[regression] [patch] [cla] dojo.query not working for attribute selector that includes a tilde (~) character

[juzzam]

dojo.query fails to query an attribute selector with a value including a tilde character which should be supported (the value encapsulated in quotes of an attribute selector is a CSS String which can have this character). This has been confirmed to work with jquery by Les Szklanny on the dojo-interest list.

A code example is attached.

[Karl Tiedt]

Updating this with the info I sent to the ML

The problem has been isolated, a fix is in the works, might be worth not jumping the jQuery ship just yet ;)

In the interest of a regex guru out there here are the problem lines

​http://scsys.co.uk:8002/38390 (ignore the line #s thats with debug code added, the real line #'s are 1187-1198

output from the debugging lines show (irrelevant debug snipped)

a[href="​http://abc.com/~123"] before infixSpaceRe a[href="​http://abc.com/ ~ 123"] after infixSpaceRe

a[href="​http://abc.com/_123"] before infixSpaceRe a[href="​http://abc.com/_123"] after infixSpaceRe

[Adam Peller]

see r20851

[James Burke]

Fixing this will be difficult without compromising the "normalize missing spaces in a query" fix. In the meantime though, this kind of chain can help avoid the issue:

dojo.query('a[href*="http://abc.com/"]').filter(function(node){ return node.href.indexOf("~") !== -1; })

[Douglas Hays]

Does this work?

var infixSpaceRe = /"[^"]*"|'[^']*'|n\+\d|([^ ])?([>~+])([^ =])?/g;

[Michael Bannister]

I'm finding the same thing with plus (+) in a query like this:

dojo.query('option[value="abc+def"]', selectNode);

I've tried it without the scope, too, though in my case the code is out of my control as it's part of setting up a dijit.form.FilteringSelect? using a <select> element. I'll try to set up and attach a test case later.

This is blocking my project upgrading to 1.4 from 1.3, and I want to start using dojo.cache! :(

[Michael Bannister]

Patch attached, including additions to test case. I would really appreciate feedback as this is my first ever contribution like this :)

[Karl Tiedt]

possible fix for 10651 (renamed to .patch)

[Karl Tiedt]

I replaced your patch with one that was named .patch (trac will colorize it then) (both .patch and .diff are treated specially)

[bill]

Mike, could you file a ​CLA so we can look at your patch?

[Michael Bannister]

Replying to bill:

Mike, could you file a ​CLA so we can look at your patch?

Done (by email) - I meant to say that I'd do this today as I don't have a printer at home!

Michael

[bill]

Cool, thanks!

[Michael Bannister]

possible fix for 10651 (corrected)

[Michael Bannister]

I discovered that Safari/Chrome? and Firefox delivered slightly different results when my RegExp? hit the end-of-string match: Safari/Chrome? passed undefined to the String.replace callback function whereas Firefox gave ""... so I was getting 'undefined' tacked onto simple selectors!

Corrected patch submitted.

[bill]

(sadly) punting seemingly abandoned ticket and meta tickets to future

[ben hockey]

sounds related to #7479

[bill]

Michael's patch fixes the bug, so I'm going to check it in.

[bill]

In [27952]:

Fix to not apply spacing normalization inside of [ ], thanks michaelbannister!, fixes #10651 !strict.

[Colin Snover]

In [30033]:

Backport [29124], [27952], and [27961] to 1.7. According to the original changesets, this fixes #14877, #14876, #10651, #11189. !strict