A bit late to the game but can't we pass a cannonicalized absolute string to
the JS ?
eg: wiki:space.sp\.ace.sp\\ace.page
and then just warn the js devs that they should use functions to manipulate the
string representation instead of hacking it manually ?
That's what nodejs does for URLs and I will say without reservation that the
user-friendlyness of their their URL parsing API is not even compatible to the
disaster that was java.net.URL.
On 26/08/15 17:36, Guillaume "Louis-Marie" Delhumeau wrote:
I don't really understand the solution 2.
We already have an EntityReference class in entityReference.js, with a
string serializer and a string parser.
If we use a JSON format to describe a reference, we will still have to
parse it to create an EntityReference object, and this parser does not
exist yet.
So what is the benefit?
I only see drawbacks:
- A JSON reference is a lot more verbose.
- From the java side, we need to serialize the reference to a JSON
representation which is probably more costly than serializing to a string.
- We use string representations in a lot of places, and suddenly we should
use JSON but only when we do JavaScript...
So +1 for the solution 1.
In any case, it seems a good idea to use references in JS code.
Thanks,
Guillaume
2015-08-06 14:25 GMT+02:00 Eduard Moraru <[email protected]>:
I agree with Thomas that it would be best to use EntityReference -> JSON as
much as possible (when passing the reference from velocity to javascript,
server-side), however, man times when you need a reference you also end up
building a new reference (on the client-side, i.e. javascript) so we can`t
really avoid having a strong javascript EntityReference API +
serializer/resolver.
Also, making a HTTP request from JS to resolve a string reference on the
server (and getting a JSON result) is not really an option IMO (specially
if you do that a lot).
So +1 to using reference in javascript code.
Thanks,
Eduard
On Thu, Jul 30, 2015 at 11:50 AM, Thomas Mortagne <
[email protected]
wrote:
Direct EntityReference -> JSON serialisation would provide a strong
standard (we already have tool to make sure we never break
EntityReference) and it would make JS and Java sides more consistent.
Now In some cases we will still need to support parsing a String
reference in JS I think.
On Thu, Jul 30, 2015 at 10:38 AM, [email protected] <[email protected]
wrote:
Hi Marius/all,
On 30 Jul 2015 at 10:26:49, Marius Dumitru Florea (
[email protected](mailto:[email protected]))
wrote:
+1
Another option could be to pass a JSON serialization of the reference.
Thomas has recently fixed some issues that were preventing JSON
serialization of an entity reference. It should work fine now:
$jsontool.serialize($documentReference)
should produce something close to:
{
name: 'Page',
type: 'DOCUMENT',
parent: {
name: 'Space2',
type: 'SPACE',
parent: {
...
}
}
}
It's more verbose obviously. For this we need to modify a bit
entityReference.js though, because it expects the entity type to be an
int.
Indeed there are only 2 global solutions:
* Solution 1: Pass the seralized String reference and have the js
perform the resolve. This also means that the java code may need to do a
serialize. So this has the drawback of doing a serialize + resolve. It
has
the advantage of being a one-liner from JS.
* Solution 2: Pass the individual elements of a reference so that the
JS
doesn’t have to call resolve. This can be done in several ways:
option 1:
<div class=“metadata”>
<div class=“sourcewiki”>wiki</div>
<div class=“sourcepage”>page</div>
<div class=“sourcespaces”>
<div>space1</div>
<div>space2</div>
</div>
</div>
option 2:
The JSON you suggest above. However, how would the java code pass this
into HTML? Inside a tag’s text, as in:
<div class=“metadata”>
… json here?...
</div>
Inside a <script> tag directly?
Note that option2 has a big advantage over option 1: JS knows JSON
natively and thus there’s no need to implement any parsing at the JS
level.
Compared to Solution 1, the advantage would to avoid a resolve() call
on
the JS side. On the java side we’ll still need the serialize() call
(which
will possibly take slightly longer than the string serialization).
Overall Solution 2/option2 could be a better solution indeed.
What do others think?
Thanks
-Vincent
Thanks,
Marius
On Wed, Jul 29, 2015 at 4:34 PM, [email protected] wrote:
Hi devs,
In the past we use to pass the wiki, space and page as 3 variables
to
JS code. For example in dashboard.js we currently have:
this.sourcePage = this.element.down('.metadata
.sourcepage').innerHTML;
this.sourceSpace = this.element.down('.metadata
.sourcespace').innerHTML;
this.sourceWiki = this.element.down('.metadata
.sourcewiki').innerHTML;
Now that we need to handle Nested Spaces, it’s more complex and I’m
proposing to change our best practice and instead to pass a full
reference,
as in:
this.sourceReference =
XWiki.Model.resolve(this.element.down('.metadata .source').innerHTML,
XWiki.EntityType.DOCUMENT);
More generally the idea would be to do the same as we do in Java
code, i.e. to start stopping passing several parameters to functions and
instead to use XWiki.EntityReference (or XWiki.DocumentReference, etc).
For example for dashboard.js this allows to replace:
var link = new Element('a', {'href' : this.sourceURL});
link.update(this.sourceWiki + ':' + this.sourceSpace + '.' +
this.sourcePage);
With:
var link = new Element('a', {'href' : this.sourceURL});
link.update(XWiki.Model.serialize(this.sourceReference));
WDYT?
Thanks
-Vincent
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs
--
Thomas Mortagne
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs
