Hi,

I really do not get the tone of these replies, but perhaps my mail was not
clear. So, let me start over.

You asked for who is interested in joining co-coding, and I said that I am
interested particularly in the area of the logger.

Specifically, I would be interested in splitting the Log class into two:
- a Log class holding the current class side methods
- a LogEvent (or LogEntry, or LogItem or whatever other name indicating the
result of a single log statement) that holds the instance side of the
current Log class.

As it is now, when I execute:
Log critical: 'message'

I get an instance of Log. This is confusing, and not needed. We can nicely
get an instance of LogEvent. This will not complicate things at all.

Next, I said that as we already have a Zinc logging mechanism, I would be
interested in consolidating Zinc to use the new mechanism. This would also
provide a case study for having explicit classes of LogEvents and see the
implications of that.

And finally, I mentioned the Fuel serialization. My idea was to offer a
cheap way to serialize a log into a file by serializing each log entry as a
fuel object, but I did not know that we already have a JSON serialization.
So, this is not that much of a priority anymore.

I will propose code for the splitting.

Doru


On Wed, May 21, 2014 at 9:05 AM, stepharo <[email protected]> wrote:

>
>
>>>> As a general matter, I would be interested in:
>>>> - unifying with the ZnLogEvent
>>>>
>>>
>>>  SystemLogger is infrastructure low level. So I do not see why Zn and
> SystemLogger should be unified.
>
>
>  If this means talking to Sven this is a definite yes. I don’t think we
>> need to unify with the current ZnLogEvent. Syncing with zinc is probably
>> not the way to go anyway. Zinc is still an external framework (hence the
>> namespace prefix). I did a zinc bridge which operates an a single event
>> bridging from zinc to SystemLogger. This is not clever performance wise. A
>> better approach is to plug into zincs LogSupport. We should hook into there
>> so we just get the invocations from zinc and we can emit the proper
>> SystemLogger events into the Log framework.
>> Sven will come up with something announcement based. And that is still a
>> task for us, too. To look where functionality of SystemLogger is doubled
>> with announcements.
>>
>
> We should pay attention that it stays small and simple.
>
>
>
>  first we should integrate it and make sure that all the Transcript show
>>> are funnelled to the Log frameworks.
>>>
>>
>> Yes. Maybe we need to sync our understanding of log levels first. The
>> exchange adds a log level and adding a tag would be good, too.
>>
>
> I do not get it but this is not a problem. No need to reply.
>
>
>  - Split the Log class
>>>>
>>>
>>> we will not split it because we iterated over the design since 8 months
>>> with norbert :)
>>>
>>
>> I cannot find the proposals of Doru. But I’m interested. I splitted the
>> Log class already into BasicLog and Log. I’m not fully satisfied with it so
>> … My concern about the Log class is more how get it right supporting
>> different kind of Log classes and still have a working filtering in the
>> Logger. But that we can discuss in the groups that form I guess
>>
> I would like to stop iterating. And again we should keep the system small.
> I studied all the logger framework available and I think that we can look
> for the graal forever or get done and I want to get done
> because there are other many many topics that can get fixed.
> Right now we went from 0% to 80% and this is a good ratio. We iterated
> with you. Now I do not want to pass time to go to an hypothetical 85%.
>
>
>
>  - Possibly support Fuel serialization for each log entry
>>>>
>>> Please go ahead.
>>>
>>
>> Yes. I don’t even understand what it means :)
>>
> Send code!
>
>
>


-- 
www.tudorgirba.com

"Every thing has its own flow"

Reply via email to