Anko SQLite

Posted 2021-04-30 未来演化师

Have you ever been tired of parsing SQLite query results using android cursors? You have to write lots of boilerplate code just to parse query result rows, and enclose it in countless try..finally blocks to properly close all opened resources.

Anko provides lots of extension functions to simplify working with SQLite databases.‍

Contents

Using Anko SQLite in your project
Accessing database
Creating and dropping tables
Inserting data
Querying data
Deleting data
Parsing query results
Custom row parsers
Cursor streams
Updating values

Transactions

Using Anko SQLite in your project
Add the anko-sqlite dependency to your build.gradle:

dependencies {
    compile "org.jetbrains.anko:anko-sqlite:$anko_version"
}

Accessing the database
If you use SQLiteOpenHelper, you generally call getReadableDatabase() or getWritableDatabase() (result is actually the same in production code), but then you must be sure to call the close() method on the received SQLiteDatabase. Also, you have to cache the helper class somewhere, and if you use it from several threads, you must be aware of the concurrent access. All this is pretty tough. That is why Android developers are not really keen on default SQLite API and prefer to use fairly expensive wrappers such as ORMs instead.
Anko provides a special class ManagedSQLiteOpenHelper that seamlessly replaces the default one. Here's how you can use it:

class MyDatabaseOpenHelper(ctx: Context) : ManagedSQLiteOpenHelper(
        ctx, "MyDatabase", null, 1) {
    companion object {
        private var instance: MyDatabaseOpenHelper? = null
        @Synchronized
        fun getInstance(ctx: Context): MyDatabaseOpenHelper {
            if (instance == null) {
                instance = MyDatabaseOpenHelper(ctx.getApplicationContext())
            }
            return instance!!
        }
    }

    override fun onCreate(db: SQLiteDatabase) {
        // Here you create tables
        db.createTable("Customer", true,
                "id" to INTEGER + PRIMARY_KEY + UNIQUE,
                "name" to TEXT,
                "photo" to BLOB)
    }

    override fun onUpgrade(db: SQLiteDatabase, oldVersion: Int, newVersion: Int) {
        // Here you can upgrade tables, as usual
        db.dropTable("User", true)
    }
}
// Access property for Contextval
Context.database: MyDatabaseOpenHelper
get() = MyDatabaseOpenHelper.getInstance(getApplicationContext())

So what's the sense? Instead of enclosing your code into try blocks, now you can just write this:

database.use {
// `this` is a SQLiteDatabase instance
}

The database will definitely be closed after executing all code inside {}.
Asynchronous call example:

class SomeActivity : Activity() {
    private fun loadAsync() {
        async(UI) {
            val result = bg {
                database.use { ... }
            }
            loadComplete(result)
        }
    }
}

These and all methods mentioned below may throw SQLiteException. You have to handle it by yourself because it would be unreasonable for Anko to pretend that errors don't occur.

Creating and dropping tables

With Anko you can easily create new tables and drop existing ones. The syntax is straightforward.

database.use {
    createTable("Customer", true,
            "id" to INTEGER + PRIMARY_KEY + UNIQUE,
            "name" to TEXT,
            "photo" to BLOB)
}

In SQLite, there are five main types: NULL, INTEGER, REAL, TEXT and BLOB. But each column may have some modifiers like PRIMARY KEY or UNIQUE. You can append such modifiers with "adding" them to the primary type name.
To drop a table, use the dropTable function:

dropTable("User", true)

Inserting data

Usually, you need a ContentValues instance to insert a row into the table. Here is an example:

val values = ContentValues()
values.put("id", 5)
values.put("name", "John Smith")
values.put("email", "[e-mail]")
db.insert("User", null, values)

Anko lets you eliminate such ceremonies by passing values directly as arguments for the insert()function:

// Where db is an SQLiteDatabase
// eg: 
val db = database.writeableDatabasedb.insert("User",
        "id" to 42,
        "name" to "John",
        "email" to "[e-mail]")

or from within database.use as:

database.use {
    insert("User",
            "id" to 42,
            "name" to "John",
            "email" to "[e-mail]"
    )
}

Please note, in the above example database is a database helper instance, and db is a SQLiteDatabase object
Functions insertOrThrow(), replace(), replaceOrThrow() also exist and have the similar semantics.
Querying data
Anko provides a convenient query builder. It may be created with db.select(tableName, vararg columns) where db is an instance of SQLiteDatabase.

Method	Description
column(String)	Add a column to select query
distinct(Boolean)	Distinct query
whereArgs(String)	Specify raw String where query
whereArgs(String, args) ★	Specify a where query with arguments
whereSimple(String, args)	Specify a where query with ? mark arguments
orderBy(String, [ASC/DESC])	Order by this column
groupBy(String)	Group by this column
limit(count: Int)	Limit query result row count
limit(offset: Int, count: Int)	Limit query result row count with an offset
having(String)	Specify raw having expression
having(String, args) ★	Specify a having expression with arguments

‍

Functions marked with ★ arse its arguments in a special way. They allow you to provide values in any ‍order and support escaping seamlessly.

db.select("User", "name")
.whereArgs("(_id > {userId}) and (name = {userName})",
"userName" to "John",
"userId" to 42)

Here, {userId} part will be replaced with 42 and {userName} — with 'John'. The value will be escaped if its type is not numeric (Int, Float etc.) or Boolean. For any other types, toString() representation will be used.
whereSimple function accepts arguments of String type. It works the same as query() from SQLiteDatabase (question marks ? will be replaced with actual values from arguments).
How can we execute the query? Using the exec() function. It accepts an extension function with the type of Cursor.() -> T. It simply launches the received extension function and then closes Cursor so you don't need to do it by yourself:

db.select("User", "email").exec {
// Doing some stuff with emails
}

Parsing query results

So we have some Cursor, and how can we parse it into regular classes? Anko provides functions parseSingle, parseOpt and parseList to do it much more easily.

Method	Description
parseSingle(rowParser): T	Parse exactly one row
parseOpt(rowParser): T?	Parse zero or one row
parseList(rowParser): List<T>	Parse zero or more rows

Note that parseSingle() and parseOpt() will throw an exception if the received Cursor contains more than one row.
Now the question is: what is rowParser? Well, each function supports two different types of parsers: RowParser and MapRowParser:

interface RowParser<T> {
    fun parseRow(columns: Array<Any>): T
}

interface MapRowParser<T> {
    fun parseRow(columns: Map<String, Any>): T
}

If you want to write your query in a very efficient way, use RowParser (but then you must know the index of each column). parseRow accepts a list of Any (the type of Any could practically be nothing but Long, Double, String or ByteArray). MapRowParser, on the other hand, lets you get row values by using column names.
Anko already has parsers for simple single-column rows:

ShortParser
IntParser
LongParser
FloatParser
DoubleParser
StringParser
BlobParser
Also, you can create a row parser from the class constructor. Assuming you have a class:

class Person(val firstName: String, val lastName: String, val age: Int)

The parser will be as simple as:

val rowParser = classParser<Person>()

For now, Anko does not support creating such parsers if the primary constructor has optional parameters. Also, note that constructor will be invoked using Java Reflection so writing a custom RowParser is more reasonable for huge data sets.
If you are using Anko db.select() builder, you can directly call parseSingle, parseOpt or parseList on it and pass an appropriate parser.
Custom row parsers
For instance, let's make a new parser for columns (Int, String, String). The most naive way to do so is:

class MyRowParser : RowParser<Triple<Int, String, String>> {
    override fun parseRow(columns: Array<Any>): Triple<Int, String, String> {
        return Triple(columns[0] as Int, columns[1] as String, columns[2] as String)
    }
}

Well, now we have three explicit casts in our code. Let's get rid of them by using the rowParser function:

val parser = rowParser { id: Int, name: String, email: String ->
    Triple(id, name, email)
}

And that's it! rowParser makes all casts under the hood and you can name the lambda parameters as you want.
Cursor streams
Anko provides a way to access SQLite Cursor in a functional way. Just call cursor.asSequence() or cursor.asMapSequence() extension functions to get a sequence of rows. Do not forget to close the Cursor :)
Updating values
Let's give a new name to one of our users:

update("User", "name" to "Alice")
.where("_id = {userId}", "userId" to 42)
.exec()

Update also has a whereSimple() method in case you want to provide the query in a traditional way:

update("User", "name" to "Alice")
.`whereSimple`("_id = ?", 42)
.exec()

Deleting Data
Let's delete a row (notice the delete method doesn't have any whereSimple() method; instead you supply the query directly in the arguments):

val numRowsDeleted = delete("User", "_id = {userID}", "userID" to 37)

Transactions
There is a special function called transaction() which allows you to enclose several database operations in a single SQLite transaction.

transaction {
// Your transaction code
}

The transaction will be marked as successful if no exception was thrown inside the {} block.

If you want to abort a transaction for some reason, just throw TransactionAbortException. You don't need to handle this exception by yourself in this case.‍