On Fri, Jan 20, 2012 at 10:24 AM, Paul Libbrecht <[email protected]> wrote:
Thomas,
Le 19 janv. 2012 à 12:06, Thomas Mortagne a écrit :
A deprecation mechanism should be able to support the developer into actualizing to the right keys, chosen from a finite set of suggestions (the new keys).
Examples would be nice ;)
Think of redoing a part of "my blabla profile" where "my contributions" are displayed so that it can speak to the third person when the user is not the same: you'd start with a copy of the profile velocity views (pages, vm files), then modify them progressively.
The keys, maybe they were prefixed with "myblabla" would be refactored to be either with "ownblaba" and "othersblabla". Ideally, this should be done in one shot, so no real need of deprecation, but I do not think this can be ensured.
Do you mean that the old key is refactored in severals ? In this cases as a key cannot contains a white space in Properties specs we could have:
#@deprecated new.key.subname1 new.key.subname2 old.key.name=Default translation
Fits me perfectly!
The encoding is bound to the target usages and vice versa... Not sure I understand what you mean.
that's philosophical: I meant we need to envision the usages to make sure the knowledge representation is good engouh.
The restricted view of a key-renaming appears to me as something that could be easily automated leading to an automatic refactoring of the sources. Other deprecation mechanisms cannot be so: the deprecation without replacement can only lead to a developer warning, the deprecation with multiple alternatives can only lead to a warning with suggestion...
Sure it's hard to automate translation copy to the new key when there is 0 or more that one new keys. Do you mean that we need a syntax advanced enough to explain to the l10n wiki or any other tool how to cut the key in peaces to put it in the right keys ?
I did mean this indeed.
I'm just trying to see the big picture of the future before settling on the knowledge representation...
Sure, that's why this proposal is for: make sure we go in the right direction. But does not mean we have to do everything at once, just that we need to make sure we don't block the future ;)
We're on the same wavelength.
I think at this point, considering all the cases we've considered, I can say:
+1.
Ok thanks. I will go for #@deprecated new.key.subname1 new.key.subname2 old.key.name=Default translation as first implementation. I will also make sure of the following: * the type of syntax is indicated by a {syntaxid} at the beginning, default being a list of new keys * skip unknown syntax type without complaining. So it allows a large choice of syntaxes to extend the current syntax without breaking anything. For example: #@deprecated {regexp} match="(.*) and (.*)" new.key.subname1=1 new.key.subname2=2 old.key.name=Default translation WDYT ?
I would lover, however, to see what tools are used that use the deprecation.
As I said the first one will be l10n wiki: the goal is to make sure we don't loose translation assigned to the old key when renaming it. When l10n import the default language and see a new deprecated key it's going to copy the translations assigned t the old key on the new key so that when you export again translation from l10n wiki all the translations are now both on the old and new key. On XWiki itself we can imagine printing warning indicating which key to use instead when some code try to get deprecated key translation and even get the translation from the new key instead.
paul _______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs
-- Thomas Mortagne