Develop Reference

How do I create a new record in Android Room database?

This seems like a lazy question, but the offical docs literally don't explain how to do this.
https://developer.android.com/training/data-storage/room
I have an entity like so:
#Entity
data class Shader(
#PrimaryKey(autoGenerate = true) val uid: Int,
val name: String,
val shaderMainText: String,
val paramsJson: String?
)
And some methods on my Dao:
#Insert
fun insertAll(vararg shaders: Shader)
#Delete
fun delete(shader: Shader)
#Update
fun update(shader: Shader)
How do I make a new record and insert it?
I've gotten so far as this:
val record = Shader(name="Foo", shaderMainText="bar", paramsJson=null)
But it complains that I'm missing the argument uid. Why do I need to provide a uid? It's supposed to be auto-generated?
An example of creating a new record would be appreciated.

You can set 0 for the default value of your uid field
#Entity
data class Shader(
#PrimaryKey(autoGenerate = true) val uid: Int = 0,
val name: String,
val shaderMainText: String,
val paramsJson: String?
)
When autoGenerate is true, Room ignores the default value, so it creates auto-incremented values.
Finally, this doesn't give you a compilation error:
val record = Shader(name="Foo", shaderMainText="bar", paramsJson=null)

The docs say:
If the field type is long or int (or its TypeConverter converts it to a long or int), Insert methods treat 0 as not-set while inserting the item.
If the field's type is Integer or Long (or its TypeConverter converts it to an Integer or a Long), Insert methods treat null as not-set while inserting the item.
So I would try setting the uid to be initialized to 0 or make it nullable and set to null.
Hope that helps!
https://developer.android.com/reference/android/arch/persistence/room/PrimaryKey#autoGenerate()

You Can Also try like this
The best practice is you always create an instance Class
1st Create a Database instance class
#Database(entities = [Shader::class], version = 1)
abstract class DatabaseHelper : RoomDatabase() {
companion object {
private var instance: DatabaseHelper? = null
fun getInstance(context: Context?): DatabaseHelper? {
if (instance == null) {
synchronized(DatabaseHelper::class) {
if (context != null) {
instance = Room.databaseBuilder(context.applicationContext, DatabaseHelper::class.java, "database.db").allowMainThreadQueries().build()
}
}
}
return instance
}
}
abstract fun getDao(): DBDao?
}
Create Your Shader data class like below
#Entity(tableName = "shader")
data class Shader(
#PrimaryKey(autoGenerate = true) var id: Int?,
val name: String,
val shaderMainText: String,
val paramsJson: String?
) {
constructor(name: String, shaderMainText: String, paramsJson: String?) : this(
null,
name,
shaderMainText,
paramsJson
)
}
Here you want to create constructor
Add data in Shader and insertAll Like Below
val record = Shader(name="Foo", shaderMainText="bar", paramsJson=null)
DatabaseHelper.getInstance(this)?.getDao()?.insertAll(record)
And Your Data added is successfully

There should be always a #PrimaryKey(autoGenerate = false) while inserting data in the DB

How can I add data into Room database in a manner where it is appended to the top row

Image for reference
I am facing a situation whenever I add a new article into the database it will add the new article to the bottom of the RecyclerView, how can I implement it in such a way that it updates the top row instead?
Is it something got to do with the Adapter or do I have to tweak the Doa?
Room Database Implementation
**//ViewModel**
fun saveArticle(article: Article) = viewModelScope.launch {
newsRepository.upsert(article)
}
**
//NewsRepository**
suspend fun upsert(article: Article) = db.getArticleDao().upsert(article)
**//Implementation of the Database**
#Database(
entities = [Article::class],
version = 1
)
#TypeConverters(Converters::class)
abstract class ArticleDatabase : RoomDatabase() {
abstract fun getArticleDao(): ArticleDao
companion object {
#Volatile
private var instance: ArticleDatabase? = null
private val LOCK = Any()
operator fun invoke(context: Context) = instance ?: synchronized(LOCK) {
instance ?: createDatabase(context).also { instance = it }
}
private fun createDatabase(context: Context) =
Room.databaseBuilder(
context.applicationContext,
ArticleDatabase::class.java,
"article_db.db"
).build()
}
}
**//DOA**
#Insert(onConflict = OnConflictStrategy.REPLACE)
suspend fun upsert(article: Article): Long //returns the id's of the article
**//Entity**
#Entity(
tableName = "articles", indices = [Index(value = ["url","title"], unique = true)]
)
#Parcelize
data class Article(
#PrimaryKey(autoGenerate = true)
var id: Int? =null,
val author: String?,
val description: String?,
val source: Source?,
val title: String?,
val url: String?,
val urlToImage: String?,
val publishedAt: String?,
val content: String?
): Parcelable {
override fun hashCode(): Int {
var result = id.hashCode()
if(url.isNullOrEmpty()){
result = 31 * result + url.hashCode()
}
return result
}
}
Expecting the top row to be updated instead of the bottom row, the Room Database docs did not talk much about changing the way the article is inserted, tried to reference other people's work, couldn't find the solution

Problem in TypeConverters In Room Database

I am trying to use type converters in Android (Kotlin) so i am using the type converters class but i am getting confused like inside of the clouds i am having a single variable so i have returned it but
#Entity(tableName = "WeatherDb")
data class WeatherDTO(
val base: String,
val clouds: Clouds,
val cod: Int,
val coord: Coord,
val dt: Int,
#PrimaryKey(autoGenerate = true)
val id: Int,
val main: Main,
val name: String,
val sys: Sys,
val timezone: Int,
val visibility: Int,
val weather: List<Weather>,
val wind: Wind
)
class TypeConverters {
#TypeConverter
fun fromCloudsToDouble(clouds: Clouds): Int {
return clouds.all
}
fun fromCoordToDouble(coord: Coord): Double {
}
}
In coord class here are multiple with different datatypes how to covert this?
data class Main(
val feels_like: Double,
val grnd_level: Int,
val humidity: Int,
val pressure: Int,
val sea_level: Int,
val temp: Double,
val temp_max: Double,
val temp_min: Double
)
Clouds.kt
data class Clouds(
val all: Int
)
Coord.kt
data class Coord(
val lat: Double,
val lon: Double
)
Main.kt
data class Main(
val feels_like: Double,
val grnd_level: Int,
val humidity: Int,
val pressure: Int,
val sea_level: Int,
val temp: Double,
val temp_max: Double,
val temp_min: Double
)
Sys.kt
data class Sys(
val country: String,
val id: Int,
val sunrise: Int,
val sunset: Int,
val type: Int
)
Weather.kt
data class Weather(
val description: String,
val icon: String,
val id: Int,
val main: String
)
Wind.kt
data class Wind(
val deg: Int,
val gust: Double,
val speed: Double
)
WeatherViewModel.kt
#HiltViewModel
class WeatherViewModel #Inject constructor(
private val repo:WeatherRepository,
private val application: Application,
private val WeatherDb:WeatherDB,
private val fusedLocationProviderClient: FusedLocationProviderClient
) :ViewModel(){
private val _resp = MutableLiveData<WeatherDTO>()
val weatherResp:LiveData<WeatherDTO>
get() = _resp
private val _cord = MutableLiveData<Coord>()
val cord:LiveData<Coord>
get() = _cord
var locality:String = ""
fun getWeather(latitude:Double,longitude:Double) =
viewModelScope.launch {
repo.getWeather(latitude,longitude).let { response->
if(response.isSuccessful){
Log.d("response","${response.body()}")
WeatherDb.WeatherDao().insertWeather(response.body()!!)
_resp.postValue(response.body())
}else{
Log.d("Weather Error","getWeather Error Response: ${response.message()}")
}
}
}
fun fetchLocation():Boolean{
val task = fusedLocationProviderClient.lastLocation
if(ActivityCompat.checkSelfPermission(application,android.Manifest.permission.ACCESS_FINE_LOCATION)
!=PackageManager.PERMISSION_GRANTED &&
ActivityCompat.checkSelfPermission(application,android.Manifest.permission.ACCESS_COARSE_LOCATION)
!=PackageManager.PERMISSION_GRANTED
){
return true
}
task.addOnSuccessListener {
if(it!=null){
getWeather(it.latitude,it.longitude)
getAddressName(it.latitude,it.longitude)
Log.d("localityname", locality)
}
}
return true
}
private fun fetchLocationDetails(){
}
private fun getAddressName(lat:Double,long:Double):String{
var addressName = " "
val geoCoder = Geocoder(application, Locale.getDefault())
val address = geoCoder.getFromLocation(lat,long,1)
if (address != null) {
addressName = address[0].adminArea
}
locality = addressName
Log.d("subadmin",addressName.toString())
Log.d("Address", addressName)
return addressName
}
fun getCoordinates(cord:String){
val geocoder = Geocoder(application,Locale.getDefault())
val address = geocoder.getFromLocationName(cord,2)
val result = address?.get(0)
if (result != null) {
getWeather(result.latitude,result.longitude)
getAddressName(result.latitude,result.longitude)
}
}
}

Here is my converter in the Kotlin:
class Converters {
#TypeConverter
fun valueFromDomainToStorage(value: Value): String {
return value.convertToJson()
}
#TypeConverter
fun valueFromStorageToDomain(str: String): Value {
// we can not create an empty instance of value as TypeDecoder.java should call non-empty constructor
return Value(
"just a stub",
BigInteger.valueOf(0),
BigInteger.valueOf(0),
false,
BigInteger.valueOf(0)
)
.fromJson(str)
}
}
where .convertToJson() and .fromJson(str) implemented as extensions within Value class:
fun Value.convertToJson(): String {
val result = JSONObject()
result.put(ValueConst.OFFER_FIELD, offer)
result.put(ValueConst.AVAILABLE_SINCE, availableSince.toLong())
result.put(ValueConst.AVAILABLE_END, availabilityEnd.toLong())
result.put(ValueConst.IS_CONSUMED, isConsumed)
result.put(ValueConst.LOCKED_UNTIL, lockedUntil)
return result.toString()
}
fun Value.fromJson(json: String): Value {
val subj = JSONObject(json)
return Value(
subj.optString(ValueConst.OFFER_FIELD),
BigInteger.valueOf(subj.optLong(ValueConst.AVAILABLE_SINCE)),
BigInteger.valueOf(subj.optLong(ValueConst.AVAILABLE_END)),
subj.optBoolean(ValueConst.IS_CONSUMED),
BigInteger.valueOf(subj.optLong(ValueConst.LOCKED_UNTIL))
)
}
You should implement Converter class for each non-native class type. Do not forget to register your converters on database:
#Database(entities = [ChainTransaction::class], version = 1, exportSchema = false)
#TypeConverters(Converters::class)
abstract class AppDatabase: RoomDatabase() {
When you have compile the code and later introduce new changes, you have to increase version parameter too to make changes to take effect:
#Database(entities = [ChainTransaction::class], version = 2, exportSchema = false)
#TypeConverters(Converters::class)
abstract class AppDatabase: RoomDatabase() {
Here is official documentation and even training on this topic:
https://developer.android.com/training/data-storage/room

so i am using the type converters class but i am getting confused
SQLite (the database around which Room is an object orientated wrapper) is not an object orientated (or aware) database. It is a database that can store primitive types of data which are one of
INTEGER (such as Int or Long), REAL
REAL (such as Float or Double)
TEXT (such as String)
BLOB (such as ByteArray)
NULL
Therefore to store a type of Coord, Cloud or Weather .... you have three options:-
to embed the class, in which case the fields are copied from the embedded class (would be complicated if the embedded classes contained unsupported types). not covered in the answer
to have the class as a table in it's own right with a relationship between it and the parent (WeatherDTO). not covered in the answer
to convert the class to one of the SQLite types (of which either TEXT or BLOB would probably only be practical).
Considering option 3 (TyepConverters) converting the data is of little, if any, use just storing the data as you would not be able to retrieve the data.
As such type converters should always be paired.
One of the pair will be to convert from the class to a type that can be stored.
The other will be to convert from the stored type to the class.
As such you will need quite a few type Converters, that is 2 each for fields:-
clouds (class Clouds)
coord (class Coord)
main (class Main)
sys (class Sys)
weather (class List)
wind (class Wind)
It is the Class of the field that Room looks at to locate the respective type converter.
One of the simplest ways to convert objects (aka classes) is to convert the object to a JSON representation. Although a complexity with this is that there are many JSON libraries and they will often have differences.
For the examples that follow Googles JSON library has been used. However, use of this library with Room doesn't appear to directly support the use of List<the_class> e.g. List.
The dependency for this being (as an example) implementation 'com.google.code.gson:gson:2.10'
As a get around a new class WeatherList has ben used as per:-
data class WeatherList(
val weatherList: List<Weather>
)
and the WeatherDTO class has been changed to use it as per :-
....
//val weather: List<Weather>,
val weather: WeatherList,
....
As such the TypeConverters class could then be:-
class TypeConverters {
#TypeConverter
fun fromCloudsToJSONString(clouds: Clouds): String = Gson().toJson(clouds)
#TypeConverter
fun toCloudsFromJSONString(jsonString: String): Clouds = Gson().fromJson(jsonString,Clouds::class.java)
#TypeConverter
fun fromCoordToJSONString(coord: Coord): String = Gson().toJson(coord)
#TypeConverter
fun toCoordFromJSONString(jsonString: String): Coord = Gson().fromJson(jsonString,Coord::class.java)
#TypeConverter
fun fromMaintoJSONString(main: Main): String = Gson().toJson(main)
#TypeConverter
fun toMainFromJSONString(jsonString: String): Main = Gson().fromJson(jsonString,Main::class.java)
#TypeConverter
fun fromSysToJSONString(sys: Sys): String = Gson().toJson(sys)
#TypeConverter
fun toSysFromJSONString(jsonString: String): Sys = Gson().fromJson(jsonString,Sys::class.java)
#TypeConverter
fun fromWeatherListFromJSONString(weatherList: WeatherList): String = Gson().toJson(weatherList)
#TypeConverter
fun toWeatherListFromJSOnString(jsonString: String): WeatherList = Gson().fromJson(jsonString,WeatherList::class.java)
#TypeConverter
fun fromWindToJSONString(wind: Wind): String = Gson().toJson(wind)
#TypeConverter
fun toWindFromJSONString(jsonString: String): Wind = Gson().fromJson(jsonString,Wind::class.java)
}
As such the all the types/classes/objects that are not directly supported are converted to/from a JSON string representation of the type/class/object.
Note that you need to add the #TypeConverters(#TypeConverters( value = [<????>.TypeConverters::class]). Where has to distinguish between your projects TypeConverters class from Room's (TypeConverters is probably not the best name for the class, renaming it, would overcome the need to distinguish)
Working Example
The following puts the above into action.
As the question does not include the underlying classes, the following have been used:-
data class Coord(
val longitude: Double,
val latitude: Double
)
data class Clouds(
val cover: Double,
val type: String
)
data class Main(
val main: Double
)
data class Sys(
val sys: Double
)
data class WeatherList(
val weatherList: List<Weather>
)
data class Weather(
val weather: Double
)
data class Wind(
val wind: Double
)
The #Dao annotated interface was also made up and is simply:-
#Dao
interface AllDao {
#Insert(onConflict = OnConflictStrategy.IGNORE)
fun insert(weatherDTO: WeatherDTO)
#Query("SELECT * FROM weatherdb")
fun getAllFromWeatherDB(): List<WeatherDTO>
}
Also the #Database annotated abstract class was made up it being:-
#TypeConverters( value = [a.a.so74384736typeconverterconfusion.TypeConverters::class])
#Database(entities = [WeatherDTO::class], exportSchema = false, version = 1)
abstract class TheDatabase: RoomDatabase() {
abstract fun getAllDao(): AllDao
companion object {
private var instance: TheDatabase? = null
fun getInstance(context: Context): TheDatabase {
if (instance==null) {
instance = Room.databaseBuilder(context,TheDatabase::class.java,"the_database.db")
.allowMainThreadQueries()
.build()
}
return instance as TheDatabase
}
}
}
Note the package name used to distinguish the TypeConverters class from Room's TypeConverters class
the package name cannot be used elsewhere, so if the above is copied then it would have to be changed. There is no expectation that the code in it's entirety would be copied and used. The code is designed solely to demonstrate the TypeConverters.
Last some activity code to actually do something (store and retrieve some data):-
class MainActivity : AppCompatActivity() {
lateinit var db: TheDatabase
lateinit var dao: AllDao
override fun onCreate(savedInstanceState: Bundle?) {
super.onCreate(savedInstanceState)
setContentView(R.layout.activity_main)
db = TheDatabase.getInstance(this)
dao = db.getAllDao()
dao.insert(
WeatherDTO(
"base001",
Clouds(25.5,"cumulus"),10,
Coord(10.567,30.345),
11,
12,
Main(12345.67890),
"thename",
Sys(9.87654321),
14,
1000,
WeatherList(listOf(Weather(5.1234),Weather(6.5432), Weather(7.6543))),
Wind(23.12)
)
)
for (wdto in dao.getAllFromWeatherDB()) {
Log.d("DBINFO","base = ${wdto.base} longitude = ${wdto.coord.longitude} latitude = ${wdto.coord.latitude} etc ....")
}
}
}
RESULT
When run the log contains, as expected:-
D/DBINFO: base = base001 longitude = 10.567 latitude = 30.345 etc ....
Using App Inspection then the database looks like:-
The fields converted to a JSON string have been highlighted.
Obviously the data will very likely not exactly be as you would expect due to the made up classes.

Follow on from the previous answer #Embedded versus Type Converters
As can be seen from the previous answer, there are some issues in regard to using TypeConverters. From a database perspective the TypeConverters will inevitably contain bloat/unecessary data which is contrary to normalisation (not needlessly storing repetitive data).
As an example the JSON representation will for every row contain exactly the same the field names wasting storage, all rows will have the additional overhead of storing the delimiters ([s and ]s, {s and }s, :s ,s). Furthermore actually using the stored data can become complex due to the bloat and also due to multiple values being stored in a single column and as such can be restrictive.
It would be more efficient to not store the bloat and it could eliminate complexities and enhance the usability of the stored data from a database perspective (querying the data for retrieval) to not store multiple values in a single column.
Using the #Embedded annotation can very easily eliminate the bloat. Consider the following (an alternative version of the WeatherDTO class/entity):-
#Entity(tableName = "WeatherDbAlternative1")
data class WeatherDTOAlternative1(
val base: String,
#Embedded
val clouds: Clouds,
val cod: Int,
#Embedded
val coord: Coord,
val dt: Int,
#PrimaryKey(autoGenerate = true)
val id: Int,
#Embedded
val main: Main,
val name: String,
#Embedded
val sys: Sys,
val timezone: Int,
val visibility: Int,
//val weather: List<Weather>,
/* Unable to embed directly so not embedding */
val weather: WeatherList,
#Embedded
val wind: Wind
)
Bar the weather field all that has been done is add the #Embedded annotation. Noting that the classes of the fields all have fields of types directly supported by Room.
Adding this entity to the #Database annotation and adding a couple of additional functions in the #Dao annotated class as per:-
#Query("SELECT * FROM weatherdbalternative1")
fun getAllFromWeatherDBAlternative1(): List<WeatherDTOAlternative1>
#Insert(onConflict = OnConflictStrategy.IGNORE)
fun insert(weatherDTOAlternative1: WeatherDTOAlternative1)
And then amending the Activity code to include :-
/*ALTERNATIVE 1 All but WeatherList embedded */
dao.insert(
WeatherDTOAlternative1(
"base001A",
Clouds(25.5, "cumulus"),
10,
Coord(10.567, 30.345),
11,
12,
Main(12345.67890),
"thenameA1",
Sys(9.87654321),
14,
1000,
WeatherList(listOf(Weather(5.1234), Weather(6.5432), Weather(7.6543))),
Wind(23.12)
)
)
for (wdto in dao.getAllFromWeatherDBAlternative1()) {
Log.d(
"DBINFO",
"base = ${wdto.base} longitude = ${wdto.coord.longitude} latitude = ${wdto.coord.latitude} etc ...."
)
}
Now results in the Log including:-
D/DBINFO: base = base001 longitude = 10.567 latitude = 30.345 etc ....
D/DBINFO: base = base001A longitude = 10.567 latitude = 30.345 etc ....
i.e. effectively the same data is stored and retrievable
However the data is now stored in the database as (ignoring the weather field) as :-
i.e. the data stored is much cleaner but at the expense of additional columns (which can be advantageous).
additionally although not apparent, the fields that have the #Embedded annotation do not need the TypeConverters.

Kotlin Room Database Save Control

I save data to Room database with a button. However, every time I press the button, that item is registered and listed again and again. I want to write a control. I do not want it to save it if it is registered. I will be glad if you help.
ArticleDao
#Dao
interface ArticleDao {
#Insert(onConflict = OnConflictStrategy.REPLACE)
suspend fun upsert(article: Article): Long
NewsRepisotry
class NewsRepository(
val db: ArticleDatabase
) {
suspend fun upsert(article: Article)=db.getArticleDao().upsert(article)
fun getSavedNews()=db.getArticleDao().getAllArticles()
ArticleDatabase
#Database(
entities = [Article::class],
version = 1
)
#TypeConverters(Converters::class)
abstract class ArticleDatabase : RoomDatabase() {
abstract fun getArticleDao(): ArticleDao
companion object {
#Volatile
private var instance: ArticleDatabase? = null
private val LOCK = Any()
operator fun invoke(context: Context) = instance ?: synchronized(LOCK) {
instance ?: createDatabase(context).also { instance = it }
}
private fun createDatabase(context: Context) =
Room.databaseBuilder(
context.applicationContext,
ArticleDatabase::class.java,
"article_db.db"
).build()
}
}
Article.kt
#Entity(
tableName = "articles"
)
data class Article(
#PrimaryKey(autoGenerate = true)
var Id: Int? = null,
val author: String?,
val content: String?,
val description: String?,
val publishedAt: String?,
val source: Source?,
val title: String?,
val url: String?,
val urlToImage: String?
) : Serializable

Your issue would appear to be that #Insert(onConflict = OnConflictStrategy.REPLACE) is inserting instead of replacing.
The OnConflict strategy will only be actioned if there is a conflict. That is typically that a UNIQUE constraint is violated (see link below for other violations covered).
From your explanation of the issue there is no violation and thus instead of the row(s) being replaced, they are instead inserted.
You may wish to refere to https://sqlite.org/lang_conflict.html
In short you need a UNIQUE index to cover your scenario to ensure that a violation occurs.
Without the Article class it is impossible to say what actual change is required. However the following example shows 2 classes (entities) and how a unique index can be added to a column/property articleName.
So first a guess as what your Article class could be:-
#Entity
data class Article(
#PrimaryKey
var articelId: Long? = null,
var articleName: String,
var otherData: String
)
and here an equivalent with a UNIQUE index on the articleName property/column :-
#Entity(indices = [Index(value = ["articleName"],unique = true)])
data class ArticleV2(
#PrimaryKey
var articleId: Long? = null,
var articleName: String,
var otherData: String
)
Example/Demo
As an example of how both work differently, the latter according to your expectations, consider the following code, utilising the above, (for brevity/convenience run on the main thread) :-
The #Dao ArticleDao for both tables/entiities/classes :-
#Dao
interface ArticleDao {
#Insert(onConflict = OnConflictStrategy.REPLACE)
fun upsert(article: Article): Long
#Insert(onConflict = OnConflictStrategy.REPLACE)
fun upsert(articleV2: ArticleV2)
#Query("SELECT count(*) FROM article")
fun getArticleCount(): Long
#Query("SELECT count(*) FROM articlev2")
fun getArticleV2Count(): Long
}
The #Database ArticleDatabase class :-
#Database(
entities = [Article::class, ArticleV2::class],
version = 1
)
abstract class ArticleDatabase : RoomDatabase() {
abstract fun getArticleDao(): ArticleDao
companion object {
#Volatile
private var instance: ArticleDatabase? = null
private val LOCK = Any()
operator fun invoke(context: Context) = instance ?: synchronized(LOCK) {
instance ?: createDatabase(context).also { instance = it }
}
private fun createDatabase(context: Context) =
Room.databaseBuilder(
context.applicationContext,
ArticleDatabase::class.java,
"article_db.db"
)
.allowMainThreadQueries()
.build()
}
}
Finally an activity, that replicates your "every time I press the button" by looping 5 times inserting one row into each table per iteration and then after the loop writing the number of rows in each of the tables to the log.
class MainActivity : AppCompatActivity() {
lateinit var db: ArticleDatabase
lateinit var dao: ArticleDao
override fun onCreate(savedInstanceState: Bundle?) {
super.onCreate(savedInstanceState)
setContentView(R.layout.activity_main)
db = ArticleDatabase.invoke(this)
dao = db.getArticleDao()
for(i in 1..5) {
dao.upsert(Article(null, "Article", "otherdata$i"))
dao.upsert(ArticleV2(null,"Article","otherdata$i"))
}
Log.d("ARTICLECOUNT","Number of Articles in article table = " + dao.getArticleCount() + "... in articleV2 table = " + dao.getArticleV2Count())
}
}
When run (first time) the output in the log contains:-
D/ARTICLECOUNT: Number of Articles in article table = 5... in articleV2 table = 1
So 5 rows inserted into the article table but only 1 row in the articleV2 table (as the value used for the articleName is the same).
After running Database Inspector shows :-
The article table
The articleV2 table
As can be seen the otherData column has been replaced throughout and is otherdata5
If the OnConflict strategy were IGNORE then otherData would have a value of otherdata1
Additional
regarding the comment and Article class now added.
First of all, thank you sir. I have also added the Article class. Your explanation is very good, but I could not understand exactly what I needed to change.
You need to decide what constitutes a duplicate entry.
Perhaps the URL if unchanged could be the most likely to indicate a duplicate.
So you could use :-
#Entity( tableName = "articles", indices = [Index(value = ["url"],unique = true)] ) //<<<<<<<<< ONLY CHANGE IS THE #Entity
data class Article(
#PrimaryKey(autoGenerate = true)
var Id: Int? = null,
val author: String?,
val content: String?,
val description: String?,
val publishedAt: String?,
val source: Source?,
val title: String?,
val url: String?,
val urlToImage: String?
) : Serializable
thus a UNIQUE index (one that does not allow the same value to be stored more than once) would be created and maintained on the url column.
Note NULL is not considered a duplicate value (that is a NULL is considered to be different to any other NULL), so if this value can be null then perhaps it is not a wise choice.
However, you may decide that multiple columns should constitute a duplicate to not be inserted. Here's a variation showing how multiple columns (composite index) can be used.
#Entity( tableName = "articles", indices = [Index(value = ["author","content","description"],unique = true)] ) //<<<<<<<<< ONLY CHANGE IS THE #Entity
data class Article(
#PrimaryKey(autoGenerate = true)
var Id: Int? = null,
val author: String?,
val content: String?,
val description: String?,
val publishedAt: String?,
val source: Source?,
val title: String?,
val url: String?,
val urlToImage: String?
) : Serializable
in this case if the author, content and description were the same then it would be a duplicate.
Note in all cases allowing all null values (? after all) could be problematic if there is no validation to stop all values being null (i.e. the data would probably be useless).
You may also wish to consider clearing inputs (some only perhaps) after an insertion, thus making it less likely for a duplicate to be attemtped.

Android Room, how to save an entity with one of the variables being a sealed class object

I want to save in my Room database an object where one of the variables can either be of on type or another. I thought a sealed class would make sense, so I took this approach:
sealed class BluetoothMessageType() {
data class Dbm(
val data: String
) : BluetoothMessageType()
data class Pwm(
val data: String
) : BluetoothMessageType()
}
Or even this, but it is not necessary. I found that this one gave me even more errors as it did not know how to handle the open val, so if I find a solution for the first version I would be happy anyway.
sealed class BluetoothMessageType(
open val data: String
) {
data class Dbm(
override val data: String
) : BluetoothMessageType()
data class Pwm(
override val data: String
) : BluetoothMessageType()
}
Then the Entity class
#Entity(tableName = MESSAGES_TABLE_NAME)
data class DatabaseBluetoothMessage(
#PrimaryKey(autoGenerate = true)
val id: Long = 0L,
val time: Long = Instant().millis,
val data: BluetoothMessageType
)
I have created a TypeConverter to convert it to and from a String as well, so I assume that it is not a problem.
First, is this possible? I assume this should function in a similar way that it would with an abstract class, but I have not managed to find a working solution with that either. If it is not possible, what sort of approach should I take when I want to save some data that may be either of one or another type if not with sealed classes?

We faced such problem when we tried using Polymorphism in our domain, and we solved it this way:
Domain:
We have a Photo model that looks like this:
sealed interface Photo {
val id: Long
data class Empty(
override val id: Long
) : Photo
data class Simple(
override val id: Long,
val hasStickers: Boolean,
val accessHash: Long,
val fileReferenceBase64: String,
val date: Int,
val sizes: List<PhotoSize>,
val dcId: Int
) : Photo
}
Photo has PhotoSize inside, it looks like this:
sealed interface PhotoSize {
val type: String
data class Empty(
override val type: String
) : PhotoSize
data class Simple(
override val type: String,
val location: FileLocation,
val width: Int,
val height: Int,
val size: Int,
) : PhotoSize
data class Cached(
override val type: String,
val location: FileLocation,
val width: Int,
val height: Int,
val bytesBase64: String,
) : PhotoSize
data class Stripped(
override val type: String,
val bytesBase64: String,
) : PhotoSize
}
Data:
There is much work to do in our data module to make this happen. I will decompose the process to three parts to make it look easier:
1. Entity:
So, using Room and SQL in general, it is hard to save such objects, so we had to come up with this idea. Our PhotoEntity (Which is the Local version of Photo from our domain looks like this:
#Entity
data class PhotoEntity(
// Shared columns
#PrimaryKey
val id: Long,
val type: Type,
// Simple Columns
val hasStickers: Boolean? = null,
val accessHash: Long? = null,
val fileReferenceBase64: String? = null,
val date: Int? = null,
val dcId: Int? = null
) {
enum class Type {
EMPTY,
SIMPLE,
}
}
And our PhotoSizeEntity looks like this:
#Entity
data class PhotoSizeEntity(
// Shared columns
#PrimaryKey
#Embedded
val identity: Identity,
val type: Type,
// Simple columns
#Embedded
val locationLocal: LocalFileLocation? = null,
val width: Int? = null,
val height: Int? = null,
val size: Int? = null,
// Cached and Stripped columns
val bytesBase64: String? = null,
) {
data class Identity(
val photoId: Long,
val sizeType: String
)
enum class Type {
EMPTY,
SIMPLE,
CACHED,
STRIPPED
}
}
Then we have this compound class to unite PhotoEntity and PhotoSizeEntity together, so we can retrieve all data required by our domain's model:
data class PhotoCompound(
#Embedded
val photo: PhotoEntity,
#Relation(entity = PhotoSizeEntity::class, parentColumn = "id", entityColumn = "photoId")
val sizes: List<PhotoSizeEntity>? = null,
)
2. Dao
So our dao should be able to store and retrieve this data. You can have two daos for PhotoEntity and PhotoSizeEntity instead of one, for the sake of flexibility, but in this example we will use a shared one, it looks like this:
#Dao
interface IPhotoDao {
#Transaction
#Query("SELECT * FROM PhotoEntity WHERE id = :id")
suspend fun getPhotoCompound(id: Long): PhotoCompound
#Transaction
suspend fun insertOrUpdateCompound(compound: PhotoCompound) {
compound.sizes?.let { sizes ->
insertOrUpdate(sizes)
}
insertOrUpdate(compound.photo)
}
#Insert(onConflict = OnConflictStrategy.REPLACE)
suspend fun insertOrUpdate(entity: PhotoEntity)
#Insert(onConflict = OnConflictStrategy.REPLACE)
suspend fun insertOrUpdate(entities: List<PhotoSizeEntity>)
}
3. Adapter:
After solving the problem of saving data to SQL database, we now need to solve the problem of converting between domain and local entities. Our Photo's converter aka adapter looks like this:
fun Photo.toCompound() = when(this) {
is Photo.Empty -> this.toCompound()
is Photo.Simple -> this.toCompound()
}
fun PhotoCompound.toModel() = when (photo.type) {
PhotoEntity.Type.EMPTY -> Photo.Empty(photo.id)
PhotoEntity.Type.SIMPLE -> this.toSimpleModel()
}
private fun PhotoCompound.toSimpleModel() = photo.run {
Photo.Simple(
id,
hasStickers!!,
accessHash!!,
fileReferenceBase64!!,
date!!,
sizes?.toModels()!!,
dcId!!
)
}
private fun Photo.Empty.toCompound(): PhotoCompound {
val photo = PhotoEntity(
id,
PhotoEntity.Type.EMPTY
)
return PhotoCompound(photo)
}
private fun Photo.Simple.toCompound(): PhotoCompound {
val photo = PhotoEntity(
id,
PhotoEntity.Type.SIMPLE,
hasStickers = hasStickers,
accessHash = accessHash,
fileReferenceBase64 = fileReferenceBase64,
date = date,
dcId = dcId,
)
val sizeEntities = sizes.toEntities(id)
return PhotoCompound(photo, sizeEntities)
}
And for the PhotoSize, it looks like this:
fun List<PhotoSize>.toEntities(photoId: Long) = map { photoSize ->
photoSize.toEntity(photoId)
}
fun PhotoSize.toEntity(photoId: Long) = when(this) {
is PhotoSize.Cached -> this.toEntity(photoId)
is PhotoSize.Empty -> this.toEntity(photoId)
is PhotoSize.Simple -> this.toEntity(photoId)
is PhotoSize.Stripped -> this.toEntity(photoId)
}
fun List<PhotoSizeEntity>.toModels() = map { photoSizeEntity ->
photoSizeEntity.toModel()
}
fun PhotoSizeEntity.toModel() = when(type) {
PhotoSizeEntity.Type.EMPTY -> this.toEmptyModel()
PhotoSizeEntity.Type.SIMPLE -> this.toSimpleModel()
PhotoSizeEntity.Type.CACHED -> this.toCachedModel()
PhotoSizeEntity.Type.STRIPPED -> this.toStrippedModel()
}
private fun PhotoSizeEntity.toEmptyModel() = PhotoSize.Empty(identity.sizeType)
private fun PhotoSizeEntity.toCachedModel() = PhotoSize.Cached(
identity.sizeType,
locationLocal?.toModel()!!,
width!!,
height!!,
bytesBase64!!
)
private fun PhotoSizeEntity.toSimpleModel() = PhotoSize.Simple(
identity.sizeType,
locationLocal?.toModel()!!,
width!!,
height!!,
size!!
)
private fun PhotoSizeEntity.toStrippedModel() = PhotoSize.Stripped(
identity.sizeType,
bytesBase64!!
)
private fun PhotoSize.Cached.toEntity(photoId: Long) = PhotoSizeEntity(
PhotoSizeEntity.Identity(photoId, type),
PhotoSizeEntity.Type.CACHED,
locationLocal = location.toEntity(),
width = width,
height = height,
bytesBase64 = bytesBase64
)
private fun PhotoSize.Simple.toEntity(photoId: Long) = PhotoSizeEntity(
PhotoSizeEntity.Identity(photoId, type),
PhotoSizeEntity.Type.SIMPLE,
locationLocal = location.toEntity(),
width = width,
height = height,
size = size
)
private fun PhotoSize.Stripped.toEntity(photoId: Long) = PhotoSizeEntity(
PhotoSizeEntity.Identity(photoId, type),
PhotoSizeEntity.Type.STRIPPED,
bytesBase64 = bytesBase64
)
private fun PhotoSize.Empty.toEntity(photoId: Long) = PhotoSizeEntity(
PhotoSizeEntity.Identity(photoId, type),
PhotoSizeEntity.Type.EMPTY
)
That's it!
Conclusion:
To save a sealed class to Room or SQL, whether as an Entity, or as an Embedded object, you need to have one big data class with all the properties, from all the sealed variants, and use an Enum type to indicate variant type to use later for conversion between domain and data, or for indication in your code if you don't use Clean Architecture. Hard, but solid and flexible. I hope Room will have some annotations that can generate such code to get rid of the boilerplate code.
PS: This class is taken from Telegram's scheme, they also solve the problem of polymorphism when it comes to communication with a server. Checkout their TL Language here: https://core.telegram.org/mtproto/TL
PS2: If you like Telegram's TL language, you can use this generator to generate Kotlin classes from scheme.tl files: https://github.com/tamimattafi/mtproto
EDIT: You can use this code generating library to automatically generate Dao for compound classes, to make it easier to insert, which removes a lot of boilerplate to map things correctly.
Link: https://github.com/tamimattafi/android-room-compound
Happy Coding!

In my case I did the following:
sealed class Status2() {
object Online : Status2()
object Offline : Status2()
override fun toString(): String {
return when (this) {
is Online ->"Online"
is Offline -> "Offline"
}
}
}
class StatusConverter{
#TypeConverter
fun toHealth(value: Boolean): Status2 {
return if (value){
Status2.Online
} else{
Status2.Offline
}
}
#TypeConverter
fun fromHealth(value: Status2):Boolean {
return when(value){
is Status2.Offline -> false
is Status2.Online -> true
}
}
}
#Dao
interface CourierDao2 {
#Insert(onConflict = OnConflictStrategy.REPLACE)
suspend fun insertStatus(courier: CourierCurrentStatus)
#Query("SELECT * FROM CourierCurrentStatus")
fun getCourierStatus(): Flow<CourierCurrentStatus>
}
#Entity
data class CourierCurrentStatus(
#PrimaryKey
val id: Int = 0,
var status: Status2 = Status2.Offline
)
and it works like a charm