Accessing NoSQL or REST data sources

VoK currently provides no out-of-the-box support for accessing NoSQL databases. However, VoK offers a lot of support for showing fetched data in the Vaadin Grid, with support for sorting, filtering (including auto-generated filter bar) and paging. All you need to implement is single interface called DataLoader. For information on how to do that, please read on.

VoK REST Support

VoK provides full support for publishing and consuming entities in a CRUD fashion over REST, with all features like paging, sorting and filtering. VoK even provides an implementation of Vaadin Grid’s DataProvider which fetches data over REST.

There are two modules on VoK providing REST support:

vok-rest when you need to publish data from your VoK server to the world;
vok-rest-client when you need to consume REST endpoints published elsewhere, and maybe displaying the fetched data in a Grid.

Please click the links to read the module documentation - it should contain all information necessary for you to get started.

There is also an example project which exposes a list of entities over REST server, then self-consumes them via REST client connected to localhost:8080 and exposes them in a Grid: vaadin8-restdataprovider-example and vaadin10-restdataprovider-example.

Data Loaders

VoK data fetching revolves around a thing called a Data Loader. The data loader is responsible for loading pages of data from arbitrary data source, sorting and filtering it as necessary, converting rows into Java Beans. The data loader then can be plugged into Vaadin Grid, in order to show the data.

For example, the REST client class CrudClient is a data loader since it implements the DataLoader interface. It fetches lists of JSON maps and converts them into Java Beans with the help of Gson. In order to connect CrudClient into Vaadin Grid, you need to use DataLoaderAdapter to convert CrudClient into DataProvider. It’s always wise to call .withConfigurableFilter2() on the adapter which enables the filter row to set filters to the data provider.

Another example is the SQL database access via the vok-orm. vok-orm provides data loader which fetches instances of entities from the database. VoK then provides convenience extension method on vok-orm Dao - the Dao.dataProvider - which takes the data loader and uses DataLoaderAdapter to convert it into DataProvider. VoK also provides the sqlDataProvider() function which works on arbitrary SQL SELECTs.

Another example would be to add support for MongoDB. You only need to figure out how to fetch data with given paging and sorting, then define allowed filters and define mapping from JSON map to a Java Bean. That’s all that’s needed to implement a MongoDB-backed DataLoader. Then it’s just a matter of using DataLoaderAdapter to turn DataLoader into Vaadin DataProvider, feed that to Vaadin Grid and let VoK auto-generate filter bar for you.

Why Not Use DataProvider API Directly?

Why we need the data loaders? Can’t we simply implement Vaadin’s DataProvider API and pass it to the Grid directly? Unfortunately we can’t, for the following reasons:

The DataLoader API is simpler than the DataProvider.
The DataLoader API library (vok-dataloader) comes with filter classes provided (EqFilter, LikeFilter etc), so you don’t have to write your own filter classes.

Implementing Data Loader

This chapter documents the general information what you need to write a custom DataLoader for your particular data source. It is important to understand the basics, before you start using whatever VoK provides for you.

Feeding Grid With Data

In order to feed the Vaadin Grid with data, you need to have two things:

A bean (a data class) that is able to hold all data for one row of a Grid
A proper implementation of DataLoader which would then:
- Fetch the data from your NoSQL/REST/other source;
- Map the data to a bean so that it provides instances of that bean directly.

Note: If you are accessing a SQL database, there are already two data providers pre-provided for you: the EntityDataProvider (fetching entities from their tables) and SQLDataProvider (fetching result of any SQL SELECT query). Please read Grids and Accessing SQL Databases for more details.

To implement the DataLoader interface you only need to implement two methods:

fun getCount(filter: Filter<T>?): Long takes filters and computes number of matching rows. That enables the Grid to draw a proper scroll bar.
fun fetch(filter: Filter<T>?, sortBy: List<SortClause>, range: LongRange): List<T> function retrieves the actual data. You need to pay attention to the range parameter which specifies the paging.

Your initial implementation does not need to support any filters nor sorting, to keep things simple. Simply ignore the filter and sortBy parameters for now, and focus on implementing the conversion from your native data source to a Java Bean. You can use:

Gson for conversion from JSON to Java Bean; you can use the vok-provided Gson.fromJsonArray() to convert from an array of JSON maps, or use the following code: fromJson<List<T>>(reader, TypeToken.getParameterized(List::class.java, itemClass).type)
vok-orm to convert an outcome of a JDBC result to a Java Bean

Afer you have the skeletal implementation of your DataLoader ready, let’s focus on converting the DataLoader to Vaadin’s DataProvider so that you can use it with the Grid. Say that you fetch a list of Person:

val dataLoader = ... // your DataLoader here
val dp = DataLoaderAdapter(Person::class.java, crud, { it.id!! }).withConfigurableFilter2()
grid.dataProvider = dp

Grid will never attempt to pass in any filters on its own. If you need filtering, Grid expects you to create a filter bar and implement everything yourself. Luckily, VoK is able to do that for you. We’ll go through this in a minute.

Adding Support For Sorting

If you need certain columns in the Grid to be sortable, you need to support sorting clauses in your data provider implementation - you need to pay attention to the sortBy parameter of the DataLoader.fetch() method.

In your DataLoader implementation you need to convert the contents of the sortBy parameter into e.g. a list of query parameters (in case of REST), or a sorting clauses (in case of a NoSQL query).

Which columns are actually sortable will depend on your data source. For REST, the sorting parameters are limited to whatever is available as a query parameter in the REST http call. For NoSQL, the set of available indices limit available sorting parameters. By default it is expected that all properties of the Java Bean (e.g. the Person class) are sortable; the SortClause.propertyName in the sortBy parameter may be a name of any property in the Person class.

Of course that may be unwanted since sorting generally requires indices and indices tend to slow down insertion of new data. Therefore, it is best for your DataLoader to throw IllegalArgumentException in fetch() for any unsupported sort clause it encounters in the sortBy field. Then, document all supported sorting criteria in the kdoc for your DataLoader, so that you’ll know which columns to mark as non-sortable in the Grid.

You may also need to provide support for ‘hidden’ columns (not mapped to the Java Bean); you may therefore decide to also support native property names on SortClause, e.g. p.id from SELECT p.* FROM Person p INNER JOIN .... Just make sure everything is documented in your DataLoader kdoc. See vok-dataloader for an explanation of the difference between native properties and data loader properties, and when to support which.

Adding Support For Filtering

To add support for filters, both of your getCount() and fetch() implementations must take the filter parameter into consideration. The DataLoader comes with a predefined set of filters which you should implement:

EqFilter for equality comparisons
OpFilter for less-than, greater-than etc comparisons
LikeFilter for case-sensitive starts-with comparison
ILikeFilter for case-insensitive starts-with comparison
IsNullFilter and IsNotNullFilter
AndFilter for ANDing multiple filters
OrFilter for ORing multiple filters. This one is optional: for example the REST CRUD client doesn’t support this one; also VoK-autogenerated filter bar never produces the OR filter.

You should document, for which fields the filtering is supported, in the data loader javadoc/kdoc. Since filtering generally also requires indices, the remark from the sorting chapter above also applies. You need to support exactly the same property naming scheme which you support for sorting.

Now your data loader is ready to be plugged into the Grid. Go ahead, plug it into the Grid and play with the sorting a bit until you’re happy with the result. However, Grid never outputs any filters. In order to test filters, we need to add a filter bar.

Filter Factory

If your data loader supports all filters as required by the FilterFactory interface, then you can use the FilterFieldFactory class to automatically create the filtering UI components and auto-populate the filter bar for you.

Just implement the FilterFactory interface, then override DefaultFilterFieldFactory and tailor it towards your needs (e.g. make createField() return null for columns you don’t want to support). Then, in your code just call

grid.appendHeaderRow().generateFilterComponents(grid, itemClass, filterFieldFactory)

to create the filter row and populate it with filter components.

More Resources

Please read the documentation for the vok-util module for more information on generating filters.

Please find all technical details on how to implement the data loader in the vok-dataloader page.