2

Background: Reader Mode, as seen in Safari and other browsers, extracts the main content of article based web pages using sophisticated heuristics, and displays this with a very readable font.

All navigation, headers, footers, and other fluff is removed. The mode only works with "articles", ie. pages where there is a "main content" like a news article, scientific paper, etc.

The question: Is there an open source implementation of this for Terminals (ie. text-only)? Or alternatively, another way to accomplish the same thing?

Example: This article from The New York Times should output like so:

$ utility --reader-mode https://www.nytimes.com/2019/01/30/reader-center/polar-vortex-tips.html

SEND US YOUR IDEAS FOR WHAT TO DO DURING THE POLAR VORTEX. WE
WANT TO HEAR FROM YOU.

It’s so cold in much of the Midwest today that you could get
frostbite within five minutes once you step outside. If you’re
living through it indoors, give us your tips.

A commuter during an extremely light morning rush hour in Chicago
on Wednesday. Businesses and schools have closed as the city
copes with record low temperatures.

Across the Midwest, where wind chills were minus 51 in
Minneapolis and minus 45 in Chicago, the risks of going outside
on Wednesday were dire. So, many people simply didn’t bother,
while others took a chance to briefly experience the coldest
weather in a generation.

Whether you’re an adventurer or a hibernator, tell us your
recommendations for staying warm and busy. What are you cooking
or binge-watching? What board games are you playing? If you’re
venturing outside, what are you doing to stay safe? (Experts warn
that even a short time in the extreme cold can be very
dangerous.) How many layers of clothing are you wearing, and
which special hats and gloves are necessary? Send us your photos
and your stories.
  • 1
    determining the "main content" seems to me to be a tricky problem to solve – Jeff Schaller Jan 31 at 16:02
  • 1
    Yes. This is "solved" in best effort with various implementations of "Reader Mode" using heuristics. So it would have to be a text-only port of that, or something similar. google.com/search?q=reader+mode+source+code – forthrin Jan 31 at 16:23
0

The comment about "navigation content" is addressed by the -nolist option, e.g.,

lynx -nolist -dump www.google.com > file.txt

which shows no links, etc:

$ lynx -nolist -dump www.google.com > file.txt
$ cat file.txt 

   Search Images Maps Play YouTube News Gmail Drive More »
   Web History | Settings | Sign in

   Google

     _______________________________________________________
     Google Search  I'm Feeling Lucky                          Advanced search
                                                               Language tools

   Advertising Programs       Business  Solutions       +Google     About
   Google

                         © 2019 - Privacy - Terms

w3m gives something similar, without the option:

$ w3m -dump https://www.google.com
Search Images Maps Play YouTube News Gmail Drive More >>
Web History | Settings | Sign in

                                    Google

           [                                                         ] Advanced
                                                                       searchLanguage
                       [Google Search][I'm Feeling Lucky]              tools

           Advertising ProgramsBusiness Solutions+GoogleAbout Google

                          (C) 2019 - Privacy - Terms

links2 output looks much like w3m's (noting the missing space before About):

$ links2 -dump www.google.com                                          
   Search Images Maps Play YouTube News Gmail Drive More >>========(97,1) 31% ==
   Web History | Settings | Sign in                                             
                                     Google

    __________________________________________________________    Advanced       
              [ Google Search ] [ I'm Feeling Lucky ]             searchLanguage 
                                                                  tools          

           Advertising ProgramsBusiness Solutions+GoogleAbout Google

                           (c) 2019 - Privacy - Terms

$ links2 -dump www.google.com >file.txt 
$ cat file.txt 
   Search Images Maps Play YouTube News Gmail Drive More >>
   Web History | Settings | Sign in
                                     Google

    __________________________________________________________    Advanced       
              [ Google Search ] [ I'm Feeling Lucky ]             searchLanguage 
                                                                  tools          

           Advertising ProgramsBusiness Solutions+GoogleAbout Google

                           (c) 2019 - Privacy - Terms

(oddly enough, it also prints progress if the dump goes directly to the terminal—not a good feature) and elinks apparently only dumps the format with "navigation content" (ymmv).

From further comments, it turns out that OP is interested in something which could render the contents of a given division on the page. Comparing the sizes of the source and dump for that page gives some clues:

      Size    Buffer name          Contents
      ------- -------------------- ----------------------------------------------------------------------------------------
   0# 267624  [!lynx -source ht-1] !lynx -source https://www.nytimes.com/2019/01/30/reader-center/polar-vortex-tips.html
   1  5475    [!lynx -dump -nolis] !lynx -dump -nolist https://www.nytimes.com/2019/01/30/reader-center/polar-vortex-tips.html

shows that the dump is about 2% of the size of the source. Most of the page is non-informational, and the text-browsers show the information. But the division requested is in a two-line chunk that looks like this (only the beginning: the first line actually has 62265 characters):

<div id="app"><div class="css-v89234 e3w10z60"><div><div><div class="css-13lpfd6 e1nre7570"><header class="css-1bymuyk e1>
<script>window.__preloadedData = {"initialState":{"Article:QXJ0aWNsZTpueXQ6Ly9hcnRpY2xlLzBhODc0MTcxLWM0MjEtNWRjOS1hN2IzLW>

The first line holds the article text (plus a lot of markup), and offhand, looking at the second line, that's probably the script which the GUI browsers detect to show the article. None of the above-mentioned text-browsers has a feature for just showing a given <div>...</div>, or interpreting a script in that manner. These articles mention the absence of standard URI for reader mode in several GUI browsers:

  • Thanks for sharing! I've updated the question a bit to point out that the target for such use is article pages, where the objective is to extract the article itself. Have a look at the posting again and see if you can help. – forthrin Jan 31 at 22:03
-1

Does this satisfy your requirement? (From https://stackoverflow.com/questions/12422289/bash-command-to-convert-html-page-to-a-text-file )

lynx --dump www.google.com > file.txt
  • 1
    Nope. This dumps a ton of navigation links, eg. lynx -dump https://www.nytimes.com/2019/01/30/reader-center/polar-vortex-tips.html A solution should strip away all navigation and fluff, and leave ONLY the MAIN CONTENT, like Reader Mode in a browser does. – forthrin Jan 31 at 15:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.