Repeating images

[ducu]

Following images should be specified on extralist.txt

fcw.com -> http://fcw.com/design/gig/fcw/2012/img/fcw-logo.png
github.com -> https://github.com/apple-touch-icon-144.png
www.forbes.com -> http://images.forbes.com/images/channels/headercoverstory*
a16z.com -> http://a16z.files.wordpress.com/2014/01/7cb56ea5114a9f0e92d53bf0e171d15d.png
www.gv.com -> http://img.gv.com/wp-content/uploads*

[jonw2000]

This github article is a problem if apple-touch-icon-144.png is excluded
https://github.com/blog/1862-introducing-a-simpler-faster-github-for-mac

Either the icon is not excluded (but this may ruin other github pages) or image resizing is used to resize https://cloud.githubusercontent.com/assets/22635/3517580/2399aa10-06f1-11e4-8671-0923504c594a.png.

[jonw2000]

Added these to extralist but the one it should pick is too big.

images.forbes.com/images/channels/headercoverstory*
i.forbesimg.com/assets/img/anon_avatar.png
blogs-images.forbes.com//files///-300x300.png
images.forbes.com/images/channels/headerfeature*_panel*_original-.

[jonw2000]

Forbes images - have included some regex filters in extralist but the main one is excluded due to image size. Perhaps consider image resampling/resizing using Pillow?

[jonw2000]

a16z.com - the map and office photo are excluded but the article image is not found as it's embedded in the following:

Perhaps a custom extractor can be used here.

[ducu]

Ok so here's one option:
Let summary get up to 3 valid (filtered) images and decide which one to choose based on some measurements such as get the bigger one in size. This is probably a good choice most of the times (e.g. www.forbes.com images, logos everywhere are smaller than the article image etc.).
This should be easy to do, I remember I was playing with this option at the time.

Second, we could have custom extraction handlers based on the link url as you said.
This would be useful if we want to extend summary to get video urls for example (e.g. have special handlers for youtube.com, vimeo.com and so on). This would be the most flexible option but also the most difficult to maintain.

Re image manipulation (resizing etc) we could try http://thumbor.org/ in the future, think I mentioned it before. But let's leave this for later..

[ducu]

Looks like a candidate for PhantomJS: googleandyourbusiness.blogspot.com