Re: dogtail-devel hypertext / hyperlink objects

[Zack Cerza <zcerza redhat com>]

Marius Andreiana wrote:
> Hi,
> 
> As it's my first post here, let me start with congratulations for
> Dogtail, it's a great application!
Thanks :)

> I'm trying to get links on a webpage as separate objects. Now separate
> links are seen as a text string, e.g. 
>        Node roleName='text' name='' description='no description'
> text='Google.ro offered in: Romana  magyar  Deutsch Advertising Programs
> - About Google - Go to Google.com'
> 
> 
> I've implemented Hypertext class support to pyspi (sent patch to
> pyspi-list, I'll followup with Hyperlink) and added this to tree.py
> 249c249
> <       contained = ('__accessible', '__action', '__component',
> '__text', '__editableText')
> ---
> 
>>      contained = ('__accessible', '__action', '__component',
> 
> '__text', '__editableText', '__hypertext')
> 286a287,291
> 
>>              # Swallow the Hypertext object, if it exists
>>              hypertext = self.__accessible.getHypertext()
>>              if hypertext is not None:
>>                      self.__hypertext = hypertext
>>
> 
> 391a397,398
> 
>>              elif attr == "hypertext":
>>                      return self.__hypertext.getHypertext()
Here, you're effectively doing:
self.__accessible.getHypertext().getHypertext() - which won't work.

The whole Hypertext/Hyperlink will probably have to be implemented in
tree.py similar to the way actions work:

class Hyperlink:
	def __init__(self, node, hyperlink, index):
		self.node = node
		[...]

	def __getattr__(self, attr):
		if attr == "URI":
			return self.__hyperlink.getURI()
		[...]
		else: raise AttributeError, attr

and inside Node we'll have:
	def __getattr__(self, attr):
		[...]
		elif attr == "hyperlinks":
			hyperlinks = []
			for i in xrange(self.__hypertext.getNLinks()):
				hyperlinks.append(Hyperlink(self, self.__hyperlink, i)
			if hyperlinks: return hyperlinks
			else: raise AttributeError, attr

So, we don't need to actually expose any Hypertext objects via tree.py -
we'll just skip straight to the Hyperlinks.

> However, the dump of Epiphany tab is exactly the same as before, no
> Hypertext object. What should I do to get the links as objects (and
> click on them from Dogtail) ?
Well, as you can see, there's still a bit more work to do. I guess I
just did most of the necessary tree.py work, though. :)

Even once that is done, you won't see any objects with a roleName of
hyper*. Their roleName will just be text, since Accessible*[Tt]ext
inherit from AccessibleText. This is just the way AT-SPI works.

> And another issue, why are the text entries in web pages (<input
> type='text'...>) have the roleName "text" instead of "editableText" in
> node dump? Would make it easier to find, I don't know how to get the
> text entry in this dump:
This is for the same reason as above. If a Node with a roleName of
'text' is editable, you can just do:

node.text = 'foo'

and maybe handle the exception that will be raised in cast it's not
editable.

Yeah, this is tricky. But I couldn't find a more sane way to implement
it, given AT-SPI's own level of sanity :)

> Thanks in advance for enlightenment :)
No problem at all.

Thanks,
Zack