[PATCH 2 of 6] configitems: add an official API for extensions to register config item

[Pierre-Yves David]

On 06/29/2017 05:17 PM, Yuya Nishihara wrote:
> On Thu, 29 Jun 2017 16:37:44 +0200, Pierre-Yves David wrote:
>> On 06/29/2017 04:29 PM, Yuya Nishihara wrote:
>>> On Thu, 29 Jun 2017 23:11:57 +0900, Yuya Nishihara wrote:
>>>> On Thu, 29 Jun 2017 15:59:17 +0200, Pierre-Yves David wrote:
>>>>> On 06/29/2017 03:48 PM, Yuya Nishihara wrote:
>>>>>> On Tue, 27 Jun 2017 15:00:31 +0200, Pierre-Yves David wrote:
>>>>>>> # HG changeset patch
>>>>>>> # User Pierre-Yves David <pierre-yves.david at octobus.net>
>>>>>>> # Date 1497700100 -7200
>>>>>>> #      Sat Jun 17 13:48:20 2017 +0200
>>>>>>> # Node ID 944f8e3191e6fa37f27d52d3c94d1ef702408ae4
>>>>>>> # Parent  714ce79885e32305c45febf0da59b2ce2093967d
>>>>>>> # EXP-Topic config.register
>>>>>>> # Available At https://www.mercurial-scm.org/repo/users/marmoute/mercurial/
>>>>>>> #              hg pull https://www.mercurial-scm.org/repo/users/marmoute/mercurial/ -r 944f8e3191e6
>>>>>>> configitems: add an official API for extensions to register config item
>>>>>>>
>>>>>>> Extensions can have a 'configtable' mapping and use
>>>>>>> 'registrar.configitem(table)' to retrieve the registration function.
>>>>>>>
>>>>>>> This behave in the same way as the other way for extensions to register new
>>>>>>> items (commands, colors, etc).
>>>>>>>
>>>>>>> diff --git a/mercurial/configitems.py b/mercurial/configitems.py
>>>>>>> --- a/mercurial/configitems.py
>>>>>>> +++ b/mercurial/configitems.py
>>>>>>> @@ -13,6 +13,11 @@ from . import (
>>>>>>>         error,
>>>>>>>     )
>>>>>>>     
>>>>>>> +def loadconfigtable(ui, extname, configtable):
>>>>>>> +    """update config item known to the ui with the extension ones"""
>>>>>>> +    for section, items in  configtable.items():
>>>>>>> +        ui._knownconfig.setdefault(section, {}).update(items)
>>>>>>
>>>>>> So this effectively updates the global configitems.coreitems, and that's
>>>>>> unavoidable since the passed ui is a temporary object.
>>>>>
>>>>> Hum, yes good catch, we are missing some copying around here. However
>>>>> you mention the ui being a temporary object can you elaborate a bit?
>>>>> (I've some notion of the weird life cycle of UI object but I'm not sure
>>>>> I follow you here.
>>>>
>>>> extensions.loadall() is called with "lui", which can be a temporary copy of
>>>> the main "ui". "lui" exists for preloading repo hgrc in dispatch layer.
>>>
>>> Anyway, I think loadconfigtable() needs to update some global table. Otherwise
>>> newly created ui object wouldn't see the defaults defined by loaded extensions
>>
>> Do we need more than ui.copy() inheriting the configtable?
> 
> Yes. hgweb may create a fresh new ui object, for example. It doesn't make much
> sense to keep config tables locally because extensions are loaded per process.

Okay I'll clarify all that in a follow up.

Thanks.

-- 
Pierre-Yves David