# Nokogiri bug?

**URL:** <https://www.ruby-forum.com/t/nokogiri-bug/193502>\
**Category:** Ruby\
**Created:** [August 18, 2010, 9:58pm UTC](https://www.ruby-forum.com/t/nokogiri-bug/193502 "2010-08-18T21:58:34Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![asoto](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/a/58956e/32.png) [@asoto](https://www.ruby-forum.com/u/asoto)\
**Post date:** [August 18, 2010, 9:58pm UTC](https://www.ruby-forum.com/t/nokogiri-bug/193502/1 "2010-08-18T21:58:34Z")

</div>

Hi Folks - Having some trouble w/ Nokogiri & its handling of xpaths and  
nodesets, wondering if anyone here knows about possible existing issues.  
(Apologies if this is a bit off-topic, but I am working in Ruby…)

So I’m trying to access a form’s action URL so that I can construct my  
own query string to GET via mechanize. I’ve tested just about every  
possible xpath to get to the form’s action property, and Nokogiri is  
behaving as if it doesn’t exist.

What I’ve discovered is that, once I hit a particular div tag, which  
happens to be the parent node of the form, Nokogiri doesn’t see any of  
the div’s children.

The page has HTML to this effect:

Note the form doesn’t have an ID or class by which to directly reference  
it. Trying to access it like so:

sublnk = page.xpath("//div[@id=‘app-#{code}’]")

If I call children() on that, it’s empty. But the parent of this node  
works fine, and Nokogiri constructs its children properly, including  
containing the form as a child of “app-#{code}”:

sublnk =  
page.xpath("//div[@id=‘page-content’]/div[@id=‘summary-section’]").children

Additionally, accessing the form directly via its index won’t work  
either. It’s the second form on the page, so I’ve tried:

sublnk = page.xpath("//form[2]/@action")

…and still get nothing. It’s as if this node just doesn’t exist to  
Nokogiri, even though it clearly does when it’s a child. So am I doing  
something wrong? How come //div[@id='app-#{code}] doesn’t have any  
children? Is there a bug in Nokogiri which prevents it from accessing  
the form node? Why won’t any xpath to the form work?!? Please help me  
stop banging my head on my desk.

Thanks!

-Alex

---

<div class="post-metadata">

**Author:** ![asoto](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/a/58956e/32.png) [@asoto](https://www.ruby-forum.com/u/asoto)\
**Post date:** [August 18, 2010, 10:55pm UTC](https://www.ruby-forum.com/t/nokogiri-bug/193502/2 "2010-08-18T22:55:57Z")

</div>

Hi,

On Wed, Aug 18, 2010 at 3:58 PM, Alex S. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> Hi Folks - Having some trouble w/ Nokogiri & its handling of xpaths and  
> nodesets, wondering if anyone here knows about possible existing issues.  
> (Apologies if this is a bit off-topic, but I am working in Ruby…)

A better place to discuss Nokogiri-specific questions like this is  
probably  
the nokogiri-talk mailing list, which you can find more about here:

```
http://groups.google.com/group/nokogiri-talk

```

> So I’m trying to access a form’s action URL so that I can construct my  
> own query string to GET via mechanize. I’ve tested just about every  
> possible xpath to get to the form’s action property, and Nokogiri is  
> behaving as if it doesn’t exist.

When you email the nokogiri-talk list, please make sure you include the  
entire HTML document you’re dealing with (e.g., as a gist or pastie), so  
that someone can first reproduce your issue and then diagnose it or give  
you  
some advice.

Cheers,  
-mike

---

<div class="post-metadata">

**Author:** ![asoto](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/a/58956e/32.png) [@asoto](https://www.ruby-forum.com/u/asoto)\
**Post date:** [August 18, 2010, 11:17pm UTC](https://www.ruby-forum.com/t/nokogiri-bug/193502/3 "2010-08-18T23:17:01Z")

</div>

Thanks, I thought there might be a more appropriate forum.

But fortunately I won’t have to post there - figured it out.

To find & verify the xpaths, I’d been using firebug. Then looking in  
the code I noted that firebug’s model wasn’t matching the actual HTML,  
which in turn didn’t match what nokogiri saw - they each showed  
something different (bug elsewhere?). Once I figured out what HTML  
nokogiri was parsing, it was easy.

Thanks,  
Alex