From: Mark Thomas <mark@...>
Date: 2009-07-24T10:20:13+09:00
Subject: Re: Problem with xpath in scrubyt.

On Jul 15, 12:32 am, likhon <lik...@gmail.com> wrote:

>     fetch "http://search.half.ebay.com/#{isbn}"
>     henewbooks "//html/body/table[2]/tbody/tr/td/table[4]/tbody/tr/td
> [2]/table/tbody/tr/td/table[2]/tbody" do
>         henb "//tr[@class = 'tr-border']" do

Positional XPaths like this are fragile. I don't recommend just taking
the path from Firebug or other DOM inspectors. What happens when
Half.com decides to put a notice in an additional table early in the
page? Your XPath will break.

Instead, think of how a human identifies the item. It's the table
immediately after the Brand New label, correct? A better XPath would
be:

//table[preceding-sibling::a[1][@name="itemlist_BRAND_NEW"]]//tr
[@class="tr-border"]

This does the same thing, but is much less susceptible to unrelated
page changes. If you do a lot of scraping, I recommend really getting
familiar with all that XPath has to offer. DOM inspectors are not
smart enough.

-- Mark.