one of the edit links (JOSM or Potlatch) below. Similarly, if you see two end nodes very close together within one piece, you might have detected a gap within a circular relation.
It is not possible to analyze all relations in the same way since different relation types have different meanings. The RA tries to build a linked series of ways using some general rules and then creates a rating based on the type of the relation (although not all relation types are supported just yet).
An example of a type that is supported are "type=route" relations. Here it expects that the relation exists in one piece as a series of linked ways - only in this case you will get a OK rating.
Shows the distribution of way types in this relations. Hover over a color for more details.
red = major roads, blue = rural roads, brown = tracks, green = footways and cycleways, gray = unknown