table row_count incorrect

[Paul Denize]

The following watir code is enough to demonstrate the flaw in the
row_count method.
The table has 2 rows - but watir return 6.

The problem is tables embedded in tables and the rowcount simply
counting TR tags from the outside. Therefore it counts inner table
rows as well

Code can then fail when cycling through the rows - the code below
fails when you try to access row 3 (of the 6 rows) when in fact there
are only 2 rows.

----

ie = Watir::IE.new()
ie.goto("https://www.google.com/accounts/NewAccount")
t = ie.table(:index,1)
puts t
t.row_values(3)

[Paul Denize]

This is not the only routine and while digging I found comments like

# This method returns the table as a 2 dimensional array. Dont expect
too much if there are nested tables, colspan etc.

All the routines seem to use the following process

rows = <table object>.getElementsByTagName("TR").length

Now looks more like a limitation than a bug

[Paul Denize]

Overloading the following seems to get past the limitations. Perhaps
someone with more DOM experience can comment if there are any dangers

One thin I did notice is that it picks up <TH> data cells just like
<TD> ones


class Table
def row_count
assert_exists
return @o.rows().length
end

def to_a
assert_exists
y = []
table_rows = @o.rows()
for row in table_rows
puts row.to_a.inspect
x = []
for td in row.cells()
x << td.innerText.strip
end
y << x
end
return y
end

end

[Alan Baird]

Paul -

This issue is WTR-26 (http://jira.openqa.org/browse/WTR-26), it's been around for a while.  Your solution looks like the one that is in this ticket.

Alan

[Moochie]

Someone posted this code and it worked really well for me.

#Get actual row count when enbedded tables exists
def get_row_count_minus_enbedded_rows(table)
num_rows=0
begin
(1..table.row_count).each do |i|
num_rows = i
table[i]
end
rescue
end
return num_rows
end

On Oct 23, 10:08 pm, "Alan Baird" <aba...@bairdsnet.net> wrote:
> Paul -
> This issue is WTR-26

> (http://jira.openqa.org/browse/WTR-26)<http://jira.openqa.org/browse/WTR-26>,

> it's been around for a while.  Your solution looks like the one that is in
> this ticket.
>
> Alan
>

> On Thu, Oct 23, 2008 at 9:10 PM, Paul Denize <paul.den...@datacom.co.nz>wrote:
>
>
>
>
>
> > Overloading the following seems to get past the limitations.  Perhaps
> > someone with more DOM experience can comment if there are any dangers
>
> > One thin I did notice is that it picks up <TH> data cells just like
> > <TD> ones
>
> > class Table
> >    def row_count
> >        assert_exists
> >        return @o.rows().length
> >    end
>
> >    def to_a
> >      assert_exists
> >      y = []
> >      table_rows = @o.rows()
> >      for row in table_rows
> >        puts row.to_a.inspect
> >        x = []
> >        for td in row.cells()
> >          x << td.innerText.strip
> >        end
> >        y << x
> >      end
> >      return y
> >    end
>

> > end- Hide quoted text -
>
> - Show quoted text -

[Nathan Lane]

I'm wondering whether or not this should be a defect really, because according to this standards document about tables (http://www.w3.org/TR/html401/struct/tables.html) tables probably shouldn't be contained in tables. I know that the site you're testing is probably using tables as a means of layout, which is outdated and has always been an incorrect use of tables. One alternative to trying to fix something that isn't really broken is to tell your developers that using tables for layout is invalid HTML 4/XHTML and to update it to today's web standards. I don't really think that Watir is what's broken here, rather it is the web site that is using a broken method. I may be wrong about the site your testing using tables for layout, but if it is using tables for layout, then that's invalid HTML according to W3C.

Just some thoughts.

Nathan

--
Nathan Lane
Home, http://www.nathandelane.com
Blog, http://nathandelane.blogspot.com

[Nathan Lane]

This is the XHTML document: http://www.w3.org/TR/2004/WD-xhtml2-20040722/mod-tables.html

And attached are my two tests. "InvalidMarkup..." contains both the results of the W3C Validator test and a testtables.html containing inner tables, and "ValidMarkup..." contains both the results of the W3C Validator test and a testtables.html NOT containing inner tables.

Just for reference.

[andy sipe]

Just a comment on the standards vs reality thing:

While I understand and advocate the push to use correct and standard html, the reality is that it isn't very common to have 100% validated sites.   In fact, I cannot point to a single time where a project I worked on had every page 100% valid.   I'd be very surprised to find sites of any size and scope that meet this criteria while supporting the plethora of browsers versions available in the wild.

This coupled with the fact that many testers cannot control and/or dictate such things leaves us between an ideal that is somewhat less than obtainable.

Watir is a good tool because it does not always point back to standards as an out when something doesn't work.   In my opinion, if it can be rendered, watir should work with it.   Maybe this isn't always workable but in a lot of cases it works just fine.

-andy

[Nathan Lane]

The whole point is DON'T support the plethora of browsers available -- only the new ones that implement web standards, which is basically the law now for all major competitors - and if your using some niche browser then I'd hope you to expect problems with websites. Here are a couple of my references (not for the weak of heart ;))

http://veerle.duoh.com/blog/comments/starting_with_css_and_bug_fixing_tips/

http://veerle.duoh.com/blog/comments/does_flash_irks_me/

http://alistapart.com/articles/tohell/

In the time and age we live in, we don't have a need to support "bad browsers" or browsers that don't comply to web standards, we really don't. The vendors don't support them so why should we, really -- and this coming from somebody who is told that he "must support IE 6" because one whole division of our company will not upgrade even though they can and should. I know that it's a reality that people still write terrible web sites that try to support old browsers when they should just tell people to upgrade, but it shouldn't be required of a project used to test those websites. I reaffirm that since we have a work around, we should leave it at that and put it in the wiki somewhere for those people who must deal with poorly developed websites, especially those that use HTML and XHTML elements incorrectly. And that's why we have a wiki - we don't need to be sorry that our tool doesn't work with bad websites, and we can provide workarounds for people testing them. It just seems like a better idea than mucking up the project with stuff that isn't standard. That's my opinion, restated.

This coupled with the fact that many testers cannot control and/or dictate such things leaves us between an ideal that is somewhat less than obtainable.

Testers can dictate this information and have the ability to change what's going on and inform their developers that what they are doing is wrecking the site. That is one of the jobs of testers. Now depending on the company, and politics involved, there may need to be some intervention, but all companies irregardless of size, location, or power in the world need to realize that web standards help everybody and are built on the basis of being accessible, which apparently is written into law these days: http://alistapart.com/articles/politics/

Anyway, there you go, I'm a tester and a developer, but testing is my job and if I see something that isn't web standards on our site then I question it, write it up as a request/defect and then attend the meeting to help people understand it, deal with it, and get it fixed. If my company were using tables for layout, then I'd tell them that's the wrong way to go, but as it is they only use tables for tabular data.

Nathan

[Moochie]

Let's just use this and call it a day.

#Get actual row count when enbedded tables exists
def get_row_count_minus_enbedded_rows(table)
num_rows=0
begin
(1..table.row_count).each do |i|
num_rows = i
table[i]
end
rescue
end
return num_rows
end






On Oct 24, 11:27 am, "Nathan Lane" <nathamberl...@gmail.com> wrote:
> The whole point is DON'T support the plethora of browsers available -- only
> the new ones that implement web standards, which is basically the law now
> for all major competitors - and if your using some niche browser then I'd
> hope you to expect problems with websites. Here are a couple of my

> references (not for the weak of heart ;))http://veerle.duoh.com/blog/comments/starting_with_css_and_bug_fixing...http://veerle.duoh.com/blog/comments/does_flash_irks_me/http://alistapart.com/articles/tohell/

> > On Fri, Oct 24, 2008 at 11:42 AM, Nathan Lane <nathamberl...@gmail.com>wrote:
>
> >> This is the XHTML document:
> >>http://www.w3.org/TR/2004/WD-xhtml2-20040722/mod-tables.html
> >> And attached are my two tests. "InvalidMarkup..." contains both the
> >> results of the W3C Validator test and a testtables.html containing inner
> >> tables, and "ValidMarkup..." contains both the results of the W3C Validator
> >> test and a testtables.html NOT containing inner tables.
>
> >> Just for reference.
>

> >> On Fri, Oct 24, 2008 at 9:18 AM, Nathan Lane <nathamberl...@gmail.com>wrote:
>
> >>> I'm wondering whether or not this should be a defect really, because
> >>> according to this standards document about tables (
> >>>http://www.w3.org/TR/html401/struct/tables.html) tables probably
> >>> shouldn't be contained in tables. I know that the site you're testing is
> >>> probably using tables as a means of layout, which is outdated and has always
> >>> been an incorrect use of tables. One alternative to trying to fix something
> >>> that isn't really broken is to tell your developers that using tables for
> >>> layout is invalid HTML 4/XHTML and to update it to today's web standards. I
> >>> don't really think that Watir is what's broken here, rather it is the web
> >>> site that is using a broken method. I may be wrong about the site your
> >>> testing using tables for layout, but if it is using tables for layout, then
> >>> that's invalid HTML according to W3C.
> >>> Just some thoughts.
>
> >>> Nathan
>

> Blog,http://nathandelane.blogspot.com- Hide quoted text -

[Bret Pettichord]

I've just added new method
Watir::Table#row_count_excluding_nested_tables with this functionality.
It will be in Watir 1.6.

Bret