Hello world, mapneat!

4 minute read

One of the most common scenarios I had to solve during my Software Engineering career is manipulating data, applying some logic on top, eventually changing the way the data is structured, and then serializing it to JSON to be consumed by other systems.

mapneat does precisely this; it provides an easy to use DSL for transforming JSON to JSON, XML to JSON, POJO(s) to JSON in a declarative way. The library is written in Kotlin (JVM), but it can be easily used from Java.

Under the hood, mapneat uses:

The library is available in the JCenter repo.

But, let’s jump directly into the code:

A simple example

Let’s start with the following class structure:

class User(
    val id: Long,
    val firstName: String,
    val lastName: String,
    val birthDate: String,
    val friends: MutableList<Long>,
    val visits: Set<Visit>,
    val creditCards: Set<CreditCardInfo>,
    val pwd: String,
    val email: String,
    val userName: String
)

class Visit(
    val id: Long,
    val country: String,
    val enter: String
)

class CreditCardInfo(
    val number: String,
    val cvv: String,
    val expirationDate: String
)

This hierarchy describes a possible system where User(s) have associated to them financial information (val creditCards: Set<CreditCardInfo>), and all the countries they visited so far (val visits: Set<Visit>). Additionally, User(s) can also befriend other users (val friends: MutableList<Long>).

Now, let’s say we are writing a REST API, and we want to allow others/other systems to query our piece of software.

Of course, we won’t want to share any kind of secret information (passwords), or any financial information.

But, if we were to serialize the User directly into a JSON format, a possible output will look like this:

{
  "id" : 490,
  "firstName" : "Rhea",
  "lastName" : "Orlich",
  "birthDate" : "1925-07-09",
  "friends" : [ 380, 290, 470 ],
  "visits" : [ {
    "id" : 25700,
    "country" : "Saint Lucia",
    "enter" : "1970-08-14"
  }, {
    "id" : 25500,
    "country" : "Lesotho",
    "enter" : "2006-10-18"
  }, {
    "id" : 25800,
    "country" : "Kenya",
    "enter" : "1972-01-01"
  }, {
    "id" : 25600,
    "country" : "Malawi",
    "enter" : "2014-06-30"
  }, {
    "id" : 25900,
    "country" : "Falkland Islands (malvinas)",
    "enter" : "2016-08-17"
  } ],
  "creditCards" : [ {
    "number" : "4158272054910622",
    "cvv" : "077",
    "expirationDate" : "2035-11-11"
  }, {
    "number" : "373725891728967",
    "cvv" : "862",
    "expirationDate" : "2021-04-21"
  } ],
  "pwd" : "genomes",
  "email" : "crinedbotfly@msn.com",
  "userName" : "fainleif"
}

Here intervenes mapneat.

The library can work directly on top of the object and morph it into a (JSON) format of our choosing.

For example, instead of showing the User in the format above, let’s say we want to:

  • Remove the pwd field and everything related to creditCards. We don’t want the world to know about them;
  • Show only the visited countries, without the visiting date or the id;
    • In case the user has visited a country multiple times, we show it once;
  • We want to uppercase the lastName of the user;
  • Instead of showing the id of his friends, we want to retrieve their full name.

The equivalent mapneat transformation that performs what we’ve listed above is the following:

val users : Map<Long, User> = getUsers(100)
val aRandomUser = users[10]

val out = json(fromObject(aRandomUser)) {
    "" *= "$"
    - "visits"
    - "creditCards"
    - "pwd"
    "visited" *= {
        expression = "$.visits[*].country"
        processor = {
            val result = HashSet<String>()
            result.addAll(it as LinkedList<String>)
            result
        }
    }
    "lastName" /= { targetCtx().read<String>("$.lastName").toUpperCase() }
    "friends" /= {
        targetCtx()
            .read<ArrayList<Long>>("$.friends")
            .map { (users[it]?.firstName + " " + users[it]?.lastName) }
            .toList()
    }
}
println(out)

If we were to run the code above, the “morphed” output is:

{
  "id" : 490,
  "firstName" : "Rhea",
  "lastName" : "ORLICH",
  "birthDate" : "1925-07-09",
  "friends" : [ "Bunny Winstanley", "Dianna Imaizumi", "Verdell Aguillar" ],
  "email" : "crinedbotfly@msn.com",
  "userName" : "fainleif",
  "visited" : [ "Saint Lucia", "Falkland Islands (malvinas)", "Malawi", "Kenya", "Lesotho" ]
}

A simple example (Explanation)

A mapneat Transformation has always the format, and wraps a series of Operations:

json(source = /* ... */) {
  // operation1
  // operation2
  // ...  
} 

The first operation we’ve performed was: "" *= "$".

*= is called the Shift Operation and allows us to query an object/json/xml using a json-path notation.

The line expresses: Copy everything ($) you find in the aRandomUser source and put it inside the target.

The next three lines (Operations) are discarding the information we don’t really need:

- "visits"
- "creditCards"
- "pwd"

- is called the Delete Operation and does exactly what it says; it deletes a node and all of its children.

The next operation is also a shift (*=):

"visited" *= {
    expression = "$.visits[*].country"
    processor = {
        val result = HashSet<String>()
        result.addAll(it as LinkedList<String>)
        result
    }
}

We query the source using a json-path expression, and from the visits object we select only the country names ("$.visits[*].country").

To avoid possible duplications, we keep the results into a Set<String>.

The next operation:

 "lastName" /= { targetCtx().read<String>("$.lastName").toUpperCase() }

Is called an Assign Operation that normally allows us to retrieve information from a lambda expression.

In our case, the lastName information is already in the target (remember "" *= "$"), so we read it from the target context, and then we just capitalize the String.

And the last operation is an Assign (as we already have the friends ids in the target context) that returns the User’s friends:

"friends" /= {
    targetCtx()
        .read<ArrayList<Long>>("$.friends")
        .map { (users[it]?.firstName + " " + users[it]?.lastName) }
        .toList()
}

Moving forward

mapneat already offers extensive documentation, and that details every supported Operation (there are more!).

If this article was an interesting read, also check the existing examples.

Updated:

Comments