datahub/docs/backends.markdown

layout, title, root

layout	title	root
container	Backends	../

Backends Connect to data sources

Backends come in 2 flavours:

1. Loader backends - only implement fetch method. The data is then cached in a Memory.Store on the Dataset and interacted with there. This is best for sources which just allow you to load data or where you want to load the data once and work with it locally.
2. Store backends - these support fetch, query and, if write-enabled, save. These are suitable where the backend contains a lot of data (infeasible to load locally - for examples a million rows) or where the backend has capabilities you want to take advantage of.

Backend modules must implement the following API:

{% highlight javascript %} type: 'name-of-backend' // e.g. elasticsearch

// Initial load of dataset including initial set of records fetch: function(dataset)

// Query the backend for records returning them in bulk. // This method will be used by the Dataset.query method to search the backend // for records, retrieving the results in bulk. query: function(queryObj, dataset)

// Save changes to the backend save: function(changes, dataset) {% endhighlight %}

Details of each function below. Note that:

• Each backend function takes a dataset object. This is not a Dataset object but is simple JS object representation resulting from calling Dataset.toJSON().

  It is required because the Dataset attributes contain details of specific backend (e.g. url for ElasticSearch etc).

• Each function returns a promise API object - that is something conforming to the jquery promise API and, in particular, having a done and fail function.

fetch: function(dataset)

On success, promise callback must return an object with the following structure:

{% highlight javascript %} { // (optional) Set of record data // Either an array of arrays or an array of objects corresponding to initial set of records for this object // May not provided if data only returned by query records: [...]

// (optional) Set of field data // Either an array of string or an array of objects corresponding to Field specification (see Field above) fields: { ... } // as per recline.Model.Field

// (optional) metadata fields to set on the Dataset object metadata: { title: ..., id: ... etc }

// boolean indicating whether to use a local memory store for managing this dataset useMemoryStore: } {% endhighlight %}

query: function(queryObj, dataset)

queryObj: JS object following Query specification above.

Callbacks

On success must return a 'QueryResult' object which has the following structure:

{% highlight javascript %} { // total number of results (can be null) total: ...

// one entry for each result record hits: [ { // JS object that can be used to initialize a Record object } ],

// (optional) facets: { // facet results (as per http://www.elasticsearch.org/guide/reference/api/search/facets/) } } {% endhighlight %}

The QueryResult is partially modelled on ElasticSearch - see this issue for more details.

save: function(changes, dataset)

The save function is still being revised and its API and arguments are subject to change

changes: an object with the following structure:

{% highlight javascript %} { creates: [ record.toJSON(), record.toJSON(), ... ] updates: [ ... ] deletes: [ ... ] } {% endhighlight %}

Each key has an array of records (as simple JS objects resulting from a call to Record.toJSON()) that are in that state.

The backend should take appropriate actions for each case.