Re: Getter method for retrieving the attributes of an appender from the LoggerContext
I agree. It seems like we're preparing to do a lot of work to essentially enable users to duplicate our unit tests... Sent from my iPhone > On 2016/01/28, at 0:44, Mikael Ståldal wrote: > > OK, then the keeping config nodes approach might not be a good idea. > > However, I still don't think that the benefit of being able to inspect > appender's config justifies the cost of increasing the complexity of every > appender (including future ones and 3rd party plugins). > > On Wed, Jan 27, 2016 at 4:22 PM, Apostolis Giannakidis < > ap.giannaki...@gmail.com> wrote: > >> One use case that I have at hand at the moment is that I want to be able >> to verify that my appenders have the expected configuration attributes. For >> example, I would like to be able to verify that my syslog appender is >> connecting to the expected host,port,protocol, etc. >> >> On Wed, Jan 27, 2016 at 3:19 PM, Mikael Ståldal >> wrote: >> >>> It would be useful if Apostolis Giannakidis can explain the use case >>> behind this request, now it is a bit abstract to me. >>> On Wed, Jan 27, 2016 at 4:13 PM, Matt Sicker wrote: I mean keeping the Node tree to get attributes. It would only work from config files (and the config builder classes). We get questions every so often about modifying the config programmatically which would either need to maintain more Nodes or just be unsupported. On 27 January 2016 at 09:09, Mikael Ståldal wrote: > I don't quite understand what you mean. > >> On Wed, Jan 27, 2016 at 4:02 PM, Matt Sicker wrote: >> >> That sounds a little fragile as some people either create or modify their >> creation directly from the plugin factories. >> >> On 27 January 2016 at 07:05, Mikael Ståldal < mikael.stal...@magine.com> >> wrote: >> >>> Then perhaps we should keep the node tree and expose it for this kind > of >>> queries, something like this: >>> >>> String hostname = loggerContext.getConfiguration(). >>> getAttributesForAppender("syslogAppender").get("host"); >>> >>> This would require a new method in >>> org.apache.logging.log4j.core.config.Configuration: >>> >>> public Map getAttributesForAppender(String > appenderName); >>> >>> >>> >>> On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers < > ralph.go...@dslextreme.com >>> >>> wrote: >>> While I understand your point, the node tree is discarded after the plugins are created. We would have to keep it around for this to > work. Furthermore, each component would need to have a reference to its corresponding node, which we obviously don't currently do. Ralph > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal < >> mikael.stal...@magine.com wrote: > > To me it does not seems good to force all Appender implementations > to > implement this. Especially not since the next logical step would > then >>> be to > do the same with other components such as Layouts. That would be a >> lot >>> of > work in total, and also add more work for all future components, including > 3rd party plugins. > > I think it makes more sense, and would be less work in total, if > the > configuration system would store and expose those attributes > without > involving the components themselves. > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory < >> garydgreg...@gmail.com> > wrote: > >> Apostolis, >> >> You are warmly welcome to contribute to Log4j. You can create a > JIRA >>> and >> attach a patch in unified diff file format. Unit tests as part of >> the patch >> are a must IMO. Feel free to flush out any design or > implementation here on >> the dev ML. >> >> Thank you! >> Gary >> >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < ralph.go...@dslextreme.com> >> wrote: >> >>> All the attributes have to have String representations to be > usable >>> in >> the >>> XML, JSON & Properties configurations. Yes, the Map could contain Objects >>> but then every one of them has to be cast to its real object to > be >> usable. >>> >>> The map should be read-only because modifying its contents would >> have no >>> effect on the appender. >>> >>> The map should not be stored in an ivar but constructed whenever >> the >>> attributes are retrieved. Otherwise it will be temping to just > keep them >> in >>
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
OK, then the keeping config nodes approach might not be a good idea. However, I still don't think that the benefit of being able to inspect appender's config justifies the cost of increasing the complexity of every appender (including future ones and 3rd party plugins). On Wed, Jan 27, 2016 at 4:22 PM, Apostolis Giannakidis < ap.giannaki...@gmail.com> wrote: > One use case that I have at hand at the moment is that I want to be able > to verify that my appenders have the expected configuration attributes. For > example, I would like to be able to verify that my syslog appender is > connecting to the expected host,port,protocol, etc. > > On Wed, Jan 27, 2016 at 3:19 PM, Mikael Ståldal > wrote: > >> It would be useful if Apostolis Giannakidis can explain the use case >> behind this request, now it is a bit abstract to me. >> >> On Wed, Jan 27, 2016 at 4:13 PM, Matt Sicker wrote: >> >>> I mean keeping the Node tree to get attributes. It would only work from >>> config files (and the config builder classes). We get questions every so >>> often about modifying the config programmatically which would either need >>> to maintain more Nodes or just be unsupported. >>> >>> On 27 January 2016 at 09:09, Mikael Ståldal >>> wrote: >>> >>> > I don't quite understand what you mean. >>> > >>> > On Wed, Jan 27, 2016 at 4:02 PM, Matt Sicker wrote: >>> > >>> > > That sounds a little fragile as some people either create or modify >>> their >>> > > creation directly from the plugin factories. >>> > > >>> > > On 27 January 2016 at 07:05, Mikael Ståldal < >>> mikael.stal...@magine.com> >>> > > wrote: >>> > > >>> > > > Then perhaps we should keep the node tree and expose it for this >>> kind >>> > of >>> > > > queries, something like this: >>> > > > >>> > > > String hostname = loggerContext.getConfiguration(). >>> > > > getAttributesForAppender("syslogAppender").get("host"); >>> > > > >>> > > > This would require a new method in >>> > > > org.apache.logging.log4j.core.config.Configuration: >>> > > > >>> > > > public Map getAttributesForAppender(String >>> > appenderName); >>> > > > >>> > > > >>> > > > >>> > > > On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers < >>> > ralph.go...@dslextreme.com >>> > > > >>> > > > wrote: >>> > > > >>> > > > > While I understand your point, the node tree is discarded after >>> the >>> > > > > plugins are created. We would have to keep it around for this to >>> > work. >>> > > > > Furthermore, each component would need to have a reference to its >>> > > > > corresponding node, which we obviously don't currently do. >>> > > > > >>> > > > > Ralph >>> > > > > >>> > > > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal < >>> > > mikael.stal...@magine.com >>> > > > > >>> > > > > wrote: >>> > > > > > >>> > > > > > To me it does not seems good to force all Appender >>> implementations >>> > to >>> > > > > > implement this. Especially not since the next logical step >>> would >>> > then >>> > > > be >>> > > > > to >>> > > > > > do the same with other components such as Layouts. That would >>> be a >>> > > lot >>> > > > of >>> > > > > > work in total, and also add more work for all future >>> components, >>> > > > > including >>> > > > > > 3rd party plugins. >>> > > > > > >>> > > > > > I think it makes more sense, and would be less work in total, >>> if >>> > the >>> > > > > > configuration system would store and expose those attributes >>> > without >>> > > > > > involving the components themselves. >>> > > > > > >>> > > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory < >>> > > garydgreg...@gmail.com> >>> > > > > > wrote: >>> > > > > > >>> > > > > >> Apostolis, >>> > > > > >> >>> > > > > >> You are warmly welcome to contribute to Log4j. You can create >>> a >>> > JIRA >>> > > > and >>> > > > > >> attach a patch in unified diff file format. Unit tests as >>> part of >>> > > the >>> > > > > patch >>> > > > > >> are a must IMO. Feel free to flush out any design or >>> > implementation >>> > > > > here on >>> > > > > >> the dev ML. >>> > > > > >> >>> > > > > >> Thank you! >>> > > > > >> Gary >>> > > > > >> >>> > > > > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < >>> > > > > ralph.go...@dslextreme.com> >>> > > > > >> wrote: >>> > > > > >> >>> > > > > >>> All the attributes have to have String representations to be >>> > usable >>> > > > in >>> > > > > >> the >>> > > > > >>> XML, JSON & Properties configurations. Yes, the Map could >>> contain >>> > > > > Objects >>> > > > > >>> but then every one of them has to be cast to its real object >>> to >>> > be >>> > > > > >> usable. >>> > > > > >>> >>> > > > > >>> The map should be read-only because modifying its contents >>> would >>> > > have >>> > > > > no >>> > > > > >>> effect on the appender. >>> > > > > >>> >>> > > > > >>> The map should not be stored in an ivar but constructed >>> whenever >>> > > the >>> > > > > >>> attributes are retrieved. Otherwise it will be temping to >>> just >>> > keep >>> > > > > them >>> > > > > >> in >>> > > > > >>> a map an
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
One use case that I have at hand at the moment is that I want to be able to verify that my appenders have the expected configuration attributes. For example, I would like to be able to verify that my syslog appender is connecting to the expected host,port,protocol, etc. On Wed, Jan 27, 2016 at 3:19 PM, Mikael Ståldal wrote: > It would be useful if Apostolis Giannakidis can explain the use case > behind this request, now it is a bit abstract to me. > > On Wed, Jan 27, 2016 at 4:13 PM, Matt Sicker wrote: > >> I mean keeping the Node tree to get attributes. It would only work from >> config files (and the config builder classes). We get questions every so >> often about modifying the config programmatically which would either need >> to maintain more Nodes or just be unsupported. >> >> On 27 January 2016 at 09:09, Mikael Ståldal >> wrote: >> >> > I don't quite understand what you mean. >> > >> > On Wed, Jan 27, 2016 at 4:02 PM, Matt Sicker wrote: >> > >> > > That sounds a little fragile as some people either create or modify >> their >> > > creation directly from the plugin factories. >> > > >> > > On 27 January 2016 at 07:05, Mikael Ståldal < >> mikael.stal...@magine.com> >> > > wrote: >> > > >> > > > Then perhaps we should keep the node tree and expose it for this >> kind >> > of >> > > > queries, something like this: >> > > > >> > > > String hostname = loggerContext.getConfiguration(). >> > > > getAttributesForAppender("syslogAppender").get("host"); >> > > > >> > > > This would require a new method in >> > > > org.apache.logging.log4j.core.config.Configuration: >> > > > >> > > > public Map getAttributesForAppender(String >> > appenderName); >> > > > >> > > > >> > > > >> > > > On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers < >> > ralph.go...@dslextreme.com >> > > > >> > > > wrote: >> > > > >> > > > > While I understand your point, the node tree is discarded after >> the >> > > > > plugins are created. We would have to keep it around for this to >> > work. >> > > > > Furthermore, each component would need to have a reference to its >> > > > > corresponding node, which we obviously don't currently do. >> > > > > >> > > > > Ralph >> > > > > >> > > > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal < >> > > mikael.stal...@magine.com >> > > > > >> > > > > wrote: >> > > > > > >> > > > > > To me it does not seems good to force all Appender >> implementations >> > to >> > > > > > implement this. Especially not since the next logical step would >> > then >> > > > be >> > > > > to >> > > > > > do the same with other components such as Layouts. That would >> be a >> > > lot >> > > > of >> > > > > > work in total, and also add more work for all future components, >> > > > > including >> > > > > > 3rd party plugins. >> > > > > > >> > > > > > I think it makes more sense, and would be less work in total, if >> > the >> > > > > > configuration system would store and expose those attributes >> > without >> > > > > > involving the components themselves. >> > > > > > >> > > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory < >> > > garydgreg...@gmail.com> >> > > > > > wrote: >> > > > > > >> > > > > >> Apostolis, >> > > > > >> >> > > > > >> You are warmly welcome to contribute to Log4j. You can create a >> > JIRA >> > > > and >> > > > > >> attach a patch in unified diff file format. Unit tests as part >> of >> > > the >> > > > > patch >> > > > > >> are a must IMO. Feel free to flush out any design or >> > implementation >> > > > > here on >> > > > > >> the dev ML. >> > > > > >> >> > > > > >> Thank you! >> > > > > >> Gary >> > > > > >> >> > > > > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < >> > > > > ralph.go...@dslextreme.com> >> > > > > >> wrote: >> > > > > >> >> > > > > >>> All the attributes have to have String representations to be >> > usable >> > > > in >> > > > > >> the >> > > > > >>> XML, JSON & Properties configurations. Yes, the Map could >> contain >> > > > > Objects >> > > > > >>> but then every one of them has to be cast to its real object >> to >> > be >> > > > > >> usable. >> > > > > >>> >> > > > > >>> The map should be read-only because modifying its contents >> would >> > > have >> > > > > no >> > > > > >>> effect on the appender. >> > > > > >>> >> > > > > >>> The map should not be stored in an ivar but constructed >> whenever >> > > the >> > > > > >>> attributes are retrieved. Otherwise it will be temping to just >> > keep >> > > > > them >> > > > > >> in >> > > > > >>> a map and not as individual attributes, which would cause >> > problems. >> > > > > >>> >> > > > > >>> If you have nesting such as MyAppender extends MyBaseAppender >> > > > extends >> > > > > >>> AbstractOutputStreamAppender extends AbstractAppender, I >> > envision a >> > > > > >>> fillAttributes method at each “level” that fills in the >> > attributes >> > > it >> > > > > >> knows >> > > > > >>> about, so fillAttributeMap(map) should always call >> > > > > >>> super.fillAttributeMap(map) - except in AbstractAppender of >> > cour
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
It would be useful if Apostolis Giannakidis can explain the use case behind this request, now it is a bit abstract to me. On Wed, Jan 27, 2016 at 4:13 PM, Matt Sicker wrote: > I mean keeping the Node tree to get attributes. It would only work from > config files (and the config builder classes). We get questions every so > often about modifying the config programmatically which would either need > to maintain more Nodes or just be unsupported. > > On 27 January 2016 at 09:09, Mikael Ståldal > wrote: > > > I don't quite understand what you mean. > > > > On Wed, Jan 27, 2016 at 4:02 PM, Matt Sicker wrote: > > > > > That sounds a little fragile as some people either create or modify > their > > > creation directly from the plugin factories. > > > > > > On 27 January 2016 at 07:05, Mikael Ståldal > > > > wrote: > > > > > > > Then perhaps we should keep the node tree and expose it for this kind > > of > > > > queries, something like this: > > > > > > > > String hostname = loggerContext.getConfiguration(). > > > > getAttributesForAppender("syslogAppender").get("host"); > > > > > > > > This would require a new method in > > > > org.apache.logging.log4j.core.config.Configuration: > > > > > > > > public Map getAttributesForAppender(String > > appenderName); > > > > > > > > > > > > > > > > On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers < > > ralph.go...@dslextreme.com > > > > > > > > wrote: > > > > > > > > > While I understand your point, the node tree is discarded after the > > > > > plugins are created. We would have to keep it around for this to > > work. > > > > > Furthermore, each component would need to have a reference to its > > > > > corresponding node, which we obviously don't currently do. > > > > > > > > > > Ralph > > > > > > > > > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal < > > > mikael.stal...@magine.com > > > > > > > > > > wrote: > > > > > > > > > > > > To me it does not seems good to force all Appender > implementations > > to > > > > > > implement this. Especially not since the next logical step would > > then > > > > be > > > > > to > > > > > > do the same with other components such as Layouts. That would be > a > > > lot > > > > of > > > > > > work in total, and also add more work for all future components, > > > > > including > > > > > > 3rd party plugins. > > > > > > > > > > > > I think it makes more sense, and would be less work in total, if > > the > > > > > > configuration system would store and expose those attributes > > without > > > > > > involving the components themselves. > > > > > > > > > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory < > > > garydgreg...@gmail.com> > > > > > > wrote: > > > > > > > > > > > >> Apostolis, > > > > > >> > > > > > >> You are warmly welcome to contribute to Log4j. You can create a > > JIRA > > > > and > > > > > >> attach a patch in unified diff file format. Unit tests as part > of > > > the > > > > > patch > > > > > >> are a must IMO. Feel free to flush out any design or > > implementation > > > > > here on > > > > > >> the dev ML. > > > > > >> > > > > > >> Thank you! > > > > > >> Gary > > > > > >> > > > > > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < > > > > > ralph.go...@dslextreme.com> > > > > > >> wrote: > > > > > >> > > > > > >>> All the attributes have to have String representations to be > > usable > > > > in > > > > > >> the > > > > > >>> XML, JSON & Properties configurations. Yes, the Map could > contain > > > > > Objects > > > > > >>> but then every one of them has to be cast to its real object to > > be > > > > > >> usable. > > > > > >>> > > > > > >>> The map should be read-only because modifying its contents > would > > > have > > > > > no > > > > > >>> effect on the appender. > > > > > >>> > > > > > >>> The map should not be stored in an ivar but constructed > whenever > > > the > > > > > >>> attributes are retrieved. Otherwise it will be temping to just > > keep > > > > > them > > > > > >> in > > > > > >>> a map and not as individual attributes, which would cause > > problems. > > > > > >>> > > > > > >>> If you have nesting such as MyAppender extends MyBaseAppender > > > > extends > > > > > >>> AbstractOutputStreamAppender extends AbstractAppender, I > > envision a > > > > > >>> fillAttributes method at each “level” that fills in the > > attributes > > > it > > > > > >> knows > > > > > >>> about, so fillAttributeMap(map) should always call > > > > > >>> super.fillAttributeMap(map) - except in AbstractAppender of > > course > > > - > > > > > and > > > > > >>> should call it before filling in its own attributes so that it > > can > > > > > >> override > > > > > >>> any values provided by the base Appenders. If the primary > > Appender > > > > > does > > > > > >>> not implement fillAttributeMap then only the attributes of its > > > super > > > > > >>> classes will be included, which is actually correct for the > > > > > >> SyslogAppender. > > > > > >>> > > > > > >>> Ralph > > > > > >>> > > > > > On Jan 26, 2016, at 5:2
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
I mean keeping the Node tree to get attributes. It would only work from config files (and the config builder classes). We get questions every so often about modifying the config programmatically which would either need to maintain more Nodes or just be unsupported. On 27 January 2016 at 09:09, Mikael Ståldal wrote: > I don't quite understand what you mean. > > On Wed, Jan 27, 2016 at 4:02 PM, Matt Sicker wrote: > > > That sounds a little fragile as some people either create or modify their > > creation directly from the plugin factories. > > > > On 27 January 2016 at 07:05, Mikael Ståldal > > wrote: > > > > > Then perhaps we should keep the node tree and expose it for this kind > of > > > queries, something like this: > > > > > > String hostname = loggerContext.getConfiguration(). > > > getAttributesForAppender("syslogAppender").get("host"); > > > > > > This would require a new method in > > > org.apache.logging.log4j.core.config.Configuration: > > > > > > public Map getAttributesForAppender(String > appenderName); > > > > > > > > > > > > On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers < > ralph.go...@dslextreme.com > > > > > > wrote: > > > > > > > While I understand your point, the node tree is discarded after the > > > > plugins are created. We would have to keep it around for this to > work. > > > > Furthermore, each component would need to have a reference to its > > > > corresponding node, which we obviously don't currently do. > > > > > > > > Ralph > > > > > > > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal < > > mikael.stal...@magine.com > > > > > > > > wrote: > > > > > > > > > > To me it does not seems good to force all Appender implementations > to > > > > > implement this. Especially not since the next logical step would > then > > > be > > > > to > > > > > do the same with other components such as Layouts. That would be a > > lot > > > of > > > > > work in total, and also add more work for all future components, > > > > including > > > > > 3rd party plugins. > > > > > > > > > > I think it makes more sense, and would be less work in total, if > the > > > > > configuration system would store and expose those attributes > without > > > > > involving the components themselves. > > > > > > > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory < > > garydgreg...@gmail.com> > > > > > wrote: > > > > > > > > > >> Apostolis, > > > > >> > > > > >> You are warmly welcome to contribute to Log4j. You can create a > JIRA > > > and > > > > >> attach a patch in unified diff file format. Unit tests as part of > > the > > > > patch > > > > >> are a must IMO. Feel free to flush out any design or > implementation > > > > here on > > > > >> the dev ML. > > > > >> > > > > >> Thank you! > > > > >> Gary > > > > >> > > > > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < > > > > ralph.go...@dslextreme.com> > > > > >> wrote: > > > > >> > > > > >>> All the attributes have to have String representations to be > usable > > > in > > > > >> the > > > > >>> XML, JSON & Properties configurations. Yes, the Map could contain > > > > Objects > > > > >>> but then every one of them has to be cast to its real object to > be > > > > >> usable. > > > > >>> > > > > >>> The map should be read-only because modifying its contents would > > have > > > > no > > > > >>> effect on the appender. > > > > >>> > > > > >>> The map should not be stored in an ivar but constructed whenever > > the > > > > >>> attributes are retrieved. Otherwise it will be temping to just > keep > > > > them > > > > >> in > > > > >>> a map and not as individual attributes, which would cause > problems. > > > > >>> > > > > >>> If you have nesting such as MyAppender extends MyBaseAppender > > > extends > > > > >>> AbstractOutputStreamAppender extends AbstractAppender, I > envision a > > > > >>> fillAttributes method at each “level” that fills in the > attributes > > it > > > > >> knows > > > > >>> about, so fillAttributeMap(map) should always call > > > > >>> super.fillAttributeMap(map) - except in AbstractAppender of > course > > - > > > > and > > > > >>> should call it before filling in its own attributes so that it > can > > > > >> override > > > > >>> any values provided by the base Appenders. If the primary > Appender > > > > does > > > > >>> not implement fillAttributeMap then only the attributes of its > > super > > > > >>> classes will be included, which is actually correct for the > > > > >> SyslogAppender. > > > > >>> > > > > >>> Ralph > > > > >>> > > > > On Jan 26, 2016, at 5:24 PM, Apostolis Giannakidis < > > > > >>> ap.giannaki...@gmail.com> wrote: > > > > > > > > One thing to note here. Correct me if I am wrong, but the map > > should > > > > be Map > > > Object> because not all attributes are Strings. From the top of > my > > > > >> head, > > > > >>> I > > > > know that an attribute could also be a boolean. > > > > > > > > On Wed, Jan 27, 2016 at 12:06 AM, Gary Gregory < > > > > garydgreg...@gmail.com > > > > >
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
I don't quite understand what you mean. On Wed, Jan 27, 2016 at 4:02 PM, Matt Sicker wrote: > That sounds a little fragile as some people either create or modify their > creation directly from the plugin factories. > > On 27 January 2016 at 07:05, Mikael Ståldal > wrote: > > > Then perhaps we should keep the node tree and expose it for this kind of > > queries, something like this: > > > > String hostname = loggerContext.getConfiguration(). > > getAttributesForAppender("syslogAppender").get("host"); > > > > This would require a new method in > > org.apache.logging.log4j.core.config.Configuration: > > > > public Map getAttributesForAppender(String appenderName); > > > > > > > > On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers > > > wrote: > > > > > While I understand your point, the node tree is discarded after the > > > plugins are created. We would have to keep it around for this to work. > > > Furthermore, each component would need to have a reference to its > > > corresponding node, which we obviously don't currently do. > > > > > > Ralph > > > > > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal < > mikael.stal...@magine.com > > > > > > wrote: > > > > > > > > To me it does not seems good to force all Appender implementations to > > > > implement this. Especially not since the next logical step would then > > be > > > to > > > > do the same with other components such as Layouts. That would be a > lot > > of > > > > work in total, and also add more work for all future components, > > > including > > > > 3rd party plugins. > > > > > > > > I think it makes more sense, and would be less work in total, if the > > > > configuration system would store and expose those attributes without > > > > involving the components themselves. > > > > > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory < > garydgreg...@gmail.com> > > > > wrote: > > > > > > > >> Apostolis, > > > >> > > > >> You are warmly welcome to contribute to Log4j. You can create a JIRA > > and > > > >> attach a patch in unified diff file format. Unit tests as part of > the > > > patch > > > >> are a must IMO. Feel free to flush out any design or implementation > > > here on > > > >> the dev ML. > > > >> > > > >> Thank you! > > > >> Gary > > > >> > > > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < > > > ralph.go...@dslextreme.com> > > > >> wrote: > > > >> > > > >>> All the attributes have to have String representations to be usable > > in > > > >> the > > > >>> XML, JSON & Properties configurations. Yes, the Map could contain > > > Objects > > > >>> but then every one of them has to be cast to its real object to be > > > >> usable. > > > >>> > > > >>> The map should be read-only because modifying its contents would > have > > > no > > > >>> effect on the appender. > > > >>> > > > >>> The map should not be stored in an ivar but constructed whenever > the > > > >>> attributes are retrieved. Otherwise it will be temping to just keep > > > them > > > >> in > > > >>> a map and not as individual attributes, which would cause problems. > > > >>> > > > >>> If you have nesting such as MyAppender extends MyBaseAppender > > extends > > > >>> AbstractOutputStreamAppender extends AbstractAppender, I envision a > > > >>> fillAttributes method at each “level” that fills in the attributes > it > > > >> knows > > > >>> about, so fillAttributeMap(map) should always call > > > >>> super.fillAttributeMap(map) - except in AbstractAppender of course > - > > > and > > > >>> should call it before filling in its own attributes so that it can > > > >> override > > > >>> any values provided by the base Appenders. If the primary Appender > > > does > > > >>> not implement fillAttributeMap then only the attributes of its > super > > > >>> classes will be included, which is actually correct for the > > > >> SyslogAppender. > > > >>> > > > >>> Ralph > > > >>> > > > On Jan 26, 2016, at 5:24 PM, Apostolis Giannakidis < > > > >>> ap.giannaki...@gmail.com> wrote: > > > > > > One thing to note here. Correct me if I am wrong, but the map > should > > > be Map > > Object> because not all attributes are Strings. From the top of my > > > >> head, > > > >>> I > > > know that an attribute could also be a boolean. > > > > > > On Wed, Jan 27, 2016 at 12:06 AM, Gary Gregory < > > > garydgreg...@gmail.com > > > >>> > > > wrote: > > > > > > > I could see AbstractAppender implementing a getAttributes() > method > > > >> like > > > > this: > > > > > > > > public Map getAttributes() { > > > > Map map = new HashMap<>(); > > > > fillAttributeMap(map); > > > > // (1) should the map be read-only? why? > > > > // (2) if the map is cached in an ivar, the it must be > > updated > > > >> by > > > > each appender when an attribute changes, so > > > > // I'd say no and follow the KISS principle for now. > > > > return map; > > > > } > > > > > > > > protected
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
That sounds a little fragile as some people either create or modify their creation directly from the plugin factories. On 27 January 2016 at 07:05, Mikael Ståldal wrote: > Then perhaps we should keep the node tree and expose it for this kind of > queries, something like this: > > String hostname = loggerContext.getConfiguration(). > getAttributesForAppender("syslogAppender").get("host"); > > This would require a new method in > org.apache.logging.log4j.core.config.Configuration: > > public Map getAttributesForAppender(String appenderName); > > > > On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers > wrote: > > > While I understand your point, the node tree is discarded after the > > plugins are created. We would have to keep it around for this to work. > > Furthermore, each component would need to have a reference to its > > corresponding node, which we obviously don't currently do. > > > > Ralph > > > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal > > > wrote: > > > > > > To me it does not seems good to force all Appender implementations to > > > implement this. Especially not since the next logical step would then > be > > to > > > do the same with other components such as Layouts. That would be a lot > of > > > work in total, and also add more work for all future components, > > including > > > 3rd party plugins. > > > > > > I think it makes more sense, and would be less work in total, if the > > > configuration system would store and expose those attributes without > > > involving the components themselves. > > > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory > > > wrote: > > > > > >> Apostolis, > > >> > > >> You are warmly welcome to contribute to Log4j. You can create a JIRA > and > > >> attach a patch in unified diff file format. Unit tests as part of the > > patch > > >> are a must IMO. Feel free to flush out any design or implementation > > here on > > >> the dev ML. > > >> > > >> Thank you! > > >> Gary > > >> > > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < > > ralph.go...@dslextreme.com> > > >> wrote: > > >> > > >>> All the attributes have to have String representations to be usable > in > > >> the > > >>> XML, JSON & Properties configurations. Yes, the Map could contain > > Objects > > >>> but then every one of them has to be cast to its real object to be > > >> usable. > > >>> > > >>> The map should be read-only because modifying its contents would have > > no > > >>> effect on the appender. > > >>> > > >>> The map should not be stored in an ivar but constructed whenever the > > >>> attributes are retrieved. Otherwise it will be temping to just keep > > them > > >> in > > >>> a map and not as individual attributes, which would cause problems. > > >>> > > >>> If you have nesting such as MyAppender extends MyBaseAppender > extends > > >>> AbstractOutputStreamAppender extends AbstractAppender, I envision a > > >>> fillAttributes method at each “level” that fills in the attributes it > > >> knows > > >>> about, so fillAttributeMap(map) should always call > > >>> super.fillAttributeMap(map) - except in AbstractAppender of course - > > and > > >>> should call it before filling in its own attributes so that it can > > >> override > > >>> any values provided by the base Appenders. If the primary Appender > > does > > >>> not implement fillAttributeMap then only the attributes of its super > > >>> classes will be included, which is actually correct for the > > >> SyslogAppender. > > >>> > > >>> Ralph > > >>> > > On Jan 26, 2016, at 5:24 PM, Apostolis Giannakidis < > > >>> ap.giannaki...@gmail.com> wrote: > > > > One thing to note here. Correct me if I am wrong, but the map should > > be Map > Object> because not all attributes are Strings. From the top of my > > >> head, > > >>> I > > know that an attribute could also be a boolean. > > > > On Wed, Jan 27, 2016 at 12:06 AM, Gary Gregory < > > garydgreg...@gmail.com > > >>> > > wrote: > > > > > I could see AbstractAppender implementing a getAttributes() method > > >> like > > > this: > > > > > > public Map getAttributes() { > > > Map map = new HashMap<>(); > > > fillAttributeMap(map); > > > // (1) should the map be read-only? why? > > > // (2) if the map is cached in an ivar, the it must be > updated > > >> by > > > each appender when an attribute changes, so > > > // I'd say no and follow the KISS principle for now. > > > return map; > > > } > > > > > > protected void fillAttributeMap(Map map) { > > > // ... > > > } > > > > > > The boilerplate of creating and/or managing the map can be in > > > getAttributes(). > > > Actually filling in the map in is done in fillAttributeMap() which > > >> each > > > appender can override. > > > > > > fillAttributeMap() could be abstract to force each appender to make > > >> sure > > > developers pay attention to p
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
Then perhaps we should keep the node tree and expose it for this kind of queries, something like this: String hostname = loggerContext.getConfiguration(). getAttributesForAppender("syslogAppender").get("host"); This would require a new method in org.apache.logging.log4j.core.config.Configuration: public Map getAttributesForAppender(String appenderName); On Wed, Jan 27, 2016 at 1:27 PM, Ralph Goers wrote: > While I understand your point, the node tree is discarded after the > plugins are created. We would have to keep it around for this to work. > Furthermore, each component would need to have a reference to its > corresponding node, which we obviously don't currently do. > > Ralph > > > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal > wrote: > > > > To me it does not seems good to force all Appender implementations to > > implement this. Especially not since the next logical step would then be > to > > do the same with other components such as Layouts. That would be a lot of > > work in total, and also add more work for all future components, > including > > 3rd party plugins. > > > > I think it makes more sense, and would be less work in total, if the > > configuration system would store and expose those attributes without > > involving the components themselves. > > > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory > > wrote: > > > >> Apostolis, > >> > >> You are warmly welcome to contribute to Log4j. You can create a JIRA and > >> attach a patch in unified diff file format. Unit tests as part of the > patch > >> are a must IMO. Feel free to flush out any design or implementation > here on > >> the dev ML. > >> > >> Thank you! > >> Gary > >> > >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers < > ralph.go...@dslextreme.com> > >> wrote: > >> > >>> All the attributes have to have String representations to be usable in > >> the > >>> XML, JSON & Properties configurations. Yes, the Map could contain > Objects > >>> but then every one of them has to be cast to its real object to be > >> usable. > >>> > >>> The map should be read-only because modifying its contents would have > no > >>> effect on the appender. > >>> > >>> The map should not be stored in an ivar but constructed whenever the > >>> attributes are retrieved. Otherwise it will be temping to just keep > them > >> in > >>> a map and not as individual attributes, which would cause problems. > >>> > >>> If you have nesting such as MyAppender extends MyBaseAppender extends > >>> AbstractOutputStreamAppender extends AbstractAppender, I envision a > >>> fillAttributes method at each “level” that fills in the attributes it > >> knows > >>> about, so fillAttributeMap(map) should always call > >>> super.fillAttributeMap(map) - except in AbstractAppender of course - > and > >>> should call it before filling in its own attributes so that it can > >> override > >>> any values provided by the base Appenders. If the primary Appender > does > >>> not implement fillAttributeMap then only the attributes of its super > >>> classes will be included, which is actually correct for the > >> SyslogAppender. > >>> > >>> Ralph > >>> > On Jan 26, 2016, at 5:24 PM, Apostolis Giannakidis < > >>> ap.giannaki...@gmail.com> wrote: > > One thing to note here. Correct me if I am wrong, but the map should > be Map Object> because not all attributes are Strings. From the top of my > >> head, > >>> I > know that an attribute could also be a boolean. > > On Wed, Jan 27, 2016 at 12:06 AM, Gary Gregory < > garydgreg...@gmail.com > >>> > wrote: > > > I could see AbstractAppender implementing a getAttributes() method > >> like > > this: > > > > public Map getAttributes() { > > Map map = new HashMap<>(); > > fillAttributeMap(map); > > // (1) should the map be read-only? why? > > // (2) if the map is cached in an ivar, the it must be updated > >> by > > each appender when an attribute changes, so > > // I'd say no and follow the KISS principle for now. > > return map; > > } > > > > protected void fillAttributeMap(Map map) { > > // ... > > } > > > > The boilerplate of creating and/or managing the map can be in > > getAttributes(). > > Actually filling in the map in is done in fillAttributeMap() which > >> each > > appender can override. > > > > fillAttributeMap() could be abstract to force each appender to make > >> sure > > developers pay attention to providing an implementation. > > > > Gary > > > > > > On Tue, Jan 26, 2016 at 3:46 PM, Apostolis Giannakidis < > > ap.giannaki...@gmail.com> wrote: > > > >> Well, since the idea is to add it to the Appender interface, it > means > > that > >> all concrete Appenders need to be modified as well. So, yes, I can > >> give > > it > >> a try to implement it for all the Appenders. One other simple way > >
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
While I understand your point, the node tree is discarded after the plugins are created. We would have to keep it around for this to work. Furthermore, each component would need to have a reference to its corresponding node, which we obviously don't currently do. Ralph > On Jan 27, 2016, at 2:33 AM, Mikael Ståldal wrote: > > To me it does not seems good to force all Appender implementations to > implement this. Especially not since the next logical step would then be to > do the same with other components such as Layouts. That would be a lot of > work in total, and also add more work for all future components, including > 3rd party plugins. > > I think it makes more sense, and would be less work in total, if the > configuration system would store and expose those attributes without > involving the components themselves. > > On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory > wrote: > >> Apostolis, >> >> You are warmly welcome to contribute to Log4j. You can create a JIRA and >> attach a patch in unified diff file format. Unit tests as part of the patch >> are a must IMO. Feel free to flush out any design or implementation here on >> the dev ML. >> >> Thank you! >> Gary >> >> On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers >> wrote: >> >>> All the attributes have to have String representations to be usable in >> the >>> XML, JSON & Properties configurations. Yes, the Map could contain Objects >>> but then every one of them has to be cast to its real object to be >> usable. >>> >>> The map should be read-only because modifying its contents would have no >>> effect on the appender. >>> >>> The map should not be stored in an ivar but constructed whenever the >>> attributes are retrieved. Otherwise it will be temping to just keep them >> in >>> a map and not as individual attributes, which would cause problems. >>> >>> If you have nesting such as MyAppender extends MyBaseAppender extends >>> AbstractOutputStreamAppender extends AbstractAppender, I envision a >>> fillAttributes method at each “level” that fills in the attributes it >> knows >>> about, so fillAttributeMap(map) should always call >>> super.fillAttributeMap(map) - except in AbstractAppender of course - and >>> should call it before filling in its own attributes so that it can >> override >>> any values provided by the base Appenders. If the primary Appender does >>> not implement fillAttributeMap then only the attributes of its super >>> classes will be included, which is actually correct for the >> SyslogAppender. >>> >>> Ralph >>> On Jan 26, 2016, at 5:24 PM, Apostolis Giannakidis < >>> ap.giannaki...@gmail.com> wrote: One thing to note here. Correct me if I am wrong, but the map should be Map>>> Object> because not all attributes are Strings. From the top of my >> head, >>> I know that an attribute could also be a boolean. On Wed, Jan 27, 2016 at 12:06 AM, Gary Gregory >> wrote: > I could see AbstractAppender implementing a getAttributes() method >> like > this: > > public Map getAttributes() { > Map map = new HashMap<>(); > fillAttributeMap(map); > // (1) should the map be read-only? why? > // (2) if the map is cached in an ivar, the it must be updated >> by > each appender when an attribute changes, so > // I'd say no and follow the KISS principle for now. > return map; > } > > protected void fillAttributeMap(Map map) { > // ... > } > > The boilerplate of creating and/or managing the map can be in > getAttributes(). > Actually filling in the map in is done in fillAttributeMap() which >> each > appender can override. > > fillAttributeMap() could be abstract to force each appender to make >> sure > developers pay attention to providing an implementation. > > Gary > > > On Tue, Jan 26, 2016 at 3:46 PM, Apostolis Giannakidis < > ap.giannaki...@gmail.com> wrote: > >> Well, since the idea is to add it to the Appender interface, it means > that >> all concrete Appenders need to be modified as well. So, yes, I can >> give > it >> a try to implement it for all the Appenders. One other simple way >> would > be >> to implement it once in the AbstractAppender so that all its >> subclasses >> will inherit it. >> >> On Tue, Jan 26, 2016 at 9:15 PM, Gary Gregory < >> garydgreg...@gmail.com> >> wrote: >> >>> On Tue, Jan 26, 2016 at 1:06 PM, Apostolis Giannakidis < >>> ap.giannaki...@gmail.com> wrote: >>> Actually, since this seems to be a useful feature, I would love to >> do >> the patch myself and contribute it to the project, if you don't mind. >>> >>> Do you plan on implementing this for all Appenders? >>> >>> Gary >>> On Tue, Jan 26, 2016 at 8:56 PM, Ralph Goers < >> ralph.
Re: Getter method for retrieving the attributes of an appender from the LoggerContext
To me it does not seems good to force all Appender implementations to implement this. Especially not since the next logical step would then be to do the same with other components such as Layouts. That would be a lot of work in total, and also add more work for all future components, including 3rd party plugins. I think it makes more sense, and would be less work in total, if the configuration system would store and expose those attributes without involving the components themselves. On Wed, Jan 27, 2016 at 7:14 AM, Gary Gregory wrote: > Apostolis, > > You are warmly welcome to contribute to Log4j. You can create a JIRA and > attach a patch in unified diff file format. Unit tests as part of the patch > are a must IMO. Feel free to flush out any design or implementation here on > the dev ML. > > Thank you! > Gary > > On Tue, Jan 26, 2016 at 5:02 PM, Ralph Goers > wrote: > > > All the attributes have to have String representations to be usable in > the > > XML, JSON & Properties configurations. Yes, the Map could contain Objects > > but then every one of them has to be cast to its real object to be > usable. > > > > The map should be read-only because modifying its contents would have no > > effect on the appender. > > > > The map should not be stored in an ivar but constructed whenever the > > attributes are retrieved. Otherwise it will be temping to just keep them > in > > a map and not as individual attributes, which would cause problems. > > > > If you have nesting such as MyAppender extends MyBaseAppender extends > > AbstractOutputStreamAppender extends AbstractAppender, I envision a > > fillAttributes method at each “level” that fills in the attributes it > knows > > about, so fillAttributeMap(map) should always call > > super.fillAttributeMap(map) - except in AbstractAppender of course - and > > should call it before filling in its own attributes so that it can > override > > any values provided by the base Appenders. If the primary Appender does > > not implement fillAttributeMap then only the attributes of its super > > classes will be included, which is actually correct for the > SyslogAppender. > > > > Ralph > > > > > On Jan 26, 2016, at 5:24 PM, Apostolis Giannakidis < > > ap.giannaki...@gmail.com> wrote: > > > > > > One thing to note here. Correct me if I am wrong, but the map should > > > be Map > > Object> because not all attributes are Strings. From the top of my > head, > > I > > > know that an attribute could also be a boolean. > > > > > > On Wed, Jan 27, 2016 at 12:06 AM, Gary Gregory > > > > wrote: > > > > > >> I could see AbstractAppender implementing a getAttributes() method > like > > >> this: > > >> > > >>public Map getAttributes() { > > >>Map map = new HashMap<>(); > > >>fillAttributeMap(map); > > >>// (1) should the map be read-only? why? > > >>// (2) if the map is cached in an ivar, the it must be updated > by > > >> each appender when an attribute changes, so > > >>// I'd say no and follow the KISS principle for now. > > >>return map; > > >>} > > >> > > >>protected void fillAttributeMap(Map map) { > > >>// ... > > >>} > > >> > > >> The boilerplate of creating and/or managing the map can be in > > >> getAttributes(). > > >> Actually filling in the map in is done in fillAttributeMap() which > each > > >> appender can override. > > >> > > >> fillAttributeMap() could be abstract to force each appender to make > sure > > >> developers pay attention to providing an implementation. > > >> > > >> Gary > > >> > > >> > > >> On Tue, Jan 26, 2016 at 3:46 PM, Apostolis Giannakidis < > > >> ap.giannaki...@gmail.com> wrote: > > >> > > >>> Well, since the idea is to add it to the Appender interface, it means > > >> that > > >>> all concrete Appenders need to be modified as well. So, yes, I can > give > > >> it > > >>> a try to implement it for all the Appenders. One other simple way > would > > >> be > > >>> to implement it once in the AbstractAppender so that all its > subclasses > > >>> will inherit it. > > >>> > > >>> On Tue, Jan 26, 2016 at 9:15 PM, Gary Gregory < > garydgreg...@gmail.com> > > >>> wrote: > > >>> > > On Tue, Jan 26, 2016 at 1:06 PM, Apostolis Giannakidis < > > ap.giannaki...@gmail.com> wrote: > > > > > Actually, since this seems to be a useful feature, I would love to > do > > >>> the > > > patch myself and contribute it to the project, if you don't mind. > > > > > > > Do you plan on implementing this for all Appenders? > > > > Gary > > > > > > > > On Tue, Jan 26, 2016 at 8:56 PM, Ralph Goers < > > >>> ralph.go...@dslextreme.com > > > > > > wrote: > > > > > >> Actually, I kind of like the idea of adding a getAttributes() > > >> method > > >>> to > > >> the Appender interface. Then each concrete Appender would do: > > >> > > >> public void getAttributes() { > > >>Map attributes = new HashMap<>(); > >