ArrayList had several methods that should act on a copy and return
that copy, but was instead mutating the existing list.
We cant change this behaviour in the 3.0 line for backwards compt.
reasons, but this makes the desired behavior the default, and
makes disabling the mutation in 3.1 easier
filters was a DI property that could only be set via constructor. This meant that modules couldnt add a
filter without interfering with other modules. With this change you can now add a config block like:
Injector:
RequestProcessor:
properties:
filters:
- "%$MyFilter"
Which will add a filter to RequestProcessors list of filters
Guess the fragment based on the returned HTML.
The validation error HTTP response is generated by Form rather than
the controller (LeftAndMain), so we can't set custom PJAX headers easily.
When publishing to live, DataObject#forceChange is called, which wasnt correctly loading
in fields that were lazy (unloaded) if those fields were from composite fields
like Money. The end result is that any Money values would be forced to null on
publish to live
Also changes the API of the (internal, protected) loadLazyFields method so that
not passing a class argument just unlazys all lazy fields regardless of source table
In getField we check if the field we are getting is currently lazy (unloaded), and
load it if it is. This was only working for simple fields though - composite
fields like Money werent working
Introduce new LeftAndMain_HTTPResponse class for this purpose,
to mark a response as finished regardless of HTTP status.
This is required for ajax responses which do redirects on app layer
rather than HTTP (to avoid double processing).
Specifically required to decorate LeftAndMain->init()
in the 'translatable' module (TranslatableCMSMainExtension),
which marks the response as finished through its redirect,
avoiding further processing after init().
Implemented independently of URL state to ensure that
state is retained on every user interaction, rather than
having to add it to each URL specifically.
Same reasons for not saving it as HTML5 history metadata,
as that's only inspected on history events, not
normal CMS navigation.
When one or more fragments are requested that are not in
the current DOM, we need to force loading the outermost
fragment instead (currently hardcoded to "Content").
This mainly prevents history back navigation from breaking,
e.g. admin/pages -> admin/pages/list ->
admin/pages/list/?ParentID=99 -> admin/pages/edit/show/5 -> (back)
Was using text/json, which Form->httpSubmission() interprets
(correctly) to returning JSON validation data.
This trips up the interface, which expects HTML with
the validation errors directly in the markup.
After Hamish's suggestion. Entwine onchange would not get executed in
IE8 at all, which would have the effect of the displyed dropdown
selection not being set.
Change to onadd also mandates the changes to onadd on other parts of the
tree component - otherwise the change event can trigger before the
tree elements are added to the DOM.
Space delimiter is often confused by browsers, and encoded as %20 which
breaks the shortcode system. Change to comma delimitation has already
been implemented, this is a followup cleanup.
Ref http://open.silverstripe.org/ticket/7337
Functionality that affects the values in the form better fits in
updateFromEditor function, where we expect the form to be modified.
Redraw should only affect visibility parameters.
Also added a more robust reset code, so we can always expect to get at
least a clean form, and re-added missing "target" checkbox.
Returning a link type "internal" in the situation when no link has been
detected is confusing and makes it hard to know downstream if the link
was detected or not. Switched that to null.
Also added target option to file downloads, as we don't currently have a
mechanism to default this field to "yes" for files.
