From 9a96ce202194bb7ddc33f2fe61e8f25e8555f473 Mon Sep 17 00:00:00 2001 From: Tim Groeneveld Date: Tue, 10 Mar 2015 11:33:44 +1100 Subject: Add a bunch of Australian news sites. * au.businessinsider.com * au.news.yahoo.com * brokernews.com.au * choice.com.au * cnet.com.au * dailytelegraph.com.au * designbuildsource.com.au * gizmodo.com.au * heraldsun.com.au * itnews.com.au * marketingmag.com.au * moneymanagement.com.au * news.com.au * news.ninemsn.com.au * perthnow.com.au * smh.com.au * smh.drive.com.au * sunshinecoastdaily.com.au * theaustralian.com.au * watoday.com.au * weeklytimesnow.com.au * westernadvocate.com.au --- au.businessinsider.com.txt | 12 ++++++++++++ au.news.yahoo.com.txt | 4 ++++ brokernews.com.au.txt | 2 ++ choice.com.au.txt | 4 ++++ cnet.com.au.txt | 17 +++++++++++++++++ dailytelegraph.com.au.txt | 5 +++++ designbuildsource.com.au.txt | 2 ++ gizmodo.com.au.txt | 8 ++++++++ heraldsun.com.au.txt | 12 ++++++++++++ itnews.com.au.txt | 5 +++++ marketingmag.com.au.txt | 1 + moneymanagement.com.au.txt | 2 ++ news.com.au.txt | 3 +++ news.ninemsn.com.au.txt | 3 +++ perthnow.com.au.txt | 12 ++++++++++++ smh.com.au.txt | 14 ++++++++++++++ smh.drive.com.au.txt | 13 +++++++++++++ sunshinecoastdaily.com.au.txt | 10 ++++++++++ theaustralian.com.au.txt | 6 ++++++ watoday.com.au.txt | 7 +++++++ weeklytimesnow.com.au.txt | 4 ++++ westernadvocate.com.au.txt | 11 +++++++++++ 22 files changed, 157 insertions(+) create mode 100644 au.businessinsider.com.txt create mode 100644 au.news.yahoo.com.txt create mode 100644 brokernews.com.au.txt create mode 100644 choice.com.au.txt create mode 100644 cnet.com.au.txt create mode 100644 dailytelegraph.com.au.txt create mode 100644 designbuildsource.com.au.txt create mode 100644 gizmodo.com.au.txt create mode 100644 heraldsun.com.au.txt create mode 100644 itnews.com.au.txt create mode 100644 marketingmag.com.au.txt create mode 100644 moneymanagement.com.au.txt create mode 100644 news.com.au.txt create mode 100644 news.ninemsn.com.au.txt create mode 100644 perthnow.com.au.txt create mode 100644 smh.com.au.txt create mode 100644 smh.drive.com.au.txt create mode 100644 sunshinecoastdaily.com.au.txt create mode 100644 theaustralian.com.au.txt create mode 100644 watoday.com.au.txt create mode 100644 weeklytimesnow.com.au.txt create mode 100644 westernadvocate.com.au.txt diff --git a/au.businessinsider.com.txt b/au.businessinsider.com.txt new file mode 100644 index 0000000..46bcddf --- /dev/null +++ b/au.businessinsider.com.txt @@ -0,0 +1,12 @@ +title://div[@class="sl-layout-post"]/h1 +body: //div[@id='content_post'] +strip: //div[contains(@class, "post-sidebar")] +strip: //div[@id='related-links'] +strip: //img[@class='size_xlarge'] +author://div[@class="byline"]/a +date://div[@class="byline"]/span[@class="date"] +prune: no +tidy: no + + +test_url: http://www.businessinsider.com/microsoft-just-put-one-of-its-hardcore-technical-geniuses-on-xbox-2012-1 diff --git a/au.news.yahoo.com.txt b/au.news.yahoo.com.txt new file mode 100644 index 0000000..8e84cbb --- /dev/null +++ b/au.news.yahoo.com.txt @@ -0,0 +1,4 @@ +strip: //a[contains(text(), "RELATED:")] +author: //div[@class="info"]//span[@class="association printer-source"] +author: //div[@class="info"]//span[@class="stamp printer-date"] + diff --git a/brokernews.com.au.txt b/brokernews.com.au.txt new file mode 100644 index 0000000..814da38 --- /dev/null +++ b/brokernews.com.au.txt @@ -0,0 +1,2 @@ +author: //span[@itemprop="author"] +date: //span[@itemprop="datePublished"] diff --git a/choice.com.au.txt b/choice.com.au.txt new file mode 100644 index 0000000..0271475 --- /dev/null +++ b/choice.com.au.txt @@ -0,0 +1,4 @@ + +body: //div[@id='content']//div[@id='mainBlogContentWrapper']//*[self::p or self::img or self::ul] | //div[@class='mainArticleIntro')] + +date: //span[@class='date'] diff --git a/cnet.com.au.txt b/cnet.com.au.txt new file mode 100644 index 0000000..d5719d4 --- /dev/null +++ b/cnet.com.au.txt @@ -0,0 +1,17 @@ +title: //meta[@property="og:title"]/@content +body: //div[contains(@class, 'postBody')] +date: //div[@id='nameAndTime']/time +author: //div[@id='nameAndTime']/span[@class='author'] + +strip_id_or_class: image-credit +strip_id_or_class: noAutolink +strip_id_or_class: related +strip_id_or_class: cite + +prune: no +tidy: no + +# early end +replace_string(Download today's podcast): Download today's podcast + +test_url: http://www.cnet.com/8301-13952_1-57367607-81/the-404-981-where-the-world-is-a-vampire-podcast/ diff --git a/dailytelegraph.com.au.txt b/dailytelegraph.com.au.txt new file mode 100644 index 0000000..571e811 --- /dev/null +++ b/dailytelegraph.com.au.txt @@ -0,0 +1,5 @@ +title: //h1[@class="heading"] +author: //cite[@class='author'] +date: //li[contains(@class, 'date-and-time')] + + diff --git a/designbuildsource.com.au.txt b/designbuildsource.com.au.txt new file mode 100644 index 0000000..93d3507 --- /dev/null +++ b/designbuildsource.com.au.txt @@ -0,0 +1,2 @@ +date: substring-after(//p[@class='post_date'], 'on') + diff --git a/gizmodo.com.au.txt b/gizmodo.com.au.txt new file mode 100644 index 0000000..9dbfc15 --- /dev/null +++ b/gizmodo.com.au.txt @@ -0,0 +1,8 @@ +body: //div[@id='content_post' or @class="post-body" or contains(@class, 'illustration top')] +author: (//cite//span[@class="plus-icon"])[1] +date: //span[@class="date"] +date: //time + +prune: no + +test_url: http://gizmodo.com/5880147/kuhn-rikon-improves-their-spice-grinder-with-grade-school-science diff --git a/heraldsun.com.au.txt b/heraldsun.com.au.txt new file mode 100644 index 0000000..b0ce56c --- /dev/null +++ b/heraldsun.com.au.txt @@ -0,0 +1,12 @@ +#body: //div[@class='story-body'] +body: //div[contains(@class, 'story-body')] +title: //div[@class='story-headline']//h1 +author: //cite[contains(@class, 'author')] +date: //span[@class='datestamp'] + +strip_id_or_class: story-info +strip: //div[contains(@class, 'story-promo')] +strip: //div[contains(@class, 'story-related')] + +prune: no +tidy: no diff --git a/itnews.com.au.txt b/itnews.com.au.txt new file mode 100644 index 0000000..47cbb0f --- /dev/null +++ b/itnews.com.au.txt @@ -0,0 +1,5 @@ +title: //h1[@class='article-header'] +body: //div[@class='body-content'] +author: //span[@class='author-byline']/a[contains(@id, 'Author')] + +strip: //span[contains(@id, 'Article_SourceLabel')] diff --git a/marketingmag.com.au.txt b/marketingmag.com.au.txt new file mode 100644 index 0000000..910741f --- /dev/null +++ b/marketingmag.com.au.txt @@ -0,0 +1 @@ +strip: //h3[@class="related-posts"] diff --git a/moneymanagement.com.au.txt b/moneymanagement.com.au.txt new file mode 100644 index 0000000..9892f66 --- /dev/null +++ b/moneymanagement.com.au.txt @@ -0,0 +1,2 @@ +date: //span[@class="publishdate"]//time +author: //span[@class="byline"] diff --git a/news.com.au.txt b/news.com.au.txt new file mode 100644 index 0000000..57b89a5 --- /dev/null +++ b/news.com.au.txt @@ -0,0 +1,3 @@ +body: //div[@class='story-body'] +prune: no +tidy: no diff --git a/news.ninemsn.com.au.txt b/news.ninemsn.com.au.txt new file mode 100644 index 0000000..ddd6406 --- /dev/null +++ b/news.ninemsn.com.au.txt @@ -0,0 +1,3 @@ +strip: //a[@class="contact"] +strip: //div[@class="article-media video-item"] +date: //div[@class='display-date'] diff --git a/perthnow.com.au.txt b/perthnow.com.au.txt new file mode 100644 index 0000000..b0ce56c --- /dev/null +++ b/perthnow.com.au.txt @@ -0,0 +1,12 @@ +#body: //div[@class='story-body'] +body: //div[contains(@class, 'story-body')] +title: //div[@class='story-headline']//h1 +author: //cite[contains(@class, 'author')] +date: //span[@class='datestamp'] + +strip_id_or_class: story-info +strip: //div[contains(@class, 'story-promo')] +strip: //div[contains(@class, 'story-related')] + +prune: no +tidy: no diff --git a/smh.com.au.txt b/smh.com.au.txt new file mode 100644 index 0000000..f647f81 --- /dev/null +++ b/smh.com.au.txt @@ -0,0 +1,14 @@ +body: //div[@id='content'] +title: //h1[@class='cN-headingPage'] +author: //h3[@class='authorName'] +date: //dd[@class='updated dtstamp'] + +strip: //ul[@class='social sponsored cfix'] +strip: //div[contains(@class, 'hiddenVisually')] +strip: //dd[@class='updated dtstamp'] +strip: //h3[@class='authorName'] +strip: //ul[@class='social cfix'] +strip: //div[contains(@id, 'adspot')] + +strip: //div[contains(@class, 'overlayPlayCountdown')] +strip: //div[@class='fdVideoWof']//span[@class='gone'] diff --git a/smh.drive.com.au.txt b/smh.drive.com.au.txt new file mode 100644 index 0000000..463fd88 --- /dev/null +++ b/smh.drive.com.au.txt @@ -0,0 +1,13 @@ +body: //div[@id='content'] +title: //h1[@class='cN-headingPage'] +author: //h3[@class='authorName'] +date: //dd[@class='updated dtstamp'] + +strip: //ul[@class='social sponsored cfix'] +strip: //div[contains(@class, 'hiddenVisually')] +strip: //dd[@class='updated dtstamp'] +strip: //h3[@class='authorName'] +strip: //ul[@class='social cfix'] +strip: //div[contains(@id, 'adspot')] + +test_url: http://smh.drive.com.au/roads-and-traffic/driver-distraction-responsible-for-more-car-crashes-than-alcohol-20130503-2iyg0.html diff --git a/sunshinecoastdaily.com.au.txt b/sunshinecoastdaily.com.au.txt new file mode 100644 index 0000000..bf5e918 --- /dev/null +++ b/sunshinecoastdaily.com.au.txt @@ -0,0 +1,10 @@ +body: //section//article//p + +strip: //aside +strip: //div[@class='margin-top-15'] +strip: //p[@class='tags'] + +author: //span[@class='byline']//ul[@class='piped']//li[1] +date: //span[@class='byline']//ul[@class='piped']//li[2] + +parser: html5lib diff --git a/theaustralian.com.au.txt b/theaustralian.com.au.txt new file mode 100644 index 0000000..1245efc --- /dev/null +++ b/theaustralian.com.au.txt @@ -0,0 +1,6 @@ +body: //div[contains(@class, 'story-body')] +author: //cite[contains(@class, 'author')] +date: //span[@class='datestamp'] + +strip: //div[@class='story-info'] + diff --git a/watoday.com.au.txt b/watoday.com.au.txt new file mode 100644 index 0000000..4302ac5 --- /dev/null +++ b/watoday.com.au.txt @@ -0,0 +1,7 @@ +author: //h3[@class="authorName"] +date: //dd[@class='updated dtstamp']//time + +strip: //div[contains(@class, "adspot")] +strip: //noscript +strip: //p//small + diff --git a/weeklytimesnow.com.au.txt b/weeklytimesnow.com.au.txt new file mode 100644 index 0000000..a79871f --- /dev/null +++ b/weeklytimesnow.com.au.txt @@ -0,0 +1,4 @@ +body: //div[@class='main-col' or @class='article-image-wide'] +title: //h1[@class='article-title'] +author: substring-before(//span[@class='author'], "|") +date: //span[@class='date'] diff --git a/westernadvocate.com.au.txt b/westernadvocate.com.au.txt new file mode 100644 index 0000000..be89cb0 --- /dev/null +++ b/westernadvocate.com.au.txt @@ -0,0 +1,11 @@ +title: //header[contains(@class, "news-article-title")]//h1 +date: //div[@class="news-article-byline"]//time +author: //h2[@class="news-article-author"]//cite + +# Turns out that westernadvocate is doing funky things with the slide show images. :< +# body: //ul[@class="slides"]//img | //div[contains(@class, "news-article-body")] +body: //div[contains(@class, "news-article-body")] + +strip: //div[contains(@class, "flexslider")] + +test_url: http://:www.westernadvocate.com.au/story/1532050/roos-accept-ziebell-ban-commentators-do-not/ -- cgit v1.3.1