You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@struts.apache.org by Apache Wiki <wi...@apache.org> on 2007/02/22 01:56:33 UTC
[Struts Wiki] Update of "StrutsAction1Planning" by MichaelJouravlev
Dear Wiki user,
You have subscribed to a wiki page or wiki category on "Struts Wiki" for change notification.
The following page has been changed by MichaelJouravlev:
http://wiki.apache.org/struts/StrutsAction1Planning
The comment on the change is:
Rehashed for 1.4
------------------------------------------------------------------------------
Consider refactoring for Spring. We identified the need for adding a IOC container to Struts some time ago, but stalled on the point of which to use. Since then, Spring has gained a lot of momentum. Spring is used by the !MyFaces and Beehive teams, and its on the radar for Shale. There is already a Struts-Spring component in the Spring distribution and other common ground.
+ 2007-02-09: Struts 1.3 has been branched. 1.3.x maintenance branch has been branched out. The HEAD represents future 1.4 version and is open for improvements.
+
+ === Code Issues ===
+
+ 1. [MichaelJ] '''Consider the option to call all lifecycle functions explicitly''' from an action class, allow to disable automatic form population.
+
+ * [FrankZ] I personally would not want to see the auto-population and validation and such go away. I think them being declarative is a powerful notion. I DO however agree that a developer should be able to turn them off declaratively and do it all manually.
+
+ * Implemented on 2007-02-09: !ActionForm autopopulation is now configurable from an action mapping via "populate" attribute. Automatic call to !ActionForm.reset is now configurable from an action mapping via "reset" attribute. See http://svn.apache.org/viewvc?view=rev&revision=505640
+
+ 1. '''The ability to switch off auto-form population''' [FrankZ] The idea here is that Struts would still instantiate an !ActionForm and call reset(), but that's it. This can be useful if you want to use an !ActionForm only as an output object, but want to handle input manually. This came up for me converting a non-Struts app to Struts, where there was no notion of an !ActionForm in the previous framework.
+ * [MichaelJ] +1 I even had the [http://issues.apache.org/bugzilla/show_bug.cgi?id=38620 ticket] opened for this issue, but as Martin Cooper pointed out, this can be achieved simply by not associating the form with setup action. On the other hand, having form name explicitly in the action mapping makes config file more readable.
+ * Implemented on 2007-02-09: (see previous bullet)
+
+ 1. '''Base Action class should implement event handling a.k.a. dispatching''' [MichaelJ] If event is recognized for an action, it is dispatched to a corresponding method. If event is not found in the request, then execute() is called like for a regular action.
+
+ 1. '''Built-in support for RAP (Redirect After Post) pattern''' [FrankZ] I'm not sure how best to accomplish this, but this should be a very easy thing for a developer to do, the framework should do any required heavy lifting. Again, not so much a rough spot as it is something I think a lot of people would appreciate.
+ * [MichaelJ] To simplify redirect-after-post pattern, Struts must implement Flash scope, that is, the scope that survives after redirect, but is automatically cleaned up after redirected request finishes. Currently similar pattern is applied to the messages queued to session scope.
+
+ 1. '''Pure POJO !ActionForms''' [FrankZ] An !ActionForm should not need to extend !ActionForm. The framework would have to be smart enough to not call validate() and such in that case. This would allow for the Action to be the !ActionForm as well, and this is really the underlying goal because many people view them being separate as a rough spot.
+ * [MichaelJ] Definite +1
+
+ 1. '''Built-in AJAX support''' [FrankZ] I am not entirely sure what the best form for this is, although I believe the [http://struts.sourceforge.net/ajaxtags/index.html AjaxTags] paradigm still has a great deal of merit. In any case, in today's world, not offering some degree of AJAX out of the box is probably a rough spot for many.
+ * [MichaelJ] I believe that I can combine my [http://www.jspcontrols.net/ JSP Controls Tag Library] with Tiles, building a real component subframework on top of Struts, having built-in Ajax features as well (actually, AHAH. It is coarse-grained in-place update of a component within a composite page.) I believe that it can coexist nicely with more finegrained approach of !AjaxTags.
+
+ 1. Implement scope that lasts beyond one roundtrip, but shorter than session (a.k.a. RolloverScope, Conversation scope, Flash scope).
+
+ === Best Practices Issues ===
+
+ 1. '''Dispatch-type actions should be more front-and-center''' [MichaelJ] Dispatch-action style request processing should become a cornerstone technique instead of "yet another way of doing things". It works especially well in data entry form processing or menu processing. See DataEntryForm.
+ * [FrankZ] As an oft-stated opponent, GENERALLY, of Dispatch-type Actions, I wouldn't be thrilled with this being sold as a "best practice". I would have no problem with it being described in more detail and the pluses and minuses stated for people to be able to make a more informed decision either way.
+
+ 1. '''Preference for session-scoped !ActionForms''' [MichaelJ] Best practices should explain that there is nothing wrong in having session-scoped action forms. Best practices should teach how to build stateful web resources.
+ * [FrankZ] -1. Saying there is "nothing wrong" with it, I do not believe is true. Speaking as someone who works all day in a large distributed enterprise environment, I am very well aware of the pitfalls of storing things in session and in letting session get too big. I'm not saying I never store anything in session, of course I do! :) But I think people need to know the positives and negatives of doing so and decide what is appropriate in a given use case rather than being told this or that is a "best practice".
+ * [MichaelJ] My original wording did not have "preference". What I am saying is that session-scoped forms should not be a taboo, and their benefits (as well as pitfalls) should be clearly explained.
+ * [MichaelJ] After Rollover scope is implemented it should be preferred over session scope for redirect-after-post and other multi-request but relatively short term interactions.
+
== Struts Action Framework 1.5.x considerations ==
Based on our own work with the "experimental" members introduced in 1.3.x, we might consider some other changes.
@@ -52, +89 @@
'''Consider a "populate" method on !ActionForm'''. From an OOP standpoint, it might be cleaner if an !ActionForm populated itself rather than rely on a "god" class to populate it from the outside.
'''Consider a "!FormContext" mechanism'''. Rather than "throw-away" a request-based !ActionForm, the object could be seralized as a hidden-field or session-object and restored on the next request. Many other frameworks support this behavior now. Struts would have a slightly different spin, since we look at the form as an named entity rather than as an anonymous aggregation of other objects.
+
+ 1. '''Building stateful components''' [MichaelJ] Best practices should teach how to build stateful and independent web components with Struts.
+
+ 1. '''!ActionForm as true I/O buffer''' [MichaelJ] !ActionForm should be used as I/O buffer instead of simply collecting user input from request.
+ * [FrankZ] Could you explain this further? I'm not at all clear on what you mean.
+ * [MichaelJ] This is what 60% of Struts users do anyway. I guess you do it as well. You have actionform associated with input action. You submit a form, actionform is populated. In case of error or whatnot you render the same JSP page and pull the data from the actionform, it is already there, saved for you by Struts. I use this all the time for data entry use case, but this does not seem to be an officially endorsed and advertised practice.
== Struts Action Framework 1.6.x considerations ==
@@ -75, +118 @@
1. [MichaelJ] '''Consider bringing [https://formdef.dev.java.net/ FormDef] into Struts core''' and recommend using dynaforms with nested business objects as a best practice. Reason: I/O conversion is the suckiest part of Struts 1.
- 1. [MichaelJ] '''Consider the option to call all lifecycle functions explicitly''' from an action class, allow to disable automatic form population.
-
- * [FrankZ] I personally would not want to see the auto-population and validation and such go away. I think them being declarative is a powerful notion. I DO however agree that a developer should be able to turn them off declaratively and do it all manually.
-
1. [MichaelJ] '''Deprecate automatic validation'''. Newbies always stumble upon the fact that their action class is not called when validation fails. Instead, promote explicit validation.
* [FrankZ] Definite -1 from me. Again, I'm +1 to being able to turn it on and off, but I very much believe it should be there. Perhaps there is some room for better logging in the cases you mention?
@@ -88, +127 @@
1. [MichaelJ] '''Add "form" attribute''' with the same meaning as "name" attribute; deprecate "name".
* [FrankZ] +0. Same comment as the above point.
- 1. [MichaelJ] '''Logically differentiate forward from redirect in mappings''' "forward" tag implements both forward and redirect, this is confusing. Implement two separate tags like "render" for forward and "transfer" for redirect. "forward" tag still can be kept for compatibility purposes.
- * [FrankZ] Hmm, I'm not sure how I feel about that. Maybe simply adding a type attribute, ala Webwork, would do the trick? "forward" or "redirect" as values?
- * [MichaelJ] Check out this mapping from one of the users of Struts Dialogs 1.x (he uses Tiles too):
-
- {{{
- <component path = "/userCenter"
- type = "com.acme.action.UserCenterAction"
- form = "userCenterForm"
- view = "page.userCenter">
- <event name = "DIALOG-EVENT-CHANGE" handler = "onUpdate"/>
- <event name = "DIALOG-EVENT-DELETE" handler = "onDelete"/>
- <event name = "DIALOG-EVENT-SHOW-ORDERS" handler = "onShowOrders"/>
- <event name = "DIALOG-EVENT-SHOW-PRICE-LIST" handler = "onShowPriceList"/>
- <event name = "DIALOG-EVENT-SHOW-ORDER-DETAILS" handler = "onShowOrderDetails"/>
-
- <transfer name = "SHOW-ORDERS" path = "/showOrders.do"/>
- <transfer name = "SHOW-PRICE-LIST" path = "/showPriceList.do"/>
- <transfer name = "SHOW-ORDER-DETAILS" path = "/showOrderDetails.do"/>
-
- <render name = "ON-CC-DIALOG" path = "page.CCDialog"/>
- <render name = "ON-GET-DELIVERY-CONFIRMATION" path = "/jsp/deliveryConfirmation.jsp"/>
- </component>
- }}}
-
- This mapping uses the action mapping format introduced in Struts Dialogs project. A "component" is essentially an "action" with different defaults and some new elements. This is 1:1 correspondence between mapping and action class. Same action class handles input phase as well as render phase (I saw your remark about Redirect-After-Post, here it is, implemented).
-
- 1. [MichaelJ] '''All dispatching actions but !EventDispatchAction should be deprecated'''. Reason: having so many dispatching actions is confusing.
- * [MichaelJ] On a second thought, let's '''make all actions to be dispatching actions'''. If event is recognized for an action, it is dispatched to a corresponding method. If event is not found in the request, then execute() is called like for a regular action.
-
1. [FrankZ] '''Actions should be instantiated for each request''', therefore removing the thread safety concerns that exist today.
* [MichaelJ] -0. Not sure that this bothers me anymore, especially if dynaforms will be adopted as standard practice. In this case a dynaform would be a container for a nested business object. I would rather combine Action and !ActionForm and have an option to choose scope just right now a scope can be chosen for an !ActionForm. That is, I am against strictly per-request actions.
-
- 1. '''Pure POJO !ActionForms''' [FrankZ] An !ActionForm should not need to extend !ActionForm. The framework would have to be smart enough to not call validate() and such in that case. This would allow for the Action to be the !ActionForm as well, and this is really the underlying goal because many people view them being separate as a rough spot.
- * [MichaelJ] Definite +1
-
- 1. '''Built-in AJAX support''' [FrankZ] I am not entirely sure what the best form for this is, although I believe the [http://struts.sourceforge.net/ajaxtags/index.html AjaxTags] paradigm still has a great deal of merit. In any case, in today's world, not offering some degree of AJAX out of the box is probably a rough spot for many.
- * [MichaelJ] I believe that I can combine my [http://www.jspcontrols.net/ JSP Controls Tag Library] with Tiles, building a real component subframework on top of Struts, having built-in Ajax features as well (actually, AHAH. It is coarse-grained in-place update of a component within a composite page.) I believe that it can coexist nicely with more finegrained approach of !AjaxTags. Is it possible to combine AjaxTags with JSP Controls? My goal was to make sure that components work even if Javascript is turned off.
1. '''Custom attributes on tags''' [FrankZ] All Struts tags that render HTML should allow for arbitrary attributes. I again propose a "specCompliant" attribute on the <html:html> tag. When true, no arbitrary attributes are allowed (this would be the default if the attribute is not present). If false, any attribute can be added. This has been a hassle for me a couple of times where I wanted to store some custom information on a tag for client-side purposes.
* [MichaelJ] +1
- 1. '''The ability to switch off auto-form population''' [FrankZ] The idea here is that Struts would still instantiate an !ActionForm and call reset(), but that's it. This can be useful if you want to use an !ActionForm only as an output object, but want to handle input manually. This came up for me converting a non-Struts app to Struts, where there was no notion of an !ActionForm in the previous framework.
- * [MichaelJ] +1 I even had the [http://issues.apache.org/bugzilla/show_bug.cgi?id=38620 ticket] opened for this issue, but as Martin Cooper pointed out, this can be achieved simply by not associating the form with setup action. On the other hand, having form name explicitly in the action mapping makes config file more readable.
-
1. '''Built-in dependency injection''' [FrankZ] This should be modeled after what is offered in JSF. If we took the code from the !DependencyFilter in [http://javawebparts.sourceforge.net/ Java Web Parts], added in true injection (shouldn't be a terribly big deal), I believe we would have even better capabilities than in JSF. Spring is of course still out there for those that need/want more! This may not be so much a rough spot as just a fairly low-hanging piece of fruit (because most of the work is already done by virtue of leveraging !DependencyFilter) that I think people would appreciate having.
1. '''!ThreadLocal !ActionContext''' [FrankZ] Yes, I think this is one place we should flat-out rip off Webwork :) Backwards-compatibility would have to be considered, but I'd like Actions to have to deal with a single object, and for that object to be accessible via the !ThreadLocal mechanism. This should also open the door for POJO Actions.
-
- 1. '''Built-in support for RAP (Redirect After Post) pattern''' [FrankZ] I'm not sure how best to accomplish this, but this should be a very easy thing for a developer to do, the framework should do any required heavy lifting. Again, not so much a rough spot as it is something I think a lot of people would appreciate.
- * [MichaelJ] Redirect-After-Post is fully implemented in [http://struts.sourceforge.net/strutsdialogs/index_v1.html Struts Dialogs 1.x]. After Struts 1.2.9 got better dispatch action, I stopped advertising Struts Dialogs project, but it is still valuable in terms of samples. Check it out. EventActionDispatcher and !EventDispatchAction can implement Redirect-After-Post as well, Paul and I spent quite some time discussing how !EventDispatchAction can be used for both input and render phases.
- * [MichaelJ] To simplify redirect-after-post pattern, Struts must implement Flash scope, that is, the scope that survives after redirect, but is automatically cleaned up after redirected request finishes. Currently similar pattern is applied to the messages queued to session scope.
1. '''Pre and post-processing abilities''' [FrankZ] A developer should be able to specify a class and method to call before and after an Action executes, on a per-mapping basis. This should be independant of modifying a chain. Should just amount to adding the appropriate config file changes and two commands to the default chain. This is for handling things like common setup of view-type Actions, etc.
* [MichaelJ] +0. I prefer having 1:1 correspondence between mapping and action class. With autopopulation out of the way, I can call whatever I want right in the beginning and at the end of execute() method. Same thing, but simpler, I think.
@@ -162, +159 @@
* [FrankZ] Speaking for myself, I would need to see a good definition of this approach before I was +/- on it. I have never seen a clear explanation that made me say "oh yeah, I see, that's better!"
* [MichaelJ] Check out DataEntryForm. Well, it uses dispatching action and session-scoped form, but if you look at it from 10000 ft, it gives the idea of what web resource is about. Basically, it is one-two actions, one form and one or more JSPs. A resource does NEVER forwards to a JSP page that does not belong to the resource. In the best case we can have nice and simple 1:M relationship between action and JSPs.
- 1. '''Dispatch-type actions should be more front-and-center''' [MichaelJ] Dispatch-action style request processing should become a cornerstone technique instead of "yet another way of doing things". It works especially well in data entry form processing or menu processing. See DataEntryForm.
- * [FrankZ] As an oft-stated opponent, GENERALLY, of Dispatch-type Actions, I wouldn't be thrilled with this being sold as a "best practice". I would have no problem with it being described in more detail and the pluses and minuses stated for people to be able to make a more informed decision either way.
-
- 1. '''Preference for session-scoped !ActionForms''' [MichaelJ] Best practices should explain that there is nothing wrong in having session-scoped action forms. Best practices should teach how to build stateful web resources.
- * [FrankZ] -1. Saying there is "nothing wrong" with it, I do not believe is true. Speaking as someone who works all day in a large distributed enterprise environment, I am very well aware of the pitfalls of storing things in session and in letting session get too big. I'm not saying I never store anything in session, of course I do! :) But I think people need to know the positives and negatives of doing so and decide what is appropriate in a given use case rather than being told this or that is a "best practice".
- * [MichaelJ] My original wording did not have "preference". What I am saying is that session-scoped forms should not be a taboo, and their benefits (as well as pitfalls) should be clearly explained.
-
- 1. '''Building stateful components''' [MichaelJ] It is possible to build stateful and independent web components with Struts 1. Best practices should teach how to do that.
-
- 1. '''!ActionForm as true I/O buffer''' [MichaelJ] !ActionForm should be used as I/O buffer instead of simply collecting user input from request.
- * [FrankZ] Could you explain this further? I'm not at all clear on what you mean.
- * [MichaelJ] This is what 60% of Struts users do anyway. I guess you do it as well. You have actionform associated with input action. You submit a form, actionform is populated. In case of error or whatnot you render the same JSP page and pull the data from the actionform, it is already there, saved for you by Struts. I use this all the time for data entry use case, but this does not seem to be an officially endorsed practice.
-
1. '''Use of nested properties''' [MichaelJ] Emphasize using of nested properties, using business objects as nested properties, using dynaforms holding business objects.
== Relevant Proposals ==