From ac4d114214d820b20e18518a2dbc809337e39043 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nicolas=20L=C5=93uillet?= Date: Fri, 6 Dec 2013 10:13:03 +0100 Subject: [add] new specific configuration files --- .../site_config/standard/smithsonianmag.com.txt | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) create mode 100644 inc/3rdparty/site_config/standard/smithsonianmag.com.txt (limited to 'inc/3rdparty/site_config/standard/smithsonianmag.com.txt') diff --git a/inc/3rdparty/site_config/standard/smithsonianmag.com.txt b/inc/3rdparty/site_config/standard/smithsonianmag.com.txt new file mode 100644 index 00000000..10a3f717 --- /dev/null +++ b/inc/3rdparty/site_config/standard/smithsonianmag.com.txt @@ -0,0 +1,20 @@ +# meta data +title://h1[@id = 'articleTitle'] +author:substring-after(//ul[@id = 'byLine']/li[1],'By ') +date:substring-before(substring-after(//ul[@id = 'byLine']/li[last()],','),',') +body://div[@id = 'article-body'] + +# full content +single_page_link://td/li[@class = 'article-singlepage']/a + +# caption clean up +wrap_in(i)://span[@class='articleImageCaptionwide'] +move_into (//span[@class='articleImageCaptionwide'])://div[@id = 'articleImage']/p + + +# clean up +strip://p[@id = 'articlePaginationWrapper'] +strip://ul[contains(@class, 'cat-breadcrumb')] +strip://div [@class= 'viewMorePhotos'] + +test_url: http://www.smithsonianmag.com/history-archaeology/The-Goddess-Goes-Home.html \ No newline at end of file -- cgit v1.2.3