Seam conditional render without parsing - jsf

I'm trying to make a conditional render in my Seam application (2.2.0), to display two different controls depending on a condition.
I'm using the s:fragment tag with the render attribute, but my problem is that I want whatever the control is displayed, to have the same id:
<s:fragment render="${editable}">
<rich:calendar id="entityDate"..../>
</s:fragment>
<s:fragment render="${!editable}">
<h:outputText id="entityDate".../>
</s:fragment>
My problem is that even when the render attribute set to false, the "not to be rendered" element is parsed, and I get an exception because of the duplicated id.
I also tried with the tag <ui:remove>, which effectively removes the element before the parsing phase, so I can have something like:
<span id="myId"/>
<ui:remove>
<span id="myId"/>
</ui:remove>
Unfortunately the <ui:remove> tag doesn't allow conditional logic. Has anyone found a way to solve this?

That's only possible when you use a view build time tag such as JSTL <c:if>.
<c:if test="#{editable}">
<rich:calendar id="entityDate" />
</c:if>
<c:if test="#{!editable}">
<h:outputText id="entityDate" />
</c:if>
(note that this is not going to work within an iterablte JSF component, such as <ui:repeat>, <h:dataTable> and so on)
After all, I strongly recommend to take benefit of the disabled attribute instead, if necessary with a good shot of CSS to hide the input field borders and so on. It'll minimize the JSF view boilerplate code.
<rich:calendar id="entityDate" disabled="#{!editable}" />
Disabled inputs are separately styleable by the CSS attribute selector element[attribute], e.g.
input[disabled] {
border: 0;
}
The above removes the border of input elements with the disabled attribute present so that it look like a normal output text.

"Solve"? There is nothing to solve here: two elements in a GUI can not have the same ID. Hardly unnatural or unsound?
It's like asking: "I have a database table with two rows, I would like them both to have the same primary key value, but somehow I get these errors... has anyone managed to solve the problem and circumvent the constraints?".
Or even closer analogy: "I have two spans, one of them is invisible (has style="display: none") I would like them both to have the same id - and browsers seem not to like it, despite one of the spans being invisible".
Bottom line: rendered on not rendered, each component is still a part of the view tree, and therefore has a UNIQUE id.
I have a suspicion that you want to have some "polymorphic" code that should work with the currently visible element. Using ID for such code IS WRONG. If you show us your use case, we might find a right way to achieve the effect.

I use selenuim myself in a seam environement and i recommend using defined ids whenever possible. First you have the ability to create smaller ids which is usefull for pagesize. Second the selenium test run alot faster if you use ids for referencing instead of other selectors. I have not yet found a selenium test where you cannot handle diffrent ids. Additionally if a code fails in jsf tree creation you see which id is failing.
I see you are using sfragment with editable or not. I use the sdecorate and give the decorate an id and then "just" ed for the input and vi for the outputtext for example. This makes it easy in selenium to check the availability of edit or view components.

Would you be able to get the same results you need by putting the id tag on the fragment?
So:
<s:fragment id="entityDate">
<rich:calendar render="${editable}" />
<h:outputText render="${!editable}" />
</s:fragment>

Related

Refactoring a DataTable with a large number of attributes

I've got a data table with 17 attributes. The table can be rendered in two modes: with row selection enabled and without it.
<p:dataTable selection="#{isDefaultSelectionMode ? null : widget.selected}" />
It doesn't work because selection expects a reference to a property to be able to set/get it.
I could create a dummy property widget.ignored and it's going to work. I don't like this for the obvious reason.
<p:dataTable selection="#{isDefaultSelectionMode ? widget.ignored : widget.selected}" />
I could split the table into two separate templates. I would exclude selection from one and duplicate 16 other attributes. It's not a good one, either.
I am looking for an elegant solution to either make the attribute optional (not to render it under some condition) or to avoid defining a dummy property.
I am new to JSF and PrimeFaces, feel free to correct. Any help would be welcomed.
Fortunately, I didn't have to apply any of my terrible workarounds.
As suggested by #Kukeltje (thank you) and the links he provided, I defined the attribute conditionally
<c:if test="#{isDefaultSelectionMode}">
<f:attribute name="selection" value="#{widget.selected}"/>
</c:if>
For more details, visit these questions:
JSF 2.0 dynamic attributes without creating new components
How not to set an attribute of a component inside a composite component if it is empty?
What is f:attribute used for in this example?

Rendering hidden span text inside h:commandLink

We have an accessibility requirement to render certain command links with additional "off screen" text for screen readers. So we want to end up rendering something like (attributes omitted for clarity)
<a>Edit Details<span class="hiddenOffScreen"> for John Smith</span></a>
The problem is the standard jsf1.2 <h:commandLink> tag does not respect the escape attribute. I tried something like <h:commandLink escape="false" value="#{linkText}"/> where linkText evaluates to the contents of the a tag shown above but this renders the span tag literally (i.e escapes the < and >)
How best to go about meeting this requirement? I can of course easily add the span later with JQuery however, for my own education I'd like to have a try with a custom renderer - but not sure how I would hook in with the existing default renderer which adds the "onClick" event handler and associated javascript. My google-foo seems to have failed me when searching for custom commandLink renderer.
Bear in mind, this is JSF 1.2 and we cannot use any third party tag libraries as we're running on a braindead very old version of WebSphere Portal Server.
You can nest content, including other tags and/or implicit text, within the <h:commandLink> tag, instead of specifying text within its value attribute, to achieve your functionality:
<h:commandLink ...>
<h:outputText .../>
<span class="hidden"></span>
#{bean.someText}
</h:commandLink>
This will render exactly what you want.
Use < or > to write < >

Conditional string concatenation/building in JSF

I'm trying to convert a JSP Page to JSF (no JSP code allowed), but have stumbled upon an issue.
Note: This is academic, so no "dirty" solution will do.
I have jsp code that sets the image source and alt text of an image depending on various conditions. There are concatenations and switch conditions. This is inside a loop.
Now, I can reproduce the loop that goes through all the values, like this:
<ui:repeat value="#{gameapi.game.fieldsList}" var="field">
<h:graphicImage id="field#{field.fieldNr}" styleClass="field#{field.fieldNr}"
url="..." alt="" title=""/>
</ui:repeat>
However, there's a total of over 20 combinations for the image url and alt text. Obviously, I doubt writing a conditionally rendered or chosen graphicImage Tag for every possibility is an elegant solution, since this would only get uglier with every additional combination.
I also cannot create those strings in the underlying java code, since that would violate the idea of separating view, model and controller.
So what is the best solution to do this? It seems like a huge weakness of JSF.
Consider using a variation on the MVVM pattern. You can have managed beans that are dedicated to view logic separate from the business logic managed beans.
<ui:repeat value="#{bizBean.list}" var="_row">
<h:graphicImage
url="#{viewModelBean.images[_row.outcome]}"
alt="#{viewModelBean.alts[_row.outcome]}"
title="#{viewModelBean.titles[_row.outcome]}" />
</ui:repeat>
viewModelBean could be an application-scoped type with a bunch of map properties. This assumes a relatively simple case, but the pattern is suited to more sophisticated requirements.
Note: the repeating control should be a NamingContainer, so trying to set the client identifier in the VDL is redundant. See here for more.

Why prependId="false" in a jsf form?

I know what prependId="false" does. It set the flag so that the id of the form does not prepend the id of the form child, but why? any particular reason why you do or dont want to prepend id?
In my experience, I never use this attribute. However, in some cases it can be useful.
When you use Facelets, you can create templates or include pages inside another page. So you can imagine that a page could be included in several different pages. Take the example where the parent pages contain a form, with different id:
Page 1:
<h:form id="form1">
<ui:include src="pages/my-page.xhtml"/>
...
</h:form>
Page 2:
<h:form id="form2">
<ui:include src="pages/my-page.xhtml"/>
...
</h:form>
Now, in the my-page.xhtml, you have a <h:inputText id="foo"/>. In the first case, the real ID of the input will be form1:foo, while in the second case, it will be form2:foo. This could create complex situations if you need a direct access to this component in Javascript or in Java (using findComponent("...") method).
If you use prependId="false" (or on some components forceId="true"), the real ID will be simply foo, and then your code will be simpler as you will not have to care about the container of the input field.
However, you will have to use this attribute carefully, as you may get a duplicate ID error if you use this prepend attribute too often...
In modern jsf versions it might also break ajax, see UIForm with prependId="false" breaks <f:ajax render>
A situation where prependId=false is useful is in the login form, if you are using Spring Security, because the ids of the inputtexts have to be exactly "j_username" and "j_password". So you shouldn't put the form id before them, and using prependId=false is a good choice to acheive this.
I prefer to add prependId occasionally to make styling elements via their ID classes easier. For example, a form:
<h:form id="myform" ... >
<h:inputText id="mytext" ... />
</h:form>
Would give you an ID of myform:mytext. As the colon is reserved in CSS, you have to escape the CSS to read something like #myform\:mytext { ... } which I prefer not to do. With prependId="false" I get to use just #mytext { ... } which is much simpler & nicer to read. It also plays nicer with CSS preprocessors like LESS or Sass.
One scenario where we have to set this flag is in case of Autocomplete control of primefaces library.
I had to set this flag to false when I was trying AutoComplete control of primefaces library. I was not able to get autocomplete working but after setting this flag it worked fine. You can see this link to my question regarding this problem
WARN [Parameters] Parameters: Invalid chunk ignored. warning coming in primefaces application
In addition to making for CSS selectors easier, using prependId=false makes it easier to use JavaScript and jQuery to access specific elements.
Otherwise, without using RichFaces, to get at an elmement by id using jQuery you'll have to use an ugly escape sequence like:
jQuery("form-id\\:element-id")

JSF loop reRender

Hopefully the title isn't too cryptic ...
The problem we have is that we generate a bunch of input controls (h:inputOneMenu, h:inputText etc) from some Java List.
Works fine EXCEPT the requirement is that these inputs validate on the fly. Again not so hard except that as there controls were generated in a loop the only possible reRender action is basically the entire form or an a4j:outputPanel around each loop iteration which is basically the same thing.
Now the above two solutions technically work but they have the nasty side effect of reRendering all the page controls which makes the page feel really twitchy and clunky. We'd like to stop this from happening so ideally the only control that gets reRendered is the control that send the ajax update/validation.
Basically this is our page code:
<ui:repeat value="#{seam-outjected-list}" var="item">
<a4j:outputPanel selfRendered="true">
<h:inputText value=#{item.value}>
<a4j:support event="onblur" ajaxSingle="true" />
</h:inputText>
</a4j:outputPanel>
</ui:repeat>
I've left out a bit of stuff that just renders different controls depending on the item.
As you can see we're currently using the a4j:outputPanel solution so every time any loop generated control is updated all the controls are reRendered.
Thanks in advance if anyone has any thoughts.
My first thought is that you should try replacing your <ui:repeat> with an <a4j:repeat> and take advantage of the ajaxKeys attribute to only reRender certain rows.
From the Richfaces Docs:
The main difference of this component
from iterative components of other
libraries is a special "ajaxKeys"
attribute. This attribute defines row
keys that are updated after an Ajax
request. As a result it becomes easier
to update several child components
separately without updating the whole
page.

Resources