[TOMAHAWK-914] t:dataTable style attributes don't work with Facelets

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.1.4-SNAPSHOT, 1.1.5-SNAPSHOT, 1.1.7-SNAPSHOT
• Fix Version/s: 1.1.7
• Component/s: Extended Datatable
• Environment: MyFaces Core 1.1.5, Facelets 1.1.11
• Assignee: Unassigned
• Reporter: Jeff Bischoff
• Created: 28/Feb/07 20:52
• Updated: 03/Sep/08 04:46
• Resolved: 29/Mar/07 18:08

Description

Problem: style and styleClass attributes on t:dataTable do not work in Facelets due to use of fully-qualified names in the map.

Fix: Patch attached to resolve this by changing the names as stored in the map. Includes code to accept hacks based on the old behaviour, but warns that it is now deprecated.

Bonus: Also includes fix for problem in Facelets where the EL can not return a null style. This is due to changes in the EL spec, and prevents the former (very useful) style rollover behaviour. Fix works by converting any empty String returned by the EL in these Style attributes into null. (Reverse Coercion)

Background:

After converting my application from JSP to Facelets, I set out to make my "rowStyleClass" attribute on t:dataTable work like it used to.

First, I had to get the attribute working in Facelets. With considerable discussion on the user list (see [1] and [2]) and a lot of help from Mike, I think we've identified some pretty simple code changes to enable this and other similar attributes. I plan to introduce a patch for this, probably tomorrow.

What happened next was that during testing of this change, I could confirm that the attribute did indeed now work, but I was baffled by unexpected behaviour. My EL expression which had previously returned null in certain situations was now returning the empty String. I went to Facelets list for some clarification on this (see [3]) and it turned out to be a requirement of the new EL spec to coerce the nulls into empty string.

Getting an empty String instead of null for this ValueBinding lookup creates a problem because the extended dataTable treats a null value differently and goes looking at the more standard style attributes like rowClasses. With an empty string returned, it assumes it needs to look no further. As a user, there is no way for me to specify that under certain conditions, it should fallback to the other style attributes, e.g. rowClasses.

The best fix we have collectively come up with so far is to coerce the empty string back into a null in the dataTable, so that the renderer does the right thing. I don't see too much downside to this approach, as an empty style string has no relevance in CSS. However, I feel a proposed change like this requires extra discussion before making a decision on it. It's another one of those situation where we have to decide how we want to handle unexpected results due to changes in the newer specs.

[1] https://facelets.dev.java.net/servlets/ReadMsg?list=users&msgNo=6875
[2] http://www.nabble.com/Re%3A-Facelets-support-for-a-Tomahawk-dataTable-trick--tf3236491.html
[3] https://facelets.dev.java.net/servlets/ReadMsg?list=users&msgNo=6941

Regards,

Jeff Bischoff
Kenneth L Kurz & Associates, Inc.

Issue Links

• incorporates: TOMAHAWK-523 rowStyleClass does not resolve to an EL of var