Re: [Wicket-user] replace ognl.
After rechecking the whole thread, I couldn't find out if this change got into the 1.1 release. Can somebdoy put some light onto this? thanks a lot, ./alex -- .w( the_mindstorm )p. #: Johan Compagner changed the world a bit at a time by saying on 10/27/2005 1:41 AM :# Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- SF.Net email is sponsored by: Tame your development challenges with Apache's Geronimo App Server. Download it for free - -and be entered to win a 42 plasma tv or your very own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
no this is 1.2On 11/10/05, Alexandru Popescu [EMAIL PROTECTED] wrote: After rechecking the whole thread, I couldn't find out if this change got into the 1.1 release. Cansomebdoy put some light onto this?thanks a lot,./alex--.w( the_mindstorm )p.#: Johan Compagner changed the world a bit at a time by saying on10/27/2005 1:41 AM :# Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the _expression_ is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan---SF.Net email is sponsored by:Tame your development challenges with Apache's Geronimo App Server. Downloadit for free - -and be entered to win a 42 plasma tv or your very own Sony(tm)PSP.Click here to play: http://sourceforge.net/geronimo.php___Wicket-user mailing list Wicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
#: Johan Compagner changed the world a bit at a time by saying on 11/10/2005 2:05 AM :# no this is 1.2 Thanks Johan. ./alex -- .w( the_mindstorm )p. On 11/10/05, Alexandru Popescu [EMAIL PROTECTED] wrote: After rechecking the whole thread, I couldn't find out if this change got into the 1.1 release. Can somebdoy put some light onto this? thanks a lot, ./alex -- .w( the_mindstorm )p. #: Johan Compagner changed the world a bit at a time by saying on 10/27/2005 1:41 AM :# Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- SF.Net email is sponsored by: Tame your development challenges with Apache's Geronimo App Server. Download it for free - -and be entered to win a 42 plasma tv or your very own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- SF.Net email is sponsored by: Tame your development challenges with Apache's Geronimo App Server. Download it for free - -and be entered to win a 42 plasma tv or your very own Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Conversion is first done in Wicket itself. So if you have a Field that gets its value from a map then you just have to type youre field first: TextField datePropertyTextField = new TextField(map.property, Date.class); I already thought about looking at the current key/index and then convert it to that one. But type conversion should be done at front. And this will be handled by wicket before the object is set. johan On 11/1/05, Laurent PETIT [EMAIL PROTECTED] wrote: Hello,some thoughts:property: This can can then be a bean property with get and set method. Or if a map is given as an object it will be lookup with the property as a key when there is not get method for that property. If the model is a Map, how is the property converted ?Could it be possible to check the type of the property by doing aget() first on the old value of the property, and this old value (ifnot null) could give the type of the property ? property1.property2: Both properties are lookup as written above. If property1 evaluates to null then if there is a setMethod (or if it is a map) and the Class of the property has a default constructor then the object will be constructed and set on the object.in the case of the map, how will you guess the correct type forproperty1 ? (without strongly typed setter ? !)Thanks in advance,--laurent ---SF.Net email is sponsored by:Tame your development challenges with Apache's Geronimo App Server. Downloadit for free - -and be entered to win a 42 plasma tv or your very own Sony(tm)PSP.Click here to play: http://sourceforge.net/geronimo.php___ Wicket-user mailing list Wicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I have tried CVS HEAD, but my old codes which based on ognl is broken since I use foo.bar.baz syntax to access public field. I hope that we are not restricted to only getter/setter: private String myProperty public String getMyProperty() public void setMyProperty() but also support : public String myProperty On 10/31/05, Eelco Hillenius [EMAIL PROTECTED] wrote: I agree with Matej.EelcoOn 10/30/05, Matej Knopp [EMAIL PROTECTED] wrote: In my opinion, there's no need for another indirection. From what I've seen, new users are already confused about models, making them even more complicated isn't imho the way to go. I think that Ognl models should be separated from the core and put to extensions. That would also get rid of ognl as a dependecy. Just my 2c. -Matej Johan Compagner wrote: Ok now we released 1.1 i have committed the ognl replacement It is not yet plugable yet we first need to see what we will or can support to bring back ognl for specific things. (plugable or in wicket extentions as its own model?)This is the current javadoc it explains what is possible now: This class parses expressions to lookup or set a value on the object that is given. The supported expressions are: property: This can can then be a bean property with get and set method. Or if a map is given as an object it will be lookup with the property as a key when there is not get method for that property. property1.property2: Both properties are lookup as written above. If property1 evaluates to null then if there is a setMethod (or if it is a map) and the Class of the property has a default constructor then the object will be constructed and set on the object. property.index: If the property is a List or Array then the second property can be a index on that list like: 'mylist.0' this _expression_ will also map on a getProperty(index) or setProperty(index,value) methods. If the object is a List then the list will grow automaticaly if the index is greater then the size Index or map properties can also be written as: property[index] or property[key] On 10/28/05, *Phil Kulak* [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: I like the idea of being able to plug in OGNL if you really need it. Now that I think about it, I've got a couple ChoiceRenderers that call functions using OGNL. Although, it probably would be very easy to rewrite these by just implementing the interface myself, and much more efficient.--- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list [EMAIL PROTECTED] mailto:[EMAIL PROTECTED].net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005Visit http://www.jboss.com/services/certification for more information___ Wicket-user mailing listWicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user -- Ingram ChenJava [EMAIL PROTECTED]Institue of BioMedical Sciences Academia Sinica Taiwanblog: http://www.javaworld.com.tw/roller/page/ingramchen
Re: [Wicket-user] replace ognl.
This is being fixed (as I read on the dev list). Martijn On 11/1/05, Ingram Chen [EMAIL PROTECTED] wrote: I have tried CVS HEAD, but my old codes which based on ognl is broken since I use foo.bar.baz syntax to access public field. I hope that we are not restricted to only getter/setter: private String myProperty public String getMyProperty() public void setMyProperty() but also support : public String myProperty On 10/31/05, Eelco Hillenius [EMAIL PROTECTED] wrote: I agree with Matej. Eelco On 10/30/05, Matej Knopp [EMAIL PROTECTED] wrote: In my opinion, there's no need for another indirection. From what I've seen, new users are already confused about models, making them even more complicated isn't imho the way to go. I think that Ognl models should be separated from the core and put to extensions. That would also get rid of ognl as a dependecy. Just my 2c. -Matej Johan Compagner wrote: Ok now we released 1.1 i have committed the ognl replacement It is not yet plugable yet we first need to see what we will or can support to bring back ognl for specific things. (plugable or in wicket extentions as its own model?) This is the current javadoc it explains what is possible now: This class parses expressions to lookup or set a value on the object that is given. The supported expressions are: property: This can can then be a bean property with get and set method. Or if a map is given as an object it will be lookup with the property as a key when there is not get method for that property. property1.property2: Both properties are lookup as written above. If property1 evaluates to null then if there is a setMethod (or if it is a map) and the Class of the property has a default constructor then the object will be constructed and set on the object. property.index: If the property is a List or Array then the second property can be a index on that list like: 'mylist.0' this expression will also map on a getProperty(index) or setProperty(index,value) methods. If the object is a List then the list will grow automaticaly if the index is greater then the size Index or map properties can also be written as: property[index] or property[key] On 10/28/05, *Phil Kulak* [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: I like the idea of being able to plug in OGNL if you really need it. Now that I think about it, I've got a couple ChoiceRenderers that call functions using OGNL. Although, it probably would be very easy to rewrite these by just implementing the interface myself, and much more efficient. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list [EMAIL PROTECTED] mailto:[EMAIL PROTECTED].net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user -- Ingram Chen Java [EMAIL PROTECTED] Institue of BioMedical Sciences Academia Sinica Taiwan blog: http://www.javaworld.com.tw/roller/page/ingramchen -- Living a wicket life... Martijn Dashorst - http://www.jroller.com/page/dashorst Wicket 1.1 is out: http://wicket.sourceforge.net/wicket-1.1 --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information
Re: [Wicket-user] replace ognl.
In my opinion, there's no need for another indirection. From what I've seen, new users are already confused about models, making them even more complicated isn't imho the way to go. I think that Ognl models should be separated from the core and put to extensions. That would also get rid of ognl as a dependecy. Just my 2c. -Matej Johan Compagner wrote: Ok now we released 1.1 i have committed the ognl replacement It is not yet plugable yet we first need to see what we will or can support to bring back ognl for specific things. (plugable or in wicket extentions as its own model?) This is the current javadoc it explains what is possible now: This class parses expressions to lookup or set a value on the object that is given. The supported expressions are: property: This can can then be a bean property with get and set method. Or if a map is given as an object it will be lookup with the property as a key when there is not get method for that property. property1.property2: Both properties are lookup as written above. If property1 evaluates to null then if there is a setMethod (or if it is a map) and the Class of the property has a default constructor then the object will be constructed and set on the object. property.index: If the property is a List or Array then the second property can be a index on that list like: 'mylist.0' this expression will also map on a getProperty(index) or setProperty(index,value) methods. If the object is a List then the list will grow automaticaly if the index is greater then the size Index or map properties can also be written as: property[index] or property[key] On 10/28/05, *Phil Kulak* [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: I like the idea of being able to plug in OGNL if you really need it. Now that I think about it, I've got a couple ChoiceRenderers that call functions using OGNL. Although, it probably would be very easy to rewrite these by just implementing the interface myself, and much more efficient. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list [EMAIL PROTECTED] mailto:[EMAIL PROTECTED].net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I agree with Matej. Eelco On 10/30/05, Matej Knopp [EMAIL PROTECTED] wrote: In my opinion, there's no need for another indirection. From what I've seen, new users are already confused about models, making them even more complicated isn't imho the way to go. I think that Ognl models should be separated from the core and put to extensions. That would also get rid of ognl as a dependecy. Just my 2c. -Matej Johan Compagner wrote: Ok now we released 1.1 i have committed the ognl replacement It is not yet plugable yet we first need to see what we will or can support to bring back ognl for specific things. (plugable or in wicket extentions as its own model?) This is the current javadoc it explains what is possible now: This class parses expressions to lookup or set a value on the object that is given. The supported expressions are: property: This can can then be a bean property with get and set method. Or if a map is given as an object it will be lookup with the property as a key when there is not get method for that property. property1.property2: Both properties are lookup as written above. If property1 evaluates to null then if there is a setMethod (or if it is a map) and the Class of the property has a default constructor then the object will be constructed and set on the object. property.index: If the property is a List or Array then the second property can be a index on that list like: 'mylist.0' this expression will also map on a getProperty(index) or setProperty(index,value) methods. If the object is a List then the list will grow automaticaly if the index is greater then the size Index or map properties can also be written as: property[index] or property[key] On 10/28/05, *Phil Kulak* [EMAIL PROTECTED] mailto:[EMAIL PROTECTED] wrote: I like the idea of being able to plug in OGNL if you really need it. Now that I think about it, I've got a couple ChoiceRenderers that call functions using OGNL. Although, it probably would be very easy to rewrite these by just implementing the interface myself, and much more efficient. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list [EMAIL PROTECTED] mailto:[EMAIL PROTECTED].net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
yes i think having a switch globally is not a good use case. I think you just have to choose it at model level. The only problem is that i have changed AbstractPropertyModel. And that one is the base class for X numbmer of classes (like compound) So the only real solution (besides always have at least 4 classes copied) is that the Ognl.getValue or Objects.getValue() is abstracted a way. And by default we use our own. and you can if you want give in the constructor of the AbstractPropertyModel which one you want to use. so: new CompoundModel(user, new OgnlResolver()); that would mean that all AbstractPropertyModel implementations must have an extra field (default null) And we need an interface that that can handle all types so we have to give them a method like: getValue(_expression_, component) or something and setValue(_expression_,value,component) Then we can move Ognl to extentions or to another project and only give that Resolver that can be used by any sub class of AbstractPropertyModel. johan On 10/27/05, Laurent PETIT [EMAIL PROTECTED] wrote: Hello,On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote: AbstractPropertyModel could work with something like a propertyresolver/ delegate (one more indirection :)) for getting and setting values based on a _expression_, and it might even be a good idea to have the option of having the default delegate as a application setting or something. Would be real easy to implement too. I like the idea of one more indirection :-) to allow more than asingle implementation.But I can see a problem arise if you do so: having manyimplementations that eventually will not share the same syntax for accessing properties.This could presumably be a risk for Components Reusability, if somepackaged components embark their own models and rely on a certainsyntax for specifying access to model properties ? ---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training CourseFree Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information___Wicket-user mailing list Wicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Does Wicket also check for direct public property of the same name ? not yet, but that is easy to build. Same remark concerning maps vs beans.If find it great if we don't know if the property is contained in the Map or is a property of the Model. i am alos still in favor the current impl by just doing dots nothing more. The only thing is that for maps that are also a bean is a bit difficult. But i find that a bad idea anyway.(a map that is also a bean) if the object is an instance of a Map: first try to interpret prop as a Map key that will never fails (if it is a setter) (mymap.xxx = yyy) So i think it must be the other way around. First try to do a setter and then if that fails use it as a map. johan
Re: [Wicket-user] replace ognl.
On 10/28/05, Johan Compagner [EMAIL PROTECTED] wrote: Does Wicket also check for direct public property of the same name ? not yet, but that is easy to build. That would be great. Same remark concerning maps vs beans. If find it great if we don't know if the property is contained in the Map or is a property of the Model. i am alos still in favor the current impl by just doing dots nothing more. The only thing is that for maps that are also a bean is a bit difficult. But i find that a bad idea anyway.(a map that is also a bean) Yes, I agree, and for the following reason: the use case I thought of was that, during refactoring, the model is first implemented using a Map, and when it is needed to have an object instead (in order to do some logic in the getter/setter, or for whatever reason), the Model is replaced by an object. And then only the creation/injection of the model has to change, not all the references to getter/setter. if the object is an instance of a Map: first try to interpret prop as a Map key that will never fails (if it is a setter) (mymap.xxx = yyy) So i think it must be the other way around. First try to do a setter and then if that fails use it as a map. Yes it's true, I only thought of the get() part of the problem, shame on me. But, as we are in the semantics of accessing an existing property, what about doing a get() before the set. And if the get() returns null, deduce that this is not a Map property we want to bind to, and search javabean properties, then public properties ? This does not seem to weird to me, since the absence of the key would be as bad as the absence of a getter in this use case ? --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
don't know if that is weird (that the key is not there) it can be completely valid. What happens if i do this: map.xxx = yyy First i try to get xxx key? Not there then i try to get xxx as a property still not there the back to the map? i still think that i can first check for the _expression_ as a property then if not found it is map value (doesn't matter if the key isn't there or not) johan On 10/28/05, Laurent PETIT [EMAIL PROTECTED] wrote: On 10/28/05, Johan Compagner [EMAIL PROTECTED] wrote: Does Wicket also check for direct public property of the same name ? not yet, but that is easy to build. That would be great. Same remark concerning maps vs beans. If find it great if we don't know if the property is contained in the Map or is a property of the Model. i am alos still in favor the current impl by just doing dots nothing more. The only thing is that for maps that are also a bean is a bit difficult. But i find that a bad idea anyway.(a map that is also a bean) Yes, I agree, and for the following reason: the use case I thought ofwas that, during refactoring, the model is first implemented using aMap, and when it is needed to have an object instead (in order to do some logic in the getter/setter, or for whatever reason), the Model isreplaced by an object.And then only the creation/injection of the model has to change, notall the references to getter/setter. if the object is an instance of a Map: first try to interpret prop as a Map key that will never fails (if it is a setter) (mymap.xxx = yyy) So i think it must be the other way around. First try to do a setter and then if that fails use it as a map.Yes it's true, I only thought of the get() part of the problem, shame on me.But, as we are in the semantics of accessing an existing property, what about doing a get() before the set.And if the get() returns null, deduce that this is not a Map propertywe want to bind to, and search javabean properties, then publicproperties ?This does not seem to weird to me, since the absence of the key would be as bad as the absence of a getter in this use case ?---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005Visit http://www.jboss.com/services/certification for more information___ Wicket-user mailing listWicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Well, yes, in fact, I think we can guess that it would be dangereous an error prone for the developer to have a key of the map with the same name as the property. So, indeed, it is not a problem if you first check for the property, then the map value. +1 for that, -- Laurent On 10/28/05, Johan Compagner [EMAIL PROTECTED] wrote: don't know if that is weird (that the key is not there) it can be completely valid. What happens if i do this: map.xxx = yyy First i try to get xxx key? Not there then i try to get xxx as a property still not there the back to the map? i still think that i can first check for the expression as a property then if not found it is map value (doesn't matter if the key isn't there or not) johan On 10/28/05, Laurent PETIT [EMAIL PROTECTED] wrote: On 10/28/05, Johan Compagner [EMAIL PROTECTED] wrote: Does Wicket also check for direct public property of the same name ? not yet, but that is easy to build. That would be great. Same remark concerning maps vs beans. If find it great if we don't know if the property is contained in the Map or is a property of the Model. i am alos still in favor the current impl by just doing dots nothing more. The only thing is that for maps that are also a bean is a bit difficult. But i find that a bad idea anyway.(a map that is also a bean) Yes, I agree, and for the following reason: the use case I thought of was that, during refactoring, the model is first implemented using a Map, and when it is needed to have an object instead (in order to do some logic in the getter/setter, or for whatever reason), the Model is replaced by an object. And then only the creation/injection of the model has to change, not all the references to getter/setter. if the object is an instance of a Map: first try to interpret prop as a Map key that will never fails (if it is a setter) (mymap.xxx = yyy) So i think it must be the other way around. First try to do a setter and then if that fails use it as a map. Yes it's true, I only thought of the get() part of the problem, shame on me. But, as we are in the semantics of accessing an existing property, what about doing a get() before the set. And if the get() returns null, deduce that this is not a Map property we want to bind to, and search javabean properties, then public properties ? This does not seem to weird to me, since the absence of the key would be as bad as the absence of a getter in this use case ? --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: That NPE is already handled. When you do a get a and somewhere in the line a null is encountered then null is returned Great!!! I still am thinking of a way to create the null object then this can be automatic if the object i have to create has a default constructor and a setMethod is there for that object on its parent. Wow, that's exactly what I need!!! So I don't have to create empty objects in my model anymore! For example the parent object could implement an interface called INullHandler with one method createObject(property) Then the parent object is responsable for making the missing field that it has. But don't know about this because such an interface is wicket specific (so ui specific) and the objects are or can be database things Those worlds should be seperate. The interface could be optional: if the parent implements this interface, Wicket uses it, otherwise it tries the default child object constructor. -- Davide Savazzi --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I like the idea of being able to plug in OGNL if you really need it. Now that I think about it, I've got a couple ChoiceRenderers that call functions using OGNL. Although, it probably would be very easy to rewrite these by just implementing the interface myself, and much more efficient. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
#: Martijn Dashorst changed the world a bit at a time by saying on 10/27/2005 7:28 AM :# +1 on moving OGNL into extensions. I think the OGNL thing is good, it provides a nice base implementation. I'd like to see the current implementation still being available for those 1% that need the full OGNL expression evaluator (though I think I share the same sentiment Phil voiced). This is exactly the way I would put my +1 :-). Do we also support things like person.getChildren().get(0).getName()? Martijn I guess this is: person.children.[0].name and Johan already mentioned it. ./alex -- .w( the_mindstorm )p. On 10/27/05, Phil Kulak [EMAIL PROTECTED] wrote: I'm very +1 on this. OGNL has turned into it's own scripting language over the years. If you want to use any of those features, then you're missing the point of Wicket. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I'm also +1 on moving OGNL into extension I guess this is: person.children.[0].name and Johan already mentioned it. I think Johan mentioned, it'll be person.children.0.name Juergen On 10/27/05, Alexandru Popescu [EMAIL PROTECTED] wrote: #: Martijn Dashorst changed the world a bit at a time by saying on 10/27/2005 7:28 AM :# +1 on moving OGNL into extensions. I think the OGNL thing is good, it provides a nice base implementation. I'd like to see the current implementation still being available for those 1% that need the full OGNL expression evaluator (though I think I share the same sentiment Phil voiced). This is exactly the way I would put my +1 :-). Do we also support things like person.getChildren().get(0).getName()? Martijn I guess this is: person.children.[0].name and Johan already mentioned it. ./alex -- .w( the_mindstorm )p. On 10/27/05, Phil Kulak [EMAIL PROTECTED] wrote: I'm very +1 on this. OGNL has turned into it's own scripting language over the years. If you want to use any of those features, then you're missing the point of Wicket. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
RE: [Wicket-user] replace ognl.
Also +1 Actually our project is what triggerd this change and will bring some new features we need. Nothing fancy though. What I have gathered from johan it will be possible to get elements from maps / lists / arrays etc. The expression would be something like person.children.0.name It is very similar if not the same syntax ognl uses. If children was a map containing a key jack you could use person.children.jack.name Maurice Marrink www.topicus.nl -Original Message- From: Martijn Dashorst [mailto:[EMAIL PROTECTED] Sent: donderdag 27 oktober 2005 7:28 To: wicket-user@lists.sourceforge.net Subject: Re: [Wicket-user] replace ognl. +1 on moving OGNL into extensions. I think the OGNL thing is good, it provides a nice base implementation. I'd like to see the current implementation still being available for those 1% that need the full OGNL expression evaluator (though I think I share the same sentiment Phil voiced). Do we also support things like person.getChildren().get(0).getName()? Martijn On 10/27/05, Phil Kulak [EMAIL PROTECTED] wrote: I'm very +1 on this. OGNL has turned into it's own scripting language over the years. If you want to use any of those features, then you're missing the point of Wicket. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
yes i currently don't have support for { or [ or things like that.
Just plain text with dots. This makes it very easy and fast to parse.
but if people really want this:
person.children[0].name
does ognl has a special map _expression_? something like person.childeren{child1}.name ?
i know that it also works for just strings like Ognl.getValue(child1 , myHashMap);
so it all depends on what we really would like to see for array/list:
person.children.0.name
person.children.[0].name
person.children[0].name
and maps?
if we want [] in it then i just have to use a bit more logic like a tokenizer instead of just doing a plain indexOf loop
I like the simple approache i have now. .
On 10/27/05, Juergen Donnerstag [EMAIL PROTECTED] wrote:
I'm also +1 on moving OGNL into extension I guess this is: person.children.[0].name and Johan already mentioned it.I think Johan mentioned, it'll be person.children.0.name
JuergenOn 10/27/05, Alexandru Popescu [EMAIL PROTECTED] wrote: #: Martijn Dashorst changed the world a bit at a time by saying on10/27/2005 7:28 AM :#
+1 on moving OGNL into extensions. I think the OGNL thing is good, it provides a nice base implementation. I'd like to see the current implementation still being available for those 1% that need the full OGNL _expression_ evaluator
(though I think I share the same sentiment Phil voiced). This is exactly the way I would put my +1 :-). Do we also support things like
person.getChildren().get(0).getName()? Martijn I guess this is: person.children.[0].name and Johan already mentioned it. ./alex --
.w( the_mindstorm )p. On 10/27/05, Phil Kulak [EMAIL PROTECTED] wrote: I'm very +1 on this. OGNL has turned into it's own scripting language
over the years. If you want to use any of those features, then you're missing the point of Wicket.---
This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list
Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
--- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information
___ Wicket-user mailing list Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification
for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005Visit http://www.jboss.com/services/certification for more information___
Wicket-user mailing listWicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Yes, I agree. Keep it simple. Doesn't Groovy do it the same way, you
proposed it, Johan?
Juergen
On 10/27/05, Alexandru Popescu [EMAIL PROTECTED] wrote:
#: Johan Compagner changed the world a bit at a time by saying on 10/27/2005
10:06 AM :#
yes i currently don't have support for { or [ or things like that.
Just plain text with dots. This makes it very easy and fast to parse.
but if people really want this:
person.children[0].name
does ognl has a special map expression? something like
person.childeren{child1}.name ?
i know that it also works for just strings like Ognl.getValue(child1 ,
myHashMap);
so it all depends on what we really would like to see for array/list:
person.children.0.name http://person.children.0.name
person.children.[0].name
person.children[0].name
and maps?
if we want [] in it then i just have to use a bit more logic like a
tokenizer instead of just doing a plain indexOf loop
I like the simple approache i have now. .
If you start this to keep it simple, than we should keep it that way. IMO
person.children.0.name is
same good as person.children[0].name and as person.children.[0].name, once it
is clearly described
somewhere ;-).
./alex
--
.w( the_mindstorm )p.
On 10/27/05, Juergen Donnerstag [EMAIL PROTECTED] wrote:
I'm also +1 on moving OGNL into extension
I guess this is: person.children.[0].name and Johan already mentioned
it.
I think Johan mentioned, it'll be
person.children.0.namehttp://person.children.0.name
Juergen
On 10/27/05, Alexandru Popescu [EMAIL PROTECTED]
wrote:
#: Martijn Dashorst changed the world a bit at a time by saying on
10/27/2005 7:28 AM :#
+1 on moving OGNL into extensions.
I think the OGNL thing is good, it provides a nice base
implementation. I'd like to see the current implementation still being
available for those 1% that need the full OGNL expression evaluator
(though I think I share the same sentiment Phil voiced).
This is exactly the way I would put my +1 :-).
Do we also support things like
person.getChildren().get(0).getName()?
Martijn
I guess this is: person.children.[0].name and Johan already mentioned
it.
./alex
--
.w( the_mindstorm )p.
On 10/27/05, Phil Kulak [EMAIL PROTECTED] wrote:
I'm very +1 on this. OGNL has turned into it's own scripting language
over the years. If you want to use any of those features, then you're
missing the point of Wicket.
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Re: [Wicket-user] replace ognl.
I strongly support replacing OGNL with something more performant, as long as it is not yet another new syntax for accessing properties (we already have OGNL, Spring property access syntax, some expression languages from the JSP/Taglib world and probably some others). At least it sould be OGNL compatible for 90% of the cases. Perhaps using the Spring implementation is possible? It is widely tested and I suppose it doesn't have the OGNL performance issues. Further I remember an issue in this mailinglist that in OGNL adressing a not existing subtree results in a NullPointerException (customer.address.city.zip throws NPE if address is null). Maybe a new implementation could handle this issue without NPE? (There was the suggestion for a syntax like customer.address?.city.zip, but I'm not sure it this is a good idea). Cheers, Felix On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
That NPE is already handled. When you do a get a and somewhere in the line a null is encountered then null is returned With a set a NPE is thrown currently because if that is not handled you loose data (which is bad) I still am thinking of a way to create the null object then this can be automatic if the object i have to create has a default constructor and a setMethod is there for that object on its parent. So that are 2 big ifs... Also thinking about a NullHandler but where does that come from how do we specify that? For example the parent object could implement an interface called INullHandler with one method createObject(property) Then the parent object is responsable for making the missing field that it has. But don't know about this because such an interface is wicket specific (so ui specific) and the objects are or can be database things Those worlds should be seperate. johan On 10/27/05, Felix von Delius [EMAIL PROTECTED] wrote: I strongly support replacing OGNL with something more performant, aslong as it is not yet another new syntax for accessing properties (wealready have OGNL, Spring property access syntax, some _expression_languages from the JSP/Taglib world and probably some others). At least it sould be OGNL compatible for 90% of the cases. Perhaps usingthe Spring implementation is possible? It is widely tested and Isuppose it doesn't have the OGNL performance issues.Further I remember an issue in this mailinglist that in OGNL adressing a not existing subtree results in a NullPointerException(customer.address.city.zip throws NPE if address is null). Maybe anew implementation could handle this issue without NPE? (There was thesuggestion for a syntax like customer.address?.city.zip, but I'm notsure it this is a good idea).Cheers,FelixOn 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: HiI have written a replacement of OGNL when i test it with a very simple test (The FormInput example)then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 timesand only 1/5 of the memory garbage is generated (50MB against 10MB)The question is what do you guys use of ognl? Can i completely drop it or must i make it an option sothat you can switch in youre application for all using ognl or the homebrew wicket impl.Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all thesub classes must also be copied...(Like CompoundXX)What i do support now is this:person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element)person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the _expression_ is then the key or the index you can also put values in a mapor append/set to a list:person.addresses.homeaddress = new Address()person.addresses.10 = new Address()if the list size is smaller then 10 then it will appends null to make it that size.addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrownif you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable.So can people live with this? Does anybody uses something different of ognl?johan--- This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training CourseFree Certification Exam for All Training Attendees Through End of 2005Visit http://www.jboss.com/services/certification for more information___Wicket-user mailing listWicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
#: Maurice Marrink changed the world a bit at a time by saying on 10/27/2005 11:16 AM :# That nullpointer is our problem too and one of the reasons it is being changed right now. Don't think johan has put in the syntax you just suggested though. Afaik it just returns null. IMO evaluating the expression should stop immediately it reaches a null, but without throwing an exception (probably just some smart logging). Take a look at FreeMarker for example and its annoying default behavior regarding null handling. Another solution would be to provide a way to specify a NullPropertyHandler. ./alex -- .w( the_mindstorm )p. Maurice Marrink www.topicus.nl -Original Message- From: Felix von Delius [mailto:[EMAIL PROTECTED] Sent: donderdag 27 oktober 2005 11:10 To: wicket-user@lists.sourceforge.net Subject: Re: [Wicket-user] replace ognl. I strongly support replacing OGNL with something more performant, as long as it is not yet another new syntax for accessing properties (we already have OGNL, Spring property access syntax, some expression languages from the JSP/Taglib world and probably some others). At least it sould be OGNL compatible for 90% of the cases. Perhaps using the Spring implementation is possible? It is widely tested and I suppose it doesn't have the OGNL performance issues. Further I remember an issue in this mailinglist that in OGNL adressing a not existing subtree results in a NullPointerException (customer.address.city.zip throws NPE if address is null). Maybe a new implementation could handle this issue without NPE? (There was the suggestion for a syntax like customer.address?.city.zip, but I'm not sure it this is a good idea). Cheers, Felix On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
yes it directly returns null. i my eyes a nullpropertyhanlder only applies to when you want to set something not when you want to get something.On 10/27/05, Alexandru Popescu [EMAIL PROTECTED] wrote: #: Maurice Marrink changed the world a bit at a time by saying on10/27/2005 11:16 AM :# That nullpointer is our problem too and one of the reasons it is being changed right now. Don't think johan has put in the syntax you just suggested though. Afaik it just returns null.IMO evaluating the _expression_ should stop immediately it reaches a null, but without throwing anexception (probably just some smart logging). Take a look at FreeMarker for example and its annoying default behavior regarding null handling.Another solution would be to provide a way to specify a NullPropertyHandler../alex--.w( the_mindstorm )p. Maurice Marrink www.topicus.nl -Original Message- From: Felix von Delius [mailto:[EMAIL PROTECTED] ] Sent: donderdag 27 oktober 2005 11:10 To: wicket-user@lists.sourceforge.net Subject: Re: [Wicket-user] replace ognl. I strongly support replacing OGNL with something more performant, as long as it is not yet another new syntax for accessing properties (we already have OGNL, Spring property access syntax, some _expression_ languages from the JSP/Taglib world and probably some others). At least it sould be OGNL compatible for 90% of the cases. Perhaps using the Spring implementation is possible? It is widely tested and I suppose it doesn't have the OGNL performance issues. Further I remember an issue in this mailinglist that in OGNL adressing a not existing subtree results in a NullPointerException (customer.address.city.zip throws NPE if address is null). Maybe a new implementation could handle this issue without NPE? (There was the suggestion for a syntax like customer.address?.city.zip, but I'm not sure it this is a good idea). Cheers, Felix On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: HiI have written a replacement of OGNL when i test it with a very simple test (The FormInput example)then i see quite some improvements in cpu speedups and mem improvementsAround 40% speed increase for a submitting the forminput example page 20 timesand only 1/5 of the memory garbage is generated (50MB against 10MB)The question is what do you guys use of ognl? Can i completely drop it or must i make it an option sothat you can switch in youre application for all using ognl or the homebrew wicket impl.Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all thesub classes must also be copied...(Like CompoundXX)What i do support now is this: person.name (plain properties)person.addresses.0.street (addresses is a list and i take the first element)person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it)so maps and list are seen and the next part of the _expression_ is then the key or the index you can also put values in a map or append/set to a list:person.addresses.homeaddress = new Address()person.addresses.10 = new Address()if the list size is smaller then 10 then it will appends null to make it that size.addresses can also be an Array but then it won't be able to grow.Ofcourse the person.address.street will just be null if address is null, no exception will be thrownif you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable.So can people live with this? Does anybody uses something different of ognl?johan --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user--- This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training CourseFree Certification Exam for All Training Attendees Through End of 2005Visit http://www.jboss.com/services/certification for more information___Wicket-user mailing listWicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists
Re: [Wicket-user] replace ognl.
I'm also +1 for replacing ognl. I'm not trying to say that Wicket is slow, not at all, but any speed improvement is .an improvement :)! - Johannes Johan Compagner wrote: Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Johan Compagner wrote:
yes i currently don't have support for { or [ or things like that.
Just plain text with dots. This makes it very easy and fast to parse.
but if people really want this:
person.children[0].name
I can think of two reasons why an access operator [] could be useful:
- It might make the expression more readable, because you know that you
are operating on an array, List or Map by looking at the expression.
- The content of the access operator could be an expression itself. Example:
person.children[person.selectedChildIndex].name
But I think this is already too much functionality. I would prefer the
most simple approach with only the most basic support for arrays, Lists
and Maps using only dot-syntax.
There is one issue that's very important to me. I have run into this
using the Expression Language of JSP 2.0:
If you have a class that implements Map, you can not access a bean
property on the instance.
Let's say we have the following class:
class Bar extends HashMap {
public String getFoo() ...
}
Now create an instance and populate it:
Bar b = new Bar();
b.put(yadda, value);
The JSP EL Evaluator would return null for the following expression:
b.foo
I think the Evaluator should look for a value in the map using
containsKey() and look for a bean property if nothing was found. On a
list, you could check if the key is numeric.
This would not be a performance hit, and you are still able to use the
instance like any other object.
Timo
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I can think of two reasons why an access operator [] could be useful:- It might make the _expression_ more readable, because you know that you
are operating on an array, List or Map by looking at the _expression_.- The content of the access operator could be an _expression_ itself. Example:person.children[person.selectedChildIndex].name
If i you want to this then use ognl.
I will not support this by the basic stuff.
If you have a class that implements Map, you can not access a beanproperty on the instance.
Let's say we have the following class:class Bar extends HashMap {public String getFoo() ...}Now create an instance and populate it:Bar b = new Bar();
b.put(yadda, value);The JSP EL Evaluator would return null for the following _expression_:b.fooI think the Evaluator should look for a value in the map using
containsKey() and look for a bean property if nothing was found. On alist, you could check if the key is numeric.
I though about this (List and Maps being beans)
For a list this is not that hard because if the next expresion can be converted to an int then it is just an int for the index.
(get1() is then not supported)
And if it can't be converted then we could check if a method is there.
For maps this is different.
Currently i see List/Maps are not beans. If i always use them as a List or Map.
First check if something is there (in the map) looks strange to me.
Because what happens if you do this:
mymap.value = x
and there is a setValue() but not a map entry?
Should i use setValue or do a map update. (or both?)
I still like the current simple way of notation that i have now. And i
think that we will say: do you want more use the ognl models.
johan
Re: [Wicket-user] replace ognl.
Johan Compagner wrote: I can think of two reasons why an access operator [] could be useful: - It might make the expression more readable, because you know that you are operating on an array, List or Map by looking at the expression. - The content of the access operator could be an expression itself. Example: person.children[person.selectedChildIndex].name If i you want to this then use ognl. I will not support this by the basic stuff. No, I don't want to do this. My point is: Brackets can be useful for the two reasons I mentioned. If you do not intend to support these features, there is no reason to use brackets in the syntax. I though about this (List and Maps being beans) For a list this is not that hard because if the next expresion can be converted to an int then it is just an int for the index. (get1() is then not supported) And if it can't be converted then we could check if a method is there. For maps this is different. Currently i see List/Maps are not beans. If i always use them as a List or Map. First check if something is there (in the map) looks strange to me. It gives better performance. Because what happens if you do this: mymap.value = x and there is a setValue() but not a map entry? Should i use setValue or do a map update. (or both?) That's true, I didn't think of this situation. So it is necessary to check for getters/setters before doing map stuff. This would be the following accessor precedence: 1 getter/setter 2 List 3 Map I don't know if the performance hit is too high to allow getter/setter access on Maps and Lists, but I would certainly like this feature because it is an requirement of a framework I am using. Anyways, kudos for replacing OGNL with something more simple and efficient. Timo --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I'm not too sure about using dots for index/ map references. It differs from OGNL: http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html. And personally, I think it is usefull to be able to distinct single from indexed properties just by looking at the expression. Is it /that/ more expensive? I mean, optimizing is good, but we don't have to over-do it. Eelco --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
ognl can do map lookup with just a _expression_ (that was the first thing i had to do because that did go wrong when i started testing it) this was used: Ognl.getValue(string, anHashmap); so the need to do [sd] even with juk why is that needed ...? On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote: I'm not too sure about using dots for index/ map references. Itdiffers from OGNL:http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html .And personally, I think it is usefull to be able to distinct singlefrom indexed properties just by looking at the _expression_. Is it/that/ more expensive? I mean, optimizing is good, but we don't haveto over-do it. Eelco---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training CourseFree Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information___Wicket-user mailing list Wicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
If it equals at least one of the OGNL uses, then I'm fine with it. Eelco On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: ognl can do map lookup with just a expression (that was the first thing i had to do because that did go wrong when i started testing it) this was used: Ognl.getValue(string, anHashmap); so the need to do [sd] even with juk why is that needed ...? On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote: I'm not too sure about using dots for index/ map references. It differs from OGNL: http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html . And personally, I think it is usefull to be able to distinct single from indexed properties just by looking at the expression. Is it /that/ more expensive? I mean, optimizing is good, but we don't have to over-do it. Eelco --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
The only thing for me to use [] for an index and { } for a map (for example)
would be that this:
mymap.property would mean that you really want to have a getProperty on a map
(same goes for list)
so i will change it to use with [] (without quotes that i find totally not needed)
foo.list[0].bar
foo.map[key].bar
foo.list.bar (this is a get property of the list)
foo.map.bar (this is a get property of the map)
But i am affraid that when i do the last thing (it needs to be a property of map) that wicket fails currently
so i think i fall back to that that i will also look it up as a key if the property is not found.
This will result in more mem en cpu usage.
(i don't create any garbage currently)On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote:
ognl can do map lookup with just a _expression_
(that was the first thing i had to do because that did go wrong when i started testing it)
this was used: Ognl.getValue(string, anHashmap);
so the need to do [sd] even with juk why is that needed ...?
On 10/27/05, Eelco Hillenius [EMAIL PROTECTED]
wrote:
I'm not too sure about using dots for index/ map references. Itdiffers from OGNL:
http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html
.And personally, I think it is usefull to be able to distinct singlefrom indexed properties just by looking at the _expression_. Is it/that/ more expensive? I mean, optimizing is good, but we don't have
to over-do it.
Eelco---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training CourseFree Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information___
Wicket-user mailing list
Wicket-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
I'd prefer if it used a groovy-like syntax (or actual groovy behind the
curtain) as much as possible.
[EMAIL PROTECTED] 2005-10-27 11:15:05 AM
The only thing for me to use [] for an index and { } for a map (for
example)
would be that this:
mymap.property would mean that you really want to have a getProperty on
a
map
(same goes for list)
so i will change it to use with [] (without quotes that i find totally
not
needed)
foo.list[0].bar
foo.map[key].bar
foo.list.bar (this is a get property of the list)
foo.map.bar (this is a get property of the map)
But i am affraid that when i do the last thing (it needs to be a
property of
map) that wicket fails currently
so i think i fall back to that that i will also look it up as a key if
the
property is not found.
This will result in more mem en cpu usage.
(i don't create any garbage currently)
On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote:
ognl can do map lookup with just a expression
(that was the first thing i had to do because that did go wrong when
i
started testing it)
this was used: Ognl.getValue(string, anHashmap);
so the need to do [sd] even with juk why is that needed ...?
On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote:
I'm not too sure about using dots for index/ map references. It
differs from OGNL:
http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html
.
And personally, I think it is usefull to be able to distinct
single
from indexed properties just by looking at the expression. Is it
/that/ more expensive? I mean, optimizing is good, but we don't
have
to over-do it.
Eelco
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
__
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.
Katun Corporation -- www.katun.com
_
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Johan Compagner wrote:
The only thing for me to use [] for an index and { } for a map (for example)
would be that this:
mymap.property would mean that you really want to have a getProperty on
a map
(same goes for list)
so i will change it to use with [] (without quotes that i find totally
not needed)
foo.list[0].bar
foo.map[key].bar
foo.list.bar (this is a get property of the list)
foo.map.bar (this is a get property of the map)
Great. This looks very nice.
Timo
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
and that is?
Very similar to what you are proposing:
http://groovy.codehaus.org/Collections
and
bean?.property (allows bean to be null)
(see http://groovy.codehaus.org/Migration+From+Classic+to+JSR+syntax )
I really appreciate how OGNL is very flexible with allowing me to do
things like concatenate two properties into one without having to write
a new getter. Having a groovy-based property model would give similar
advantages. And Groovy allows for compact iterating over collections
like OGNL does (although I haven't had a need for it yet). If you got
rid of that kind of functionality altogether, I think Wicket would be
poorer for it.
My 2 cents worth,
- Jonathan
[EMAIL PROTECTED] 2005-10-27 1:23:03 PM
and that is?
On 10/27/05, Jonathan Carlson [EMAIL PROTECTED] wrote:
I'd prefer if it used a groovy-like syntax (or actual groovy behind
the
curtain) as much as possible.
[EMAIL PROTECTED] 2005-10-27 11:15:05 AM
The only thing for me to use [] for an index and { } for a map (for
example)
would be that this:
mymap.property would mean that you really want to have a getProperty
on
a
map
(same goes for list)
so i will change it to use with [] (without quotes that i find
totally
not
needed)
foo.list[0].bar
foo.map[key].bar
foo.list.bar (this is a get property of the list)
foo.map.bar (this is a get property of the map)
But i am affraid that when i do the last thing (it needs to be a
property of
map) that wicket fails currently
so i think i fall back to that that i will also look it up as a key
if
the
property is not found.
This will result in more mem en cpu usage.
(i don't create any garbage currently)
On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote:
ognl can do map lookup with just a expression
(that was the first thing i had to do because that did go wrong
when
i
started testing it)
this was used: Ognl.getValue(string, anHashmap);
so the need to do [sd] even with juk why is that needed ...?
On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote:
I'm not too sure about using dots for index/ map references. It
differs from OGNL:
http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html
.
And personally, I think it is usefull to be able to distinct
single
from indexed properties just by looking at the expression. Is it
/that/ more expensive? I mean, optimizing is good, but we don't
have
to over-do it.
Eelco
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End
of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
__
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please
notify
the system manager.
Katun Corporation -- www.katun.com http://www.katun.com
_
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
__
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.
Katun Corporation -- www.katun.com
_
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of 2005
Visit http://www.jboss.com/services/certification for more information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
RE: [Wicket-user] replace ognl.
Why not keep wicket simple? I really like the simplicity of the dot notation
johan created. I think we should use [] only as a hint. By default wicket
tries to find a getter and then checks if the object is a list/set/map and
contains the key. The [] hint would tell wicket to skip the getter and go
straight for the contains check. I also like that any null link in the chain
results in a null.
If you really want the complete groovy syntax why not build a groovy
property model using groovy expression evaluator? Same goes for ognl.
My 2 cents.
-Igor
-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of
Jonathan Carlson
Sent: Thursday, October 27, 2005 11:40 AM
To: wicket-user@lists.sourceforge.net
Subject: Re: [Wicket-user] replace ognl.
and that is?
Very similar to what you are proposing:
http://groovy.codehaus.org/Collections
and
bean?.property (allows bean to be null)
(see http://groovy.codehaus.org/Migration+From+Classic+to+JSR+syntax )
I really appreciate how OGNL is very flexible with allowing
me to do things like concatenate two properties into one
without having to write a new getter. Having a groovy-based
property model would give similar advantages. And Groovy
allows for compact iterating over collections like OGNL does
(although I haven't had a need for it yet). If you got rid
of that kind of functionality altogether, I think Wicket
would be poorer for it.
My 2 cents worth,
- Jonathan
[EMAIL PROTECTED] 2005-10-27 1:23:03 PM
and that is?
On 10/27/05, Jonathan Carlson [EMAIL PROTECTED] wrote:
I'd prefer if it used a groovy-like syntax (or actual groovy behind
the
curtain) as much as possible.
[EMAIL PROTECTED] 2005-10-27 11:15:05 AM
The only thing for me to use [] for an index and { } for a map (for
example)
would be that this:
mymap.property would mean that you really want to have a getProperty
on
a
map
(same goes for list)
so i will change it to use with [] (without quotes that i find
totally
not
needed)
foo.list[0].bar
foo.map[key].bar
foo.list.bar (this is a get property of the list)
foo.map.bar (this is
a get property of the map)
But i am affraid that when i do the last thing (it needs to be a
property of
map) that wicket fails currently
so i think i fall back to that that i will also look it up as a key
if
the
property is not found.
This will result in more mem en cpu usage.
(i don't create any garbage currently)
On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote:
ognl can do map lookup with just a expression (that was the first
thing i had to do because that did go wrong
when
i
started testing it)
this was used: Ognl.getValue(string, anHashmap);
so the need to do [sd] even with juk why is that needed ...?
On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote:
I'm not too sure about using dots for index/ map references. It
differs from OGNL:
http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/ind
exing.html
.
And personally, I think it is usefull to be able to distinct
single
from indexed properties just by looking at the
expression. Is it
/that/ more expensive? I mean, optimizing is good, but we don't
have
to over-do it.
Eelco
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course Free
Certification Exam for All Training Attendees Through End
of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
__
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to
whom they
are addressed. If you have received this email in error please
notify
the system manager.
Katun Corporation -- www.katun.com http://www.katun.com
_
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course Free
Certification Exam for All Training Attendees Through End of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
__
This email and any files transmitted
Re: [Wicket-user] replace ognl.
It's all about choice of course. I agree with most proponents of
getting rid of OGNL: it's way too heavy for 95% (or more) of the usual
cases, and less dependencies == good. But, I think it would be really
good not only to keep alternatives in e.g. the extensions project, or
with a seperate project (for the sake of keeping the number of
dependencies needed for extensions small too), but also make sure our
compound property models can work with the different options there are
(you could write a Groovy implementation if you want to, and put it in
the groovy wicket-stuff project).
AbstractPropertyModel could work with something like a
propertyresolver/ delegate (one more indirection :)) for getting and
setting values based on a expression, and it might even be a good idea
to have the option of having the default delegate as a application
setting or something. Would be real easy to implement too.
Eelco
On 10/27/05, Jonathan Carlson [EMAIL PROTECTED] wrote:
and that is?
Very similar to what you are proposing:
http://groovy.codehaus.org/Collections
and
bean?.property (allows bean to be null)
(see http://groovy.codehaus.org/Migration+From+Classic+to+JSR+syntax )
I really appreciate how OGNL is very flexible with allowing me to do
things like concatenate two properties into one without having to write
a new getter. Having a groovy-based property model would give similar
advantages. And Groovy allows for compact iterating over collections
like OGNL does (although I haven't had a need for it yet). If you got
rid of that kind of functionality altogether, I think Wicket would be
poorer for it.
My 2 cents worth,
- Jonathan
[EMAIL PROTECTED] 2005-10-27 1:23:03 PM
and that is?
On 10/27/05, Jonathan Carlson [EMAIL PROTECTED] wrote:
I'd prefer if it used a groovy-like syntax (or actual groovy behind
the
curtain) as much as possible.
[EMAIL PROTECTED] 2005-10-27 11:15:05 AM
The only thing for me to use [] for an index and { } for a map (for
example)
would be that this:
mymap.property would mean that you really want to have a getProperty
on
a
map
(same goes for list)
so i will change it to use with [] (without quotes that i find
totally
not
needed)
foo.list[0].bar
foo.map[key].bar
foo.list.bar (this is a get property of the list)
foo.map.bar (this is a get property of the map)
But i am affraid that when i do the last thing (it needs to be a
property of
map) that wicket fails currently
so i think i fall back to that that i will also look it up as a key
if
the
property is not found.
This will result in more mem en cpu usage.
(i don't create any garbage currently)
On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote:
ognl can do map lookup with just a expression
(that was the first thing i had to do because that did go wrong
when
i
started testing it)
this was used: Ognl.getValue(string, anHashmap);
so the need to do [sd] even with juk why is that needed ...?
On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote:
I'm not too sure about using dots for index/ map references. It
differs from OGNL:
http://www.ognl.org/2.6.7/Documentation/html/LanguageGuide/indexing.html
.
And personally, I think it is usefull to be able to distinct
single
from indexed properties just by looking at the expression. Is it
/that/ more expensive? I mean, optimizing is good, but we don't
have
to over-do it.
Eelco
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End
of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
__
This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please
notify
the system manager.
Katun Corporation -- www.katun.com http://www.katun.com
_
---
This SF.Net email is sponsored by the JBoss Inc.
Get Certified Today * Register for a JBoss Training Course
Free Certification Exam for All Training Attendees Through End of
2005
Visit http://www.jboss.com/services/certification for more
information
___
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Oh, and of course AbstractPropertyModel.ognlExpression should be replaced by e.g. .expression or something similar. Eelco --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Hello all, This is my very first post on this list. Glad to being able to exchange ideas with you. On 10/27/05, Igor Vaynberg [EMAIL PROTECTED] wrote: Why not keep wicket simple? I really like the simplicity of the dot notation johan created. I think we should use [] only as a hint. By default wicket tries to find a getter and then checks if the object is a list/set/map and contains the key. Does Wicket also check for direct public property of the same name ? I mean, this way, we can keep things simple as long as no extra logic is done in the getter/setter ( almost 95 to 99% of the time with GUI Models ? ), but also later on add getters/setters at will, without breaking anything ... Same remark concerning maps vs beans. If find it great if we don't know if the property is contained in the Map or is a property of the Model. Thus, a simple Map at the beginning of the project, whose properties are accessed like myObjectThatMayBeAMapOrBean.myPropertyName can later on evolve to a plain JavaBean without breaking all the String references to the properties : myObjectThatMayBeAMapOrBean.myPropertyName I can see that distinguishing maps/beans properties can be a performance hint: obj[prop] obj.prop but for me it's a shame to sacrifice such power. I can also see that adding [ ] can help identify a map. In my opinion, obj[prop] and obj.prop should be the same operation under the covers : if the object is an instance of a Map: first try to interpret prop as a Map key second try to access prop via getter/setter third try to access prop as a public property of the object if the object is not an instance of a Map: first try to access prop via getter/setter secon try to access prop as a public property of the object My apologizes if such ideas have already been thrown by somebody. My 0.02 EUROS :-) -- Laurent Petit --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Hello, On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote: AbstractPropertyModel could work with something like a propertyresolver/ delegate (one more indirection :)) for getting and setting values based on a expression, and it might even be a good idea to have the option of having the default delegate as a application setting or something. Would be real easy to implement too. I like the idea of one more indirection :-) to allow more than a single implementation. But I can see a problem arise if you do so: having many implementations that eventually will not share the same syntax for accessing properties. This could presumably be a risk for Components Reusability, if some packaged components embark their own models and rely on a certain syntax for specifying access to model properties ? --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Yeah, that's true. You're right. Maybe it isn't such a good idea then. Eelco On 10/27/05, Laurent PETIT [EMAIL PROTECTED] wrote: Hello, On 10/27/05, Eelco Hillenius [EMAIL PROTECTED] wrote: AbstractPropertyModel could work with something like a propertyresolver/ delegate (one more indirection :)) for getting and setting values based on a expression, and it might even be a good idea to have the option of having the default delegate as a application setting or something. Would be real easy to implement too. I like the idea of one more indirection :-) to allow more than a single implementation. But I can see a problem arise if you do so: having many implementations that eventually will not share the same syntax for accessing properties. This could presumably be a risk for Components Reusability, if some packaged components embark their own models and rely on a certain syntax for specifying access to model properties ? --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
RE: [Wicket-user] replace ognl.
sounds awesome, great job Johan! Im +1 for dropping ognl, you covered all my usecases. -Igor From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Johan CompagnerSent: Wednesday, October 26, 2005 4:41 PMTo: wicket-user@lists.sourceforge.netSubject: [Wicket-user] replace ognl. Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example)then i see quite some improvements in cpu speedups and mem improvementsAround 40% speed increase for a submitting the forminput example page 20 timesand only 1/5 of the memory garbage is generated (50MB against 10MB)The question is what do you guys use of ognl? Can i completely drop it or must i make it an option sothat you can switch in youre application for all using ognl or the homebrew wicket impl.Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX)What i do support now is this:person.name (plain properties)person.addresses.0.street (addresses is a list and i take the first element)person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it)so maps and list are seen and the next part of the _expression_ is then the key or the index you can also put values in a mapor append/set to a list:person.addresses.homeaddress = new Address()person.addresses.10 = new Address()if the list size is smaller then 10 then it will appends null to make it that size.addresses can also be an Array but then it won't be able to grow.Ofcourse the person.address.street will just be null if address is null, no exception will be thrownif you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable.So can people live with this? Does anybody uses something different of ognl?johan
Re: [Wicket-user] replace ognl.
I vote to get rid of OGNL too. I know that the Spring Framework MVC uses its own home grown one as well and it has worked perfectly for every use case I've ever had. --AndrewOn 10/26/05, Alexandru Popescu [EMAIL PROTECTED] wrote: #: Johan Compagner changed the world a bit at a time by saying on10/27/2005 1:41 AM :# Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB)John this really makes me wonder, cause I know quite a few projects using ognl and never heard complains about it. Have you noticed such complaints or is this completely new?Most probably the improvement may come from removing the _expression_ parser (as in the examples youhave presented such a parsing is not required), but still I am wondering if the parser should take so long for simple _expression_ as those.sorry for wondering,./alex--.w( the_mindstorm )p. The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the _expression_ is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan---This SF.Net email is sponsored by the JBoss Inc.Get Certified Today * Register for a JBoss Training CourseFree Certification Exam for All Training Attendees Through End of 2005Visit http://www.jboss.com/services/certification for more information___Wicket-user mailing listWicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
Re: [Wicket-user] replace ognl.
Please don't get the impression I haven't believed you; I was just wondering. My opinion on this matter is according to make simple things easy, and complex things possible. I am not aware of how this part is currently implemented, but I guess the best approach would be a value retrieval strategy defaulting to the new implementation, but also allow changing it with an ognl strategy. my 2euroc ./alex -- .w( the_mindstorm )p. #: Johan Compagner changed the world a bit at a time by saying on 10/27/2005 2:34 AM :# It is the expression parser. Thats why i ask it here. How many people are really using extreme things of ognl Because i think 99% is just using the basic what i have implemented now. I tested it all in YourKit. and when i just submit the FormInput example of us 20 times (with some different data and some times errors) i see those results. It is all comes from Ognl.parsExpression() and its call to ognl.JavaCharStream, all the time and all the garbage is comming from that area. And if 20 times submitting a form really kost 40MB (extra) of garbage then i really don't want to know what it cost when 200 persons are working at the same time on my application... I think ognl is just to much for what we want. i attached 2 screen dumps of yourkit. 1 cpu: you see that doPost/doGet (is the same for wicket app) does take 59% of total runtime time but 58% of the total time is spend in Ognl! thats onle 1% that does something else. 2 mem: you see the dump of the singleselectchoice. One of the components of forminput and only that component generated 2MB of garbage in the Ognl part johan On 10/27/05, Alexandru Popescu [EMAIL PROTECTED] wrote: #: Johan Compagner changed the world a bit at a time by saying on 10/27/2005 1:41 AM :# Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) John this really makes me wonder, cause I know quite a few projects using ognl and never heard complains about it. Have you noticed such complaints or is this completely new? Most probably the improvement may come from removing the expression parser (as in the examples you have presented such a parsing is not required), but still I am wondering if the parser should take so long for simple expression as those. sorry for wondering, ./alex -- .w( the_mindstorm )p. The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
RE: [Wicket-user] replace ognl.
Im against introducing yet another strategy into this. We already provide IModel interface which is easy to implement. If someone wants to use ognl instead of the wicket's default they are more then welcome to roll their own impementation of IModel. Maybe we can rename/save our current ognl implementations and offer them as alternatives in the wicket-extensions project. -Igor -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Alexandru Popescu Sent: Wednesday, October 26, 2005 6:49 PM To: wicket-user@lists.sourceforge.net Subject: Re: [Wicket-user] replace ognl. Please don't get the impression I haven't believed you; I was just wondering. My opinion on this matter is according to make simple things easy, and complex things possible. I am not aware of how this part is currently implemented, but I guess the best approach would be a value retrieval strategy defaulting to the new implementation, but also allow changing it with an ognl strategy. my 2euroc ./alex -- .w( the_mindstorm )p. #: Johan Compagner changed the world a bit at a time by saying on 10/27/2005 2:34 AM :# It is the expression parser. Thats why i ask it here. How many people are really using extreme things of ognl Because i think 99% is just using the basic what i have implemented now. I tested it all in YourKit. and when i just submit the FormInput example of us 20 times (with some different data and some times errors) i see those results. It is all comes from Ognl.parsExpression() and its call to ognl.JavaCharStream, all the time and all the garbage is comming from that area. And if 20 times submitting a form really kost 40MB (extra) of garbage then i really don't want to know what it cost when 200 persons are working at the same time on my application... I think ognl is just to much for what we want. i attached 2 screen dumps of yourkit. 1 cpu: you see that doPost/doGet (is the same for wicket app) does 1 take 59% of total runtime time but 58% of the total time is spend in Ognl! thats onle 1% that does something else. 2 mem: you see the dump of the singleselectchoice. One of the 2 components of forminput and only that component generated 2MB of garbage in the Ognl part johan On 10/27/05, Alexandru Popescu [EMAIL PROTECTED] wrote: #: Johan Compagner changed the world a bit at a time by saying on 10/27/2005 1:41 AM :# Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) John this really makes me wonder, cause I know quite a few projects using ognl and never heard complains about it. Have you noticed such complaints or is this completely new? Most probably the improvement may come from removing the expression parser (as in the examples you have presented such a parsing is not required), but still I am wondering if the parser should take so long for simple expression as those. sorry for wondering, ./alex -- .w( the_mindstorm )p. The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name http://person.name http://person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the expression is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan --- This SF.Net
Re: [Wicket-user] replace ognl.
+1 but... I believe that introduceing new syntax of _expression_ language make Wicket hard to learn, especially for those developers with JSP/JSTL background (most of us !) It's better if we can reuse EL (_expression_ language) syntax, a J2EE standard, for new implmentation. On 10/27/05, Johan Compagner [EMAIL PROTECTED] wrote: Hi I have written a replacement of OGNL when i test it with a very simple test (The FormInput example) then i see quite some improvements in cpu speedups and mem improvements Around 40% speed increase for a submitting the forminput example page 20 times and only 1/5 of the memory garbage is generated (50MB against 10MB) The question is what do you guys use of ognl? Can i completely drop it or must i make it an option so that you can switch in youre application for all using ognl or the homebrew wicket impl. Or make seperate classes (like AbstractPropertyModel) but this is not really doable because then all the sub classes must also be copied...(Like CompoundXX) What i do support now is this: person.name (plain properties) person.addresses.0.street (addresses is a list and i take the first element) person.addresses.homeaddress.street (addresses is map and i take the address with the key 'homeaddress' out of it) so maps and list are seen and the next part of the _expression_ is then the key or the index you can also put values in a map or append/set to a list: person.addresses.homeaddress = new Address() person.addresses.10 = new Address() if the list size is smaller then 10 then it will appends null to make it that size. addresses can also be an Array but then it won't be able to grow. Ofcourse the person.address.street will just be null if address is null, no exception will be thrown if you try to set something on a null object a exception is still thrown, Maybe we could make some null handlers for that somehow that are easy useable. So can people live with this? Does anybody uses something different of ognl? johan -- Ingram ChenJava [EMAIL PROTECTED]Institue of BioMedical Sciences Academia Sinica Taiwanblog: http://www.javaworld.com.tw/roller/page/ingramchen
Re: [Wicket-user] replace ognl.
+1 on moving OGNL into extensions. I think the OGNL thing is good, it provides a nice base implementation. I'd like to see the current implementation still being available for those 1% that need the full OGNL expression evaluator (though I think I share the same sentiment Phil voiced). Do we also support things like person.getChildren().get(0).getName()? Martijn On 10/27/05, Phil Kulak [EMAIL PROTECTED] wrote: I'm very +1 on this. OGNL has turned into it's own scripting language over the years. If you want to use any of those features, then you're missing the point of Wicket. --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user --- This SF.Net email is sponsored by the JBoss Inc. Get Certified Today * Register for a JBoss Training Course Free Certification Exam for All Training Attendees Through End of 2005 Visit http://www.jboss.com/services/certification for more information ___ Wicket-user mailing list Wicket-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wicket-user
