Gremlin Snippets are typically short and fun dissections of some aspect of the Gremlin language. For a full list of all steps in the Gremlin language see the Reference Documentation of Apache TinkerPop™. This snippet is based on Gremlin 3.4.6.This snippet demonstrates its lesson using the data of the "modern" toy graph (image).Please consider bringing any discussion or questions about this snippet to Discord or the Gremlin Users Mailing List.

Many advanced and even intermediate-level Gremlin examples are peppered with the use of the fold() step and the related unfold() step. As a reminder, fold() gathers all elements in the stream to that point and reduces them to a List and unfold() does the opposite, taking a List and unrolling it to its individual items and placing each back in the stream:

gremlin> g.V().has('person','name','marko').out().fold()
gremlin> g.V().has('person','name','marko').out().fold().unfold()

While the above sort of usage tends to present an obvious example, we typically see fold() tucked away in by() modulators where its usage appears more enigmatic:

gremlin> g.V().has('person','name','marko').
......1>   project('name','knows').
......2>     by('name').
......3>     by(out('knows').values('name').fold())

The above bit of Gremlin takes the “marko” vertex and converts it to a Map with project() where that Map will have two keys: “name” and “knows”. The value of the key is determined by the order oriented by() modulators that follow. The first grabs the “name” property value from the “marko” vertex and the second by() modulator executes a traversal using the “marko” vertex as the starting point. Specifically, it traverses outgoing “knows” edges and gets the “name” value from the adjacent vertices. Finally, it uses fold() to reduce that stream of “name” values to a List that becomes the value supplied to the “knows” key in the project()-step’s returned Map.

While this is the correct way to write this sort of traversal, the immediate inclination is to assume that the above traversal could be written as:

gremlin> g.V().has('person','name','marko').
......1>   project('name','knows').
......2>     by('name').
......3>     by(out('knows').values('name'))

We can see that without fold() the by() modulator only gathers the first item in the stream to add to the “knows” key which for our purposes is not the desired outcome. It is important to remember that by() behaves just like map() in the sense that it is essentially treats the child traversal as a single transformative function call (filter() behaves in the same fashion and therefore Gremlin is consistent in this sort of semantics). Another way to think about these semantics is to envision by() as only calling next() on the traversal one time to get its result.

gremlin> g.V().has('person','name','marko').map(out().values('name'))
gremlin> g.V().has('person','name','marko').map(out().values('name').fold())
gremlin> g.V().has('person','name','marko').flatMap(out().values('name'))

Given this behavior, it is important to include some form of reducing step to by() unless of course, there is only a desire for the first stream result which is sometimes the case. The fold() step is a most typical reducer, but count(), sum(), etc. all exhaust the stream into a single value and therefore also make for good examples of these semantics.