|
|
|
| Re: Command / Query / Advice (r10662) [message #487225 is a reply to message #487213] |
Tue, 22 September 2009 13:53   |
|
On 09/22/2009 03:19 PM, Henrik Lindberg wrote:
> This version is very much simplified as it uses a generic mechanism for
> advice. Each advice is simply an advice to set a value of a particular
> type (string, boolean, version, etc) with an escape mechanism to
> instantiate any type.
>
I think one important IAdvice implementation will be the FilterAdvice. It should allow you to and/or/replace a filter to
an existing filter. The replace would consist of two filters. Old (the part to be replaced) and new.
- thomas
|
|
|
|
|
| Re: Command / Query / Advice (r10662) [message #487315 is a reply to message #487290] |
Tue, 22 September 2009 18:10 |
|
On 09/22/2009 06:32 PM, Henrik Lindberg wrote:
> Thomas Hallgren wrote:
>
>> I think one important IAdvice implementation will be the FilterAdvice.
>> It should allow you to and/or/replace a filter to an existing filter.
>> The replace would consist of two filters. Old (the part to be
>> replaced) and new.
>>
>
> I added filter advice, but now with specification of the part to replace
> - that would need to be some sort of pattern matching - can't see a
> specification of a Filter being used (remember that we support having a
> combination of OSGi and pattern filters.
>
Can't that be resolved by using two distinct IAdvice's. An OSGiFilterAdvice and a PatternFilterAdvice.
> Think it is too complicated to
> specify the part to replace - can just as well re-specify the filter as
> it is supposed to be...
>
> If the intent is to negate part of the filter, an AND with a NOT + the
> normalization should do the trick don't you think?
>
No, not really. (!a AND a) will always yield false. Simply removing a will not.
What's wrong with using two filters for a replace? One is the filter to remove. The other is its replacement. Simple and
easy to grok.
- thomas
|
|
|
| Re: Command / Query / Advice (r10662) [message #487318 is a reply to message #487315] |
Tue, 22 September 2009 18:24 |
|
On 09/22/2009 08:10 PM, Thomas Hallgren wrote:
>> Think it is too complicated to
>> specify the part to replace - can just as well re-specify the filter as
>> it is supposed to be...
>>
There's also the case when you don't know what the final filter is supposed to be. You just know that if a certain
filter is there, it should be replaced by something else. You don't care or have knowledge about other parts of the filter.
- thomas
|
|
|
|
|
|
| Re: Command / Query / Advice (r10662) [message #577739 is a reply to message #487213] |
Tue, 22 September 2009 13:53 |
|
On 09/22/2009 03:19 PM, Henrik Lindberg wrote:
> This version is very much simplified as it uses a generic mechanism for
> advice. Each advice is simply an advice to set a value of a particular
> type (string, boolean, version, etc) with an escape mechanism to
> instantiate any type.
>
I think one important IAdvice implementation will be the FilterAdvice. It should allow you to and/or/replace a filter to
an existing filter. The replace would consist of two filters. Old (the part to be replaced) and new.
- thomas
|
|
|
|
|
| Re: Command / Query / Advice (r10662) [message #577769 is a reply to message #487290] |
Tue, 22 September 2009 18:10 |
|
On 09/22/2009 06:32 PM, Henrik Lindberg wrote:
> Thomas Hallgren wrote:
>
>> I think one important IAdvice implementation will be the FilterAdvice.
>> It should allow you to and/or/replace a filter to an existing filter.
>> The replace would consist of two filters. Old (the part to be
>> replaced) and new.
>>
>
> I added filter advice, but now with specification of the part to replace
> - that would need to be some sort of pattern matching - can't see a
> specification of a Filter being used (remember that we support having a
> combination of OSGi and pattern filters.
>
Can't that be resolved by using two distinct IAdvice's. An OSGiFilterAdvice and a PatternFilterAdvice.
> Think it is too complicated to
> specify the part to replace - can just as well re-specify the filter as
> it is supposed to be...
>
> If the intent is to negate part of the filter, an AND with a NOT + the
> normalization should do the trick don't you think?
>
No, not really. (!a AND a) will always yield false. Simply removing a will not.
What's wrong with using two filters for a replace? One is the filter to remove. The other is its replacement. Simple and
easy to grok.
- thomas
|
|
|
| Re: Command / Query / Advice (r10662) [message #577788 is a reply to message #487315] |
Tue, 22 September 2009 18:24 |
|
On 09/22/2009 08:10 PM, Thomas Hallgren wrote:
>> Think it is too complicated to
>> specify the part to replace - can just as well re-specify the filter as
>> it is supposed to be...
>>
There's also the case when you don't know what the final filter is supposed to be. You just know that if a certain
filter is there, it should be replaced by something else. You don't care or have knowledge about other parts of the filter.
- thomas
|
|
|
|
Powered by
FUDForum. Page generated in 0.03337 seconds
] 
Search
Help
Register
Login
Home
