How to add 1 milion items in Realm correctly?

I want to choose between native SQLiteDatabase and Realm to deal with a big amount of data.

For benchmark I add to storage 1 milion of Product entities:

{id:integer,sku:string,name:string,data_creating:string}

Using SQLiteDatabase it takes near 1 minute 34 seconds on my device.

Using Realm it takes more them 10 minutes.

My code is:

Realm realm = Realm.getInstance(getApplicationContext());
realm.beginTransaction();
for(int i = 0 ; i < 1000000;i++){
    Product product = realm.createObject(Product.class);
    product.setId(i+1);
    product.setName("Product_"+i);
    product.setSku("SKU__"+i);
    product.setDateCreated(new Date());
}
realm.commitTransaction();

How can I improve my code for better time performance?

Solution

You have to be aware that SQLite and Realm are two very different things. Realm is an object store and you are creating a lot of objects in the code shown above. Depending on your model class and the number of rows/objects, you will often see that Realm is a bit slower on inserts. To do a fair comparison, you could compare Realm with one of the many excellent ORMs out there.

Said that, Realm offers a low-level interface (io.realm.internal). I wouldn't recommend you to use it as it is currently undocumented. Your example would look like this:

long numberOfObjects = 1000000;
SharedGroup sharedGroup = new SharedGroup("default.realm");            
WriteTransaction writeTransaction = sharedGroup.beginWrite();
Table table = writeTransaction.getTable("class_Product");
table.addEmptyRows(numberOfObjects);
for (int i = 0; i < numberOfObjects; i++) {
    table.setLong(0, i, i);              // id
    table.setString(1, i, "Product_"+i); // name
    table.setString(2, i, "SKU__"+i);    // sku
    table.SetDate(3, i, new Date());     // date
}
writeTransaction.commit();
sharedGroup.close();

You can now compare two table/row oriented data stores, and you will probably find that Realm is a bit faster than SQLite.

At Realm, we have a few ideas on how to get our object interface to run faster, and we hope to be able to implement them in the near future.