Xpath to get information in next sibling tag using Scrapy

Question

I'm trying to get my hands on Scrapy and right now I try to extract information from an etymology website: http://www.etymonline.com Right now, I just want to get the words and their raw description. This is how a usual HTML code chunk is presented in etymonline:

<dt>
  <a href="/index.php?term=address&allowed_in_frame=0">address (n.)</a>
  <a href="http://dictionary.reference.com/search?q=address" class="dictionary" title="Look up address at Dictionary.com">
    <img src="graphics/dictionary.gif" width="16" height="16" alt="Look up address at Dictionary.com" title="Look up address at Dictionary.com"/>
  </a>
</dt>
<dd>
  1530s, "dutiful or courteous approach," from <a href="/index.php?term=address&allowed_in_frame=0" class="crossreference">address</a> (v.) and from French <span class="foreign">adresse</span>. Sense of "formal speech" is from 1751. Sense of "superscription of a letter" is from 1712 and led to the meaning "place of residence" (1888).
</dd>

The word is contained in the <dt>tag and the description in the next sibling, <dd> tag. To get the list of word on a page like http://www.etymonline.com/index.php?l=a&p=9&allowed_in_frame=0, it's possible to write word = sel.xpath('//dl/dt/a/text()').extract().

Then I tried to loop over this list of words and extract the relevant information using this line of code info = selInfo.xpath("//dl/dt[a='"+word[i]+"']/following-sibling::dd"). But it doesn't seem to work. Any ideas ?

Answer 1

To get to the <dd> after a <dt>, you can use the following-sibling axis, you're correct.

following-sibling::dd with select all dd elements after the context node. Therefore you need to restrict the XPath to only the first one, using a position predicate [1].

For each dt element you get out of //dl/dt, you select following-sibling::dd[1].

Here's a sample session using scrapy shell for the term "address":

$ scrapy shell "http://www.etymonline.com/index.php?allowed_in_frame=0&search=address&searchmode=none"
...
2014-11-26 10:34:53+0100 [default] DEBUG: Crawled (200) <GET http://www.etymonline.com/index.php?allowed_in_frame=0&search=address&searchmode=none> (referer: None)
[s] Available Scrapy objects:
[s]   crawler    <scrapy.crawler.Crawler object at 0x7f1396cc6950>
[s]   item       {}
[s]   request    <GET http://www.etymonline.com/index.php?allowed_in_frame=0&search=address&searchmode=none>
[s]   response   <200 http://www.etymonline.com/index.php?allowed_in_frame=0&search=address&searchmode=none>
[s]   settings   <scrapy.settings.Settings object at 0x7f1397399bd0>
[s]   spider     <Spider 'default' at 0x7f13966c05d0>
[s] Useful shortcuts:
[s]   shelp()           Shell help (print this help)
[s]   fetch(req_or_url) Fetch request (or URL) and update local objects
[s]   view(response)    View response in a browser

In [1]: for dt in response.xpath('//dl/dt'):
    print "Word:", dt.xpath('string(a)').extract()
    print "Definition:", dt.xpath('string(following-sibling::dd[1])').extract()
    print
   ...:     
Word: [u'address (n.)']
Definition: [u'1530s, "dutiful or courteous approach," from address (v.) and from French adresse. Sense of "formal speech" is from 1751. Sense of "superscription of a letter" is from 1712 and led to the meaning "place of residence" (1888).']

Word: [u'addressee (n.)']
Definition: [u'1810; see address (v.) + -ee.']

Word: [u'address (v.)']
Definition: [u'early 14c., "to guide or direct," from Old French adrecier "go straight toward; straighten, set right; point, direct" (13c.), from Vulgar Latin *addirectiare "make straight," from Latin ad "to" (see ad-) + *directiare, from Latin directus "straight, direct" (see direct (v.)). Late 14c. as "to set in order, repair, correct." Meaning "to write as a destination on a written message" is from mid-15c. Meaning "to direct spoken words (to someone)" is from late 15c. Related: Addressed; addressing.']

Word: [u'salutatorian (n.)']
Definition: [u'1841, American English, from salutatory "of the nature of a salutation," here in the specific sense "designating the welcoming address given at a college commencement" (1702) + -ian. The address was originally usually in Latin and given by the second-ranking graduating student.']

...

Word: [u'reverend (adj.)']
Definition: [u'early 15c., "worthy of respect," from Middle French reverend, from Latin reverendus "(he who is) to be respected," gerundive of revereri (see reverence). As a form of address for clergymen, it is attested from late 15c.; earlier reverent (late 14c. in this sense). Abbreviation Rev. is attested from 1721, earlier Revd. (1690s). Very Reverend is used of deans, Right Reverend of bishops, Most Reverend of archbishops.']

Word: [u'nun (n.)']
Definition: [u'Old English nunne "nun, vestal, pagan priestess, woman devoted to religious life under vows," from Late Latin nonna "nun, tutor," originally (along with masc. nonnus) a term of address to elderly persons, perhaps from children\'s speech, reminiscent of nana (compare Sanskrit nona, Persian nana "mother," Greek nanna "aunt," Serbo-Croatian nena "mother," Italian nonna, Welsh nain "grandmother;" see nanny).']


In [2]: 

Answer 2

The idea for your xpath to work is not to loop the extracted list, but within a parent node in xpath.

I don't have scrapy on my mac at the moment, but the technique here should similarly apply, like this:

# I use lxml for loose html string parsing
from lxml import html

s = '''<dt><a href="/index.php?term=address&allowed_in_frame=0">address (n.)</a> <a href="http://dictionary.reference.com/search?q=address" class="dictionary" title="Look up address at Dictionary.com"><img src="graphics/dictionary.gif" width="16" height="16" alt="Look up address at Dictionary.com" title="Look up address at Dictionary.com" /></a></dt>
<dd>1530s, "dutiful or courteous approach," from <a href="/index.php?term=address&allowed_in_frame=0" class="crossreference">address</a> (v.) and from French <span class="foreign">adresse</span>. Sense of "formal speech" is from 1751. Sense of "superscription of a letter" is from 1712 and led to the meaning "place of residence" (1888).</dd>'''

sel = html.fromstring(s)

# rather than extracting the words straight away, you loop from the parent xpath
for nodes in sel.xpath('//dt'):
    # then access a node to get the text
    print nodes.xpath('a/text()')
    # and go back to parent and search the dd node
    print nodes.xpath('../dd/text()')

# sample results
['address (n.)']
['1530s, "dutiful or courteous approach," from ', ' (v.) and from French ', '. Sense of "formal speech" is from 1751. Sense of "superscription of a letter" is from 1712 and led to the meaning "place of residence" (1888).']

Hope this helps.