Is there any way to automatically apply naming conventions to Moshi when serializing data using Moshi?
e.g.
I have this class which follows Kotlin naming conventions for properties (camel case)
class Player {
var currentHealth: Int = 100
var firstName = "John"
var lastName = "Doe"
}
I would like to serialize it to something which will stick to the naming conventions we have on the server (snake case):
{
current_health: 100,
first_name: "John",
last_name: "Doe"
}
I already know we can customize column names on each property manually. Just looking for a way to do that automatically (for instance if we add properties to the POJO or on server side).
I'm one of the Moshi developers and it doesn't have this feature because I think it is harmful. I've described why in full detail here. In brief, case mapping breaks your ability to search across your codebase.
Related
While writing code for RecyclerView to get data I figured out there's a data class in Kotlin.
Following codes are taken from two different projects which are linked above.
#Serializable
data class MarsPhoto(
val id: String,
#SerialName(value = "img_src")
val imgSrc: String
)
class Contacts {
#SerializedName("country")
private val country:String? = null
fun getCountry():String?{
return country
}
}
I know that both classes are doing same job. So what does differentiate them? I also wonder in the MarsPhoto data class how they can get the id without declaring SerialName just the way they did for imgSrc. (I am just on the way to learning kotlin now, so I'm absolute beginner).
Basically for "data" class the compiler automatically derives the following members from all properties declared in the primary constructor:
equals()/hashCode() pair
toString() of the form "MarsPhoto(id=1, imgSrc=asdf)"
componentN() functions corresponding to the properties in their order of declaration.
copy()
You can read a lot more at enter link description here
On the SerializedName part of your question. if you are dealing with Gson lib by default it is using fields name as "SerializedName". And only if you want to use something different then field name, you can use SerializedName annotation and pass your custom value there. But usually, everybody just writes #SerializedName() with duplication of field names as value for every field.
It's a good idea if you are receiving and Serializing data from server from Json. Because Backend developers can use a bad keys in response, which you don't want to use in your code, so #SerializedName will be the only place where you will have to see this key, and you can name your fields however you like.
#Serializable used to mark class as serializable to disk or like into a file( alternative is Parcel able in android) special useful in case of process death or configuration changes and #SerializedName("country") used for json parsing when u receive the response from server
You get the id without #SerializedName because the JSON property field is the same as your variable name, but imgSrc and img_src is not. Still, even if they are the same, you should always use #SerializedName, because your variable names could be converted to random letters during code optimization, and obfuscation.
I have an app which is mixed Java and Kotlin.
In the Kotlin code I use Moshi to convert an object to Json in a convertor for a Room database table.
I have one case that works perfectly but another one produces the error:
Not enough information to infer type variable T
This is what my code looks like:
val type: Type = Types.newParameterizedType(
MutableMap::class.java,
LayerTwoConn::class.java,
TWeFiState::class.java,
WfMeasureFileMgr::class.java,
Traffic::class.java,
ThroughputCalculator::class.java,
CellSubTechThroughput::class.java,
LongValuesAverageCalculator::class.java,
LayerTwoConn.SenselessTraffic::class.java
)
val json = Moshi.Builder().build().adapter(type).toJson(layerTwoConn)
I have included all the classes that are used in the objects.
What have I missed?
This case works perfectly:
val type: Type = Types.newParameterizedType(
MutableList::class.java,
CnrScan::class.java,
)
val jsonAdapter: JsonAdapter<List<CnrScan>> = Moshi.Builder().build().adapter(type)
val json = jsonAdapter.toJson(list)
In this object, all the internally used classes are standard Java class and not my own.
Have I missed something simple?
I don't know if this is important but the class LayerTwoConn's constructor is private.
I think you are trying to convert too many classes into one type, try to convert MutableMap class and LayerTwoConn class.
Do note that Room uses SQL architecture, so try to predict what you want your table to contain
I would like to know what is the best way to integrate Retrofit with MoShi on my Android Studio project.
First of all, I use the moshi converter of retrofit :
implementation 'com.squareup.retrofit2:converter-moshi:2.9.0'
I use the popular plugin "Json to Kotlin class" for the generation of my POJO :
https://plugins.jetbrains.com/plugin/9960-json-to-kotlin-class-jsontokotlinclass-
When I create a new POJO, I use the annotation "MoShi (Reflect)" :
It generate me a Pojo with fields like this one :
#Json(name = "image_url")
val imageUrl: String?
The problem is Retrofit2 seem don't like underscore in my fields name because I get null results with the underscored names.
Searching in the Github issues, I found a solution. It work if I use this annotation :
#field:Json(name = "image_url")
val imageUrl: String?
So my questions are :
Using the #field:Json annotation is the best way to use MoShi with Retrofit ?
If yes, how to generate easily a Kotlin class with this annotation (using the same plugin or an other) ?
As apparent from this part of Moshi's documentation: "There’s no field naming strategy, versioning, instance creators, or long serialization policy. Instead of naming a field visibleCards and using a policy class to convert that to visible_cards, Moshi wants you to just name the field visible_cards as it appears in the JSON."
The preferred way is for you to not use the annotation, but instead name the fields the same way as in the JSON. With Kotlin you can use backticks for names that wouldn't otherwise be valid, e.g.val `image-url`: String?
Of course you wouldn't want to be working with such names in other parts of your code, and for that you should have separate objects for describing the data as it appears on the back-end, and another object for how it should appear in your application. We call those DTOs (Data-Transfer Objects.)
I'm trying to implement JSON parsing in my Android application written in Kotlin using com.squareup.moshi (v1.10.0).
Within the JSON file there are some properties that are not interesting in my case. Let's say, I only need the position to be able to mark the place on a map and the JSON looks like this:
"location":{
"address":{
"country":"..."
},
"position":{
"lat":47.469866,
"lon":19.062435
}
}
If I'm right, the data class in Kotlin should look like this if I'd like to parse that JSON:
#Parcelize
data class Location(
val address: Address,
val position: Position
): Parcelable
#Parcelize
data class Address(
val country: String
): Parcelable
#Parcelize
data class Position(
val lat: Double,
val lon: Double
): Parcelable
In Moshi's documentation I could find the transient keyword to skip values which in Kotlin works as an annotation (#Transient). As the documentation says:
Transient fields are omitted when writing JSON. When reading JSON, the field is skipped even if the JSON contains a value for the field. Instead it will get a default value.
Does it mean that if I don't want to have the address object, I should use the following code?
#Parcelize
data class Location(
#Transient val address: Address? = null,
val position: Position
): Parcelable
Also, what about in general terms? What if I have huge list of properties within a JSON object but I know I only need the 'position' object? Do I still have to create null values to parse the JSON file field-by-field?
I think you are looking for something similar to GSON's #Expose annotations, wherein all model fields are excluded from parsing except those annotated.
This functionality is currently not available in Moshi, so your current implementation using the #Transient annotation seems to be the most optimal solution. (See Moshi issues conversation here.)
Extra food for thought:
You may also wish to use #IgnoredOnParcel on your transient fields since you are implementing the parcelable interface. (Have a look here for some implementation pointers.)
Alternatively you could separate your data model into 2 models - one for use in your app and one which reflects the server (JSON) schema (just as you have done above). The main data model for your app (which could implement parcelable) would contain only the fields you use (for example, the position field). When you parse your data, you then convert that data to your primary data model using some simple adapter. (This is often good practice anyhow, since server-side schemas are inherent to change. This way, any changes in the JSON schema wouldn't end having any ripple effect throughout your code.)
https://github.com/square/moshi#omit-fields-with-transient
Omit fields with transient
Some models declare fields that shouldn’t be included in JSON. For example, suppose our blackjack hand has a total field with the sum of the cards:
public final class BlackjackHand {
private int total;
...
}
By default, all fields are emitted when encoding JSON, and all fields are accepted when decoding JSON. Prevent a field from being included by adding Java’s transient keyword:
public final class BlackjackHand {
private transient int total;
...
}
Transient fields are omitted when writing JSON. When reading JSON, the field is skipped even if the JSON contains a value for the field. Instead it will get a default value.
I'm using Kotlin objects to work with my Firebase Database models, as described in the guide. I have many fields that are stored as strings, but really are enums, so to be type-safe I have enum fields in the models, plus a string delegated property that returns the firebase stored value (as suggested in a question I asked some time ago). Now, these fields work if I get/set the string delegate in code, but firebase libs seem to skip them when converting to/from database's json format.
A simple example:
abstract class BaseModel {
#Exclude
open var path: String? = null // fails even if I delete this field!
}
class Weight() : BaseModel() {
constructor(v: Double, u: WeightUnit) : this() {
value = v
unitEnum = u
}
var value: Double = 0.0
#Exclude
var unitEnum: WeightUnit = WeightUnit.KG
var unit: String by EnumStringLowercaseConverter(WeightUnit::class.java).getDelegate(Weight::unitEnum)
}
[...]
val testWeight = Weight(7.0, "kg")
db.getReference("/valid/path/to/save/testWeight").setValue(testWeight)
.addOnSuccessListener { r -> Log.d(LOG_TAG, "set successful") }
.addOnFailureListener { e -> Log.e(LOG_TAG, "set error", e) }
The setValue always gives a Permission Denied error, but works, if I delete unitEnum field and make unit a normal String property.
It's similar for reading: Firebase gives no errors when getting a Weight object, but the weightUnit field is never set to anything else than the default. But, if I manually do weight.unit = "lb", the unitEnum field properly returns WeightUnit.LB.
I'm using firebase libs v10.0.1
Now, the questions:
What can I do to make the delegated properties work correctly with firebase? I can try a different approach to the delegated enum fields, as long as the points from my original question are satisfied (readable, concise and type-safe code).
is there any way to see how exactly do firebase libs convert objects to/from json? Or at least see the converted json? Maybe then I could tweak things myself. Unfortunately, everything firebase-related shows as /* compiled code */ in AndroidStudio.
UPDATE: I could of course add a toMap() method to each model, where I would construct a map containing all the properties needed in firebase, but it would be tiresome to do this for every model, and it solves the saving issue only, the enum fields still wouldn't be set when getting.
The delegated props are also skipped when serializing with GSON. So maybe is there a generic way to make the delegated properties look like regular fields?
Try this code, it should work.
#get:Exclude #set:Exclude
var unitEnum: WeightUnit = WeightUnit.KG
var unit: String
get() = unitEnum.name
set(v) { unitEnum = WeightUnit.valueOf(v) }