Hi all,
a bit late to the party, I just found out today that $msg.get is now
deprecated.
On 03/15/2013 01:45 PM, Vincent Massol wrote:
Hi,
On Mar 15, 2013, at 1:21 PM, Eduard Moraru<[email protected]> wrote:
Are we sure we want to drop the $msg binding in the future [1]?
IMO yes, for consistency, but it's a good question to ask. It's not just about $msg but
about all bindings. If we allow "shortcut aliases for bindings" we need to make
this general and allow any services binding to offer shortcuts.
Well... it's complicated. We would also remove $xwiki in the future?
maybe $xcontext as well? it will all be prefixed with $services?
I think that while it's a good idea for code beauty reasons, it's highly
impractical. Script authors are already used with $msg.get(), the new
$services.localization.render() does at least the same thing as $msg.get
(or doesn't it?), it's longer to write **and to read** ("services"
itself is longer than "msg.get" so no matter what you put after, it's
already longer) , and it will generate tons and tons and tons and tons
and tons (I cannot emphasize this enough) of logs saying that script X
is using a deprecated function, until you change all your scripts and
all the extensions on extensions.xwiki.org so that when you install
something your logs are still readable (or maybe we can disable these logs?)
One reason we introduced $services was to keep the xcontext namespace free for
applications to use.
Compared to other services, $services.localization would be used heavily
I had personally proposed "l10n" to make it a bit shorter but it wasn't
accepted. My proposal was: $services.l10n
inside scripts and we would basically have 2 options:
1. use $services.localization.render('key') (you fall asleep writing this
every time)
2. always declare a variable in your script like #set ($keys =
$services.localization) and then do $keys.render('key')
3. Use the translation macro if you're in wiki pages: {{translation key="…"/}}
This is, imho, a little quarter of a solution. In practice, html macro
with wiki=false is used far more than we'd like.
Now, sure, we can choose to ignore that, but that would be another
discussion.
Also, the problem of all 3 solutions above is that they require
refactoring of a lot of code (I for one use a lot the localization tool,
not only for localization, but also to make it easy for people that
don't know code to decide what the interface of the application should say).
AFAIK, we have already considered in the past that a similarly used service
like $services.model is already becoming a bit annoying having to write the
oh-so-very-long "$services.model.createDocumentReference(wiki,
space,name)"; do we want to get the same effect with the new localization
module?
@Edy, we really need to commit the auto completion feature ;)
I understand and agree that the new localization module is more powerful
and flexible than the XWikiMessageTool, but I feel that those new features
are not required in day to day use and unnecessarily crowd/pollute the
regularly used API.
You're loosing me here. I thought you were only talking about the binding name
and here you start to say that you want to go back to XWikiMessageTool?
This transition should IMO be smoother
Smoother than what?
I think what Edi means, and I agree with him, is that he doesn't really
understand why can't we just have the old $msg.get function call the new
localization service and be done with it?
and with less impact than what is
currently being proposed. If the new localization module can provide all
the features of the XWikiMessageTool, then I propose that we simply reuse
the $msg binding as it is and, in the background, transform
XWikiMessageTool into a facade (as I see we have already pretty much done
to preserve backwards compatibility), but agree that we *keep* $msg as the
simplified translations binding, to be used in day to day operations and,
for more complex tasks, the dev needs to use $services.localization instead.
I don't understand what you mean. If your idea is to get $msg.get() rather than
$msg.render() which is the new API then I don't agree. If what you propose is just to
have a generic "script service alias" notion where $msg would be the same as
$services.localization then I'm more inclined to think about it :)
See above.
BTW I would not want to reuse the same name ($msg) because it's misleading.
Why would that be misleading? Isn't the new localization service an
improvement of the old one? It's a new tool?
Since I contribute less and less, I am now more a user of the xwiki
platform than a developer of it, and honestly, as a user, I perceive the
work that was done on localization as an improvement of what was there
before and I don't like the idea that for an improvement of the platform
I use I need to rewrite all my code, and use a new, longer, function
call, and learn new function names et all: as a script author, I'm
actually doing the same thing but now I shuld change my whole code to do
the same thing in a different way, because the platform was improved.
Basically, I`m proposing to apply the KIS(s) principle. :)
Another option is to not change anything at the level of java (i.e. no "script
service alias") but in xwikivars.vm add a velocity variable named $l10n:
#set ($l10n = $services.localization)
if that set could be
#set($msg.get = $services.localization.render)
then sure, I agree with this solution.
Thanks,
Anca
This is basically your option 2) above but made generic by putting it in
xwikivars.vm.
Thanks
-Vincent
WDYT?
Thanks,
Eduard
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs
_______________________________________________
devs mailing list
[email protected]
http://lists.xwiki.org/mailman/listinfo/devs
