Specter is useful for navigating through nested data structures, and then returning (selecting) or transforming what it finds. Specter can also work its magic **recursively**. Many of Specter's most important and powerful use cases in your codebase will require you to use Specter's recursive features. However, just as recursion can be difficult to grok at first, using Specter recursively can be challenging. This guide is designed to help you feel at home using Specter recursively.
Before we review how recursion works with Specter, it might be worth a brief refresher on recursion more broadly. If you're familiar with this material, feel free to skip this section. But a brief review will allow us to separate reviewing the concept of recursion from learning how to combine that concept with Specter's functionality.
Instead of calling `square`, itself, `square` calls another function, `*` or multiply. This, of course, serves the purpose of squaring the input. You don't always need recursion.
But sometimes you need a function to call itself - to recur. To take another example from mathematics, you need recursion to implement factorials. Factorials are equal to the product of every positive number equal to or less than the number. For example, the factorial of 5 ("5!") is `5 * 4 * 3 * 2 * 1 = 120`.
Here, the function definition *calls itself* with the symbol `factorial`. You still supply input - the value x - but producing the function's return value requires calling the function itself.
You can see that there are three arguments, which match onto our list above. (They are, however, in a different order.) The `self-sym` is the name; the `params` are the optional input arguments; and the `path` is the body of the recursion. This path can be recursive, referencing itself by the `self-sym`.
Rather than providing a recursive function, `recursive-path` provides a recursive path that can be composed with other navigators. Accordingly, it is often useful - but not necessary - to define a recursive-path using `def`, so that it can be called in multiple places, not just the place where you initially need it.
You can define a navigator to all the leaves of the tree like this:
```clojure
=> (def TREE-VALUES
(recursive-path [] p
(if-path vector?
[ALL p]
STAY)))
```
`recursive-path` allows your path definition to refer to itself, in this case using `p`. This path definition leverages the `if-path` navigator which uses the predicate to determine which path to continue on. If currently navigated to a vector, it recurses navigation at all elements of the vector. Otherwise, it uses `STAY` to stop traversing and finish navigation at the current point. The effect of all this is a depth-first traversal to each leaf node.
Now we can get all of the values nested within vectors:
That you can define how to get to the values you care about once and easily reuse that logic for both querying and transformation is invaluable. And, as always, the performance is near-optimal for both querying and transformation.
# Applications
Here are some other examples of using Specter recursively with `recursive-path`.
## Find the "index route" of a value within a data structure
This example comes from [a Stack Overflow question](https://stackoverflow.com/questions/45764946/how-to-find-indexes-in-deeply-nested-data-structurevectors-and-lists-in-clojur).
