It's the recommended approach in Redux docs, and more performant (key-based).
It'll also simplify a move to Immutable.js later on.
The PHP APIs still return unkeyed data, which is standard REST API behaviour,
but for Redux state we transform it to be keyed.
Required for REST URLs which require data to be populated on call.
For example, PUT admin/campaigns/set/:id/publish to publish a campaign.
At the time when createEndpointFetcher() is called, the ":id" parameter can not be resolved,
since the "data" argument is passed in once the returned function is evaluated (since the component using it).
Since the whole idea of createEndpointFetcher() is that components shouldn't know about the backend internals,
its not the right place to interpolate urls.
I'm sending through any payload data in addition to replacing URLs, to keep the API simple for the component.
Again, it shouldn't worry about the presence URL interpolation. The trade off here is that data will be submitted
both in the payload (in case of a POST, PUT or DELETE request), as well as (optionally) in the URL.
Tests were previously excluded from eslint calls. This leads to messier
code inside tests, and potentially errors.
Now they all meet the listing rules, and so we don’t need to ignore
tests in .eslintignore.
SilverStripeBackend now has a createEndpointFetcher method, described
in its docblock in more detail.
It’s based the asset-admin code for apiCallerFromEndpoint(). It’s
refactored and generalised into framework in order to provide a
general-purpose API for tying Javascript components to backend APIs.
@todo: tests
FIX: Ensure post has correct content-type header.
Exporting silverstripe-backend lets other modules (such as asset-admin)
use it.
In addition, the Content-type header of a post request of URL-encoded
data was being set to text/plain by default, which isn’t correct and
stopped PHP from interpreting it.
Renaming state operations from 'campaign' to 'record'.
Implemented API endpoint retrieval of GridField data.
Added more mock data into CampaignAdmin (rather than hardcoding in client),
to be replaced by CampaignAdmin API endpoint querying the real datamodel.
Using more native isomorphic-fetch instead of jQuery.ajax
to minimise dependencies and get into a more forward-thinking API.
Also catching errors in ReactJS API backend:
Emulate jQuery.ajax() behaviour. Might change at a later point
if we have a general purpose backend with a promise-based catch()
implementation.
Keyed by URL instead of anonymous object maps which need to be iterated on
Removed the 'schema.forms' namespace, unnecessary since all 'schema' items should be forms
Remove once we have a schema-driven GridField in place.
Edit form should be a GridFieldDetailForm, but there's more work to do for schema generation
from this component.
jQuery uses encodeURIComponent, but also this line here:
return s.join( "&" ).replace( r20, "+" );
https://github.com/jquery/jquery/blob/1.7.2/src/ajax.js#L797
This breaks when passing the result through $.path.addSearchParams(),
which in turn uses jQuery.param - any '%20' would've been wrongly encoded as '+' by jQuery.serialise(),
which in turn gets rightly encoded as '%2B' by jQuery.param().
This had the effect of breaking CMS searches with spaces in them.
