<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by hartator on Medium]]></title>
        <description><![CDATA[Stories by hartator on Medium]]></description>
        <link>https://medium.com/@hartator?source=rss-d009ce9b7027------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*ZCgGfHA2U6wo4C7VtUjULw.png</url>
            <title>Stories by hartator on Medium</title>
            <link>https://medium.com/@hartator?source=rss-d009ce9b7027------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 06:22:48 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@hartator/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[A vision for Twitter: pushing free speech borders]]></title>
            <link>https://medium.com/@hartator/a-vision-for-twitter-pushing-free-speech-borders-26bb21f52ae4?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/26bb21f52ae4</guid>
            <category><![CDATA[liberty]]></category>
            <category><![CDATA[united-states]]></category>
            <category><![CDATA[libertarianism]]></category>
            <category><![CDATA[free-speech]]></category>
            <category><![CDATA[twitter]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Sun, 08 May 2022 22:24:48 GMT</pubDate>
            <atom:updated>2022-05-08T22:24:48.306Z</atom:updated>
            <content:encoded><![CDATA[<p>Making Twitter a free speech absolutist dream is a very welcome change.</p><p>However, I don’t think saying this is enough. We need to define what the perfect platform for free speech entails and then execute strictly on that vision.</p><p>I believe Twitter has now the unique opportunity to rise to be the champion of free speech everywhere in the world and push the borders of what should be free speech. This opportunity should not be lost and be thoroughly cherished.</p><p>Twitter should become a legal company first.</p><p>Twitter should fight for its users’ ability to post anything. While having the best freedom of speech laws and jurisprudence in the world, the U.S. is far from perfect. The Twitter legal team should fight in all courts to extend and to protect free speech. Its CEO must be ready to put his own freedom into play to safeguard Twitter users’ ability to post. We should be at the frontlines of the fight for free speech, not way behind them in low risk zones.</p><p>Twitter should accept closing down foreign offices when keeping them open would mean compromising on free speech. e.g., EU Rights to be forgotten, which has been unfortunately but not surprisingly used mainly by corrupt officials for automated censorship of their own convictions, is a very clear aberration against free speech. We should not abide by that. After office closure, it will be up to the foreign governments to decide if they want to follow China and Russia’s lead and block the full twitter.com domain name. Or accept actual exchanges of ideas.</p><p>Twitter finally should not lose momentum in U.S. internal political fightings. Free speech should unite all Americans. No matter political affiliations, religions, origins, or other convictions. Twitter should draw the line there and terminate people who, unfortunately, don’t share free speech as a core value of their own political framework.</p><p>I am hopeful that very good days are ahead of us.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=26bb21f52ae4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How a routine gem update ended up creating $73k worth of subscriptions]]></title>
            <link>https://medium.com/serpapi/how-a-routine-gem-update-ended-up-creating-73k-worth-of-subscriptions-8f4250ee6518?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/8f4250ee6518</guid>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[stripe]]></category>
            <category><![CDATA[ruby]]></category>
            <category><![CDATA[tech]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Sat, 08 Jan 2022 01:14:13 GMT</pubDate>
            <atom:updated>2022-01-08T01:14:13.349Z</atom:updated>
            <content:encoded><![CDATA[<p>On November 5, 2021 (a Friday of course), we’ve deployed innocent-looking gem updates. Minor versions of Ruby on Rails, Ruby Sentry client, Ruby Slack client, http libraries, Puma, Devise, OmniAuth Ruby client, Mongoid, and a few test gems.</p><p>However, something went very wrong.</p><p>We saw odd Stripe errors on Airbrake. Then, on our Stripe account, we saw this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NHPWM-QEP7KrwKOj.png" /></figure><p>In less than 1h, we’ve created 474 new subscriptions for a total of $73,271.36.</p><p>We weren’t expecting that much business.</p><p>Our app for some reason was creating new subscriptions from old accounts that were canceled or disabled a long time ago. We had no idea why but we decided to react proactively as fast as we can as the issue was still very ongoing.</p><p>We immediately rolled back the gem updates to their safe older versions, disabled our Stripe API keys to lock our own application out, and refunded everyone impacted:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yOc9nKh7XKSvySvh.png" /></figure><p>After refunding everyone, we manually double-checked the billing state of each account one by one and sent emails to apologize to each customer one by one; all 475 of them.</p><p>We still didn’t know why.</p><p>Most of the updates seemed inoffensive. And after deeper inspection, it was indeed minor gem updates on several small things. We worked on several hypotheses: caching issues, odd race conditions, or some thread safety issues as Puma changed their threading model in that update.</p><p>However, we’ve finally found it. One method of our billing logic particularly seemed a code smell (Obligatory disclaimer: I personally wrote this.):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*z9W4C7-Icp2srlFG.png" /></figure><p>This is part of a feature that seems good on paper: We offer to automatically renew your plan if you run out of searches on our service.</p><p>The goal of renew_early_protected is to avoid the case of someone submitting multiple searches at the exact same time, subsequently getting their subscription renewed then their credit card charged several times.</p><p>Notice the or() method at line 47.</p><p>In Mongoid (The MongoDB driver for Ruby) 7.0.8 , or() meant filter documents that contain any of the argument conditions. In our case, filter users that have a specific id AND that have either never had a renewing_early_lock or their renewing_early_lock is less than 1 hour ago.</p><p>Here&#39;s what it got computed to for a specific user:</p><pre>computed_selector_in_mongoid_7_0_8 = {<br>  &quot;_id&quot; =&gt; BSON::ObjectId(&#39;59af54094-----------64&#39;), &quot;$or&quot; =&gt; [{<br>    &quot;renewing_early_lock_at&quot; =&gt; {<br>      &quot;$lte&quot; =&gt; 2022-01-07 20:14:53.44744 UTC<br>    }<br>  }, {<br>    &quot;renewing_early_lock_at&quot; =&gt; nil<br>  }]<br>}</pre><p>In Mongoid 7.3.3 , or() now means filter documents that contain any of the argument conditions OR any of previous method conditions! In our case filter users that have either a specific id OR that have never had a renewing_early_lock OR their renewing_early_lock is less than 1 hour ago.</p><p>Here’s what it looks like for the same user:</p><pre>computed_selector_in_mongoid_7_3_3 = {<br>  &quot;$or&quot; =&gt; [{<br>    &quot;_id&quot; =&gt; BSON::ObjectId(&#39;59af54094-----------64&#39;)<br>  }, {<br>    &quot;renewing_early_lock_at&quot; =&gt; {<br>      &quot;$lte&quot; =&gt; 2022-01-07 19:00:09.571034 UTC<br>    }<br>  }, {<br>    &quot;renewing_early_lock_at&quot; =&gt; nil<br>  }]<br>}</pre><p>Notice that the user ID selector got moved into an optional or()!</p><p>This ultimately means we were charging random customers new subscriptions instead of the actual customer who wanted to renew early:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*s6rGSNUHyclRM5Gs.png" /></figure><p>Which is obviously awful.</p><p>We are deeply sorry if you were impacted by this issue. This is not acceptable.</p><p>Even if that Mongoid shouldn’t have changed existing methods behaviors between minor versions, my implementation was a true code smell as it was unclear what it did. This code should never have been deployed to production. But also and more importantly, the feature itself was a bad idea. Using our API to scrape search engine results shouldn’t trigger a renewal of a credit card subscription in the first place. And we’ll be removing that feature as soon as possible.</p><p>Sorry again to all of our users.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8f4250ee6518" width="1" height="1" alt=""><hr><p><a href="https://medium.com/serpapi/how-a-routine-gem-update-ended-up-creating-73k-worth-of-subscriptions-8f4250ee6518">How a routine gem update ended up creating $73k worth of subscriptions</a> was originally published in <a href="https://medium.com/serpapi">SerpApi</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Birth of SerpApi]]></title>
            <link>https://medium.com/serpapi/the-birth-of-serpapi-464a9c9c5689?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/464a9c9c5689</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Tue, 27 Jul 2021 21:45:24 GMT</pubDate>
            <atom:updated>2021-07-27T21:45:24.139Z</atom:updated>
            <content:encoded><![CDATA[<p>It feels like it was a long time ago. I was experimenting with porting machine learning models to iOS in order to make a couple of picture analysing apps just for fun. One was identifying food and trying to guess calories (I know it’s almost impossible to be accurate!) and another was trying to guess dog breeds. Here’s a snapshot of what the dog identifier one looked like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*z5N4HcgFDfIoYVhQ.png" /><figcaption>Dog Breed Identifier App on iOS</figcaption></figure><p>Unfortunately back in 2017, iOS didn’t have a proper machine learning framework. Core ML wasn’t released yet and the only way to have a machine learning framework running on iOS is to have C/Objective C bindings pointing to actual Caffe C code. Most people ended just using remote web API to do their machine learning logic. However I wanted something local to the iPhone. Without the need of an Internet connection. So I needed to make iPhones execute actual machine learning logic. At the time, it was challenging: iPhones’ RAM were limited, no GPU or no dedicated chip acceleration were available, using only CPU was painfully slow, and compiling/porting C code from machine learning framework like Tensorflow or Caffe to iOS wasn’t straightforward. Oddly, all of this wasn’t the most difficult part of this project.</p><p>Collecting images from Google Images was.It was hard for multiple reasons: You are getting blocked from scraping image search results super fast by Google, you needed then to buy proxies, find a way to rotate proxies in a cost effective way, buy some CAPTCHA solving services as you still get blocked and proxies are expensive, run a full browser like PhantomJS or headless Chrome to load the pages, glue all of this together somehow, and finally parse the HTML to extract the images you wanted from a layout that has variations and can change daily. It was a pain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*MHONzSEuxex8AUcK.png" /><figcaption>Sample of greasy food images extracted for calorie guessing</figcaption></figure><p>In these projects, 80% of my time ended up being spent on scraping and parsing Google Images. And maybe only 20% on actual machine learning model training, UI design of the actual apps, and iOS programming.</p><p>I thought there must be a better way! However there was not. You couldn’t pay a service that does that for you. There was no simple scraping API for Google Images or even regular Google searches. That was upsetting. I decided to work on solving this and started building prototypes. And SerpApi was born! Here’s a screenshot of one of the very first versions:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XEH1ulPxREseNTJi.png" /><figcaption>Screenshot of an early version of SerpApi</figcaption></figure><p>We’ve grown a lot since then. However it’s interesting to know that collecting text and images from Google search results in order to train machine learning models is actually still one of the main use cases of SerpApi!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=464a9c9c5689" width="1" height="1" alt=""><hr><p><a href="https://medium.com/serpapi/the-birth-of-serpapi-464a9c9c5689">The Birth of SerpApi</a> was originally published in <a href="https://medium.com/serpapi">SerpApi</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ruby Schedulers: Whenever vs Sidekiq Cron vs Sidekiq Scheduler]]></title>
            <link>https://medium.com/serpapi/ruby-schedulers-whenever-vs-sidekiq-cron-vs-sidekiq-scheduler-b229d7ca5256?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/b229d7ca5256</guid>
            <category><![CDATA[cron]]></category>
            <category><![CDATA[ruby]]></category>
            <category><![CDATA[sidekiq]]></category>
            <category><![CDATA[scheduling]]></category>
            <category><![CDATA[rails]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Mon, 03 May 2021 05:28:10 GMT</pubDate>
            <atom:updated>2021-05-03T23:50:10.731Z</atom:updated>
            <content:encoded><![CDATA[<p>We’ve been fully on board the fact that we’re a Web API at <a href="https://serpapi.com/">SerpApi</a>. And have been using as much as possible HTTP requests to make things work externally but also internally!</p><p>HTTP requests even for internal tasks are indeed great. They are easily debuggable as they are part of our usual workflow. Web servers are already there, carefully monitored, and load balanceable. All programming languages have numerous and super solid libraries for them. Toolings (❤️ cURL) are fabulous. And they allow us to not treat external API requests as second class citizens.</p><p>However sometimes you don’t have the choice but to use a good old’ cron jobs. eg, Backups, credit renewals, and other time constrained tasks. We will study here 3 Ruby solutions to schedule tasks: Whenever, sidekiq-cron and sidekiq-scheduler. Whenever gem is Ruby wrapper around classic crontabs jobs. Basically giving you a nicer syntax. Sidekiq-cron and sidekiq-scheduler use a more novel approach and seems to abuse Redis in the background to do the work of scheduling things.</p><p>To compare these, we will focus mostly on ease of setting up new tasks, ease of monitoring, and ease of access to logs. Ease of monitoring and logging are particularly important as it’s usually where things bit you in the ass when you set things and they fail silently. Looking at you old backup code that didn’t run for months.</p><h4>Whenever gem</h4><p><a href="https://github.com/javan/whenever">Repository</a></p><p>Config config/schedule.rb is pretty straightforward:</p><pre>set :output, &quot;log/cron_log.log&quot;<br>#<br>every 1.minute do<br>  rake &quot;users:renew_credits&quot;<br>end</pre><p>Then:</p><pre>whenever --update-crontab</pre><p>And that’s it!</p><p>Running whenever command also takes cares of updating and removing existings crons generated by it in addition of adding new cron jobs. And it’s working as expected:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PJf-74c3LOMLJfrWzKn6jw.png" /></figure><p>Logging is a not as straightforward. You’ll have to rely on an external errors monitoring tools like Sentry or Airbrake. Or you’ll have to tail the log/crong_log.log file to get an idea what went wrong:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FLhdcXYNZm-NktYcvtSaUw.png" /></figure><p>The pros of this solution are we don’t have to maintain sidekiq workers or a Redis database but it’s lacking an UI and a clear view if cron jobs have been running or not. Let’s check the sidekiq based solutions now!</p><h4>Sidekiq-Cron gem</h4><p><a href="https://github.com/ondrejbartas/sidekiq-cron">Repository</a></p><p>Scheduling a cron job is also kind of straightforward:</p><pre>class Users::RenewCreditsWorker<br>  include Sidekiq::Worker</pre><pre>def perform(*args)<br>    # Mock a simple job to try out different Ruby on Rails scheduling options<br>    # (Obviously to be run monthly in production not every minute)<br>    User.where(email: &#39;<a href="/service/mailto:test@serpapi.com">test@serpapi.com</a>&#39;).first.renew_montly_credit<br>  end<br>end</pre><pre>Sidekiq::Cron::Job.create(name: &#39;Renew user credit worker - every 1min&#39;, cron: &#39;1 * * * *&#39;, class: &#39;Users::RenewCreditsWorker&#39;)</pre><p>Although there is 4–5 different ways to configure this and it’s not super clear if this one is a preferred option or not. Another con: It’s not super clear what happen when “Sidekiq::Cron::Job.create” is run each time the Rails app is restarting. And actually, it’s not loading this file and adding the cron job at all even after restarting the Rails server and sidekiq several time:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-9hHsVwsmRQnI6Q-ptsTdw.png" /></figure><p>Which is a little unexpected. Just calling “Users::RenewCreditsWorker” is enough to load the worker and create the cron job.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*muj2ryiH-vLbALct8df2cA.png" /></figure><p>As you see, it also triggers a warning about “Redis#exists(key)” usage:</p><pre>`Redis#exists(key)` will return an Integer in redis-rb 4.3. `exists?` returns a boolean, you should use it instead. To opt-in to the new behavior now you can set Redis.exists_returns_integer =  true. To disable this message and keep the current (boolean) behaviour of &#39;exists&#39; you can set `Redis.exists_returns_integer = false`, but this option will be removed in 5.0. (/Users/.rbenv/versions/2.7.2/lib/ruby/gems/2.7.0/gems/sidekiq-cron-1.2.0/lib/sidekiq/cron/job.rb:464:in `block in save&#39;)</pre><p>Not sure reassuring on the quality of the gem but it does add it to the cron tab now though:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Zh2ZW56jt4AXr7SeaT0_UQ.png" /></figure><p>However puma does now try to re-create the cron job itself as it’s loading the same “Users::RenewCreditsWorker” class:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5OL0sriXZ-slp-78pa0ScQ.png" /></figure><p>Which is not okay. Switching to one of the alternative ways of configuring schedules does seem to work, but you now need an initializer that contains:</p><pre>schedule_file = &quot;config/schedule.yml&quot;</pre><pre>if File.exist?(schedule_file) &amp;&amp; Sidekiq.server?<br>  Sidekiq::Cron::Job.load_from_hash YAML.load_file(schedule_file)<br>end</pre><p>And a schedule.yml file that contains:</p><pre>users_renew_credits:<br>  cron: &quot;* * * * *&quot;<br>  class: &quot;Users::RenewCreditsWorker&quot;</pre><p>And it does get scheduled without any other trickery:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PX1M-a7Q1MPz0xRgAECDtg.png" /></figure><p>It’s cool to be able to see when was the last job scheduled, to disable/enable the cron job, or to force the cron job to run by just a click. You’ll also have one dedicated page per cron job:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w3xS1NACNq5umiTxVQfcmg.png" /></figure><p>And — most importantly — it does work now:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ze66oPLSTaQ3oD0MPfEoXQ.png" /></figure><p>Notice that the scheduling seems be less on the minute than system cron. It might be due to scheduling with sidekiq-cron is run on a set polling intervals. 30s I believe. Which is not a super big deal if it does work reliability as anyway most cron jobs are meant to be run hourly or daily. However more annoying is the non-straightforward way of configuring it — do we really need several ways of doing this when default already seems to be lacking good instructions ? — and the random warnings we’re getting hint the gem hasn’t been updated in a little bit.</p><p>Let’s give shot to sidekiq-scheduler.</p><h4>Sidekiq-Scheduler gem</h4><p><a href="https://github.com/moove-it/sidekiq-scheduler">Repository</a></p><p>Installation and configuration is very straightforward. Their Readme is way better. Given a file defining a worker:</p><pre>require &#39;sidekiq-scheduler&#39;</pre><pre>class Users::RenewCreditsWorker<br>  include Sidekiq::Worker<br>  sidekiq_options retry: false</pre><pre>def perform(*args)<br>    # Mock a simple job to try out different Ruby on Rails scheduling options<br>    # (Obviously to be run monthly in production not every minute)<br>    User.where(email: &#39;<a href="/service/mailto:test@serpapi.com">test@serpapi.com</a>&#39;).first.renew_montly_credit<br>  end<br>end</pre><p>And sidekiq.yml defining a schedule:</p><pre>:schedule:<br>  users_renew_credits:<br>      every: 1m<br>      class: Users::RenewCreditsWorker</pre><p>It does autocreate the recurring job at sidekiq startup without any other tickering needed:</p><pre>2021-05-03T04:38:17.899Z pid=2427 tid=ihv INFO: Scheduling users_renew_credits {&quot;every&quot;=&gt;&quot;1m&quot;, &quot;class&quot;=&gt;&quot;Users::RenewCreditsWorker&quot;, &quot;queue&quot;=&gt;&quot;default&quot;}</pre><p>Note that we have same warning as sidekiq-cron bout “Redis#exists(key)” usage:</p><pre>`Redis#exists(key)` will return an Integer in redis-rb 4.3. `exists?` returns a boolean, you should use it instead. To opt-in to the new behavior now you can set Redis.exists_returns_integer =  true. To disable this message and keep the current (boolean) behaviour of &#39;exists&#39; you can set `Redis.exists_returns_integer = false`, but this option will be removed in 5.0. (/Users/ioxtrem/.rbenv/versions/2.7.2/lib/ruby/gems/2.7.0/gems/sidekiq-scheduler-3.0.1/lib/sidekiq-scheduler/redis_manager.rb:99:in `block in schedule_exist?&#39;)</pre><p>It seems both gems were using “Sidekiq.redis { |r| r.exists(:schedules) }”-like syntax that got deprecated for “Sidekiq.redis { |r| r.exists?(:schedules) }”: <a href="https://github.com/resque/redis-namespace/pull/171">https://github.com/resque/redis-namespace/pull/171</a></p><p>Sidekiq-scheduler does work right away though:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ugrdfw3mGcy_8tpW6ZKkRw.png" /></figure><p>Notice also the scheduling seems to be more regular than Sidekiq-cron. It might be due to the default polling being way more often. 1s for Sidekiq-scheduler instead 30s of Sidekiq-cron.</p><p>The added UI by sidekiq-scheduler is nice but is very minimal:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*T0IiQkMgaJGP6EmQbI9r3g.png" /></figure><p>Sidedekiq-cron UI is indeed better. It contains more details and more actions in the overview page but also separated pages for each recurring job with even more details and more actions. Which matters more than a well written Readme.</p><p>In conclusion, all 3 of these gems do the job of adding recurrence to your Ruby or Ruby on Rails application. Whenever uses a classic approach of relying on system cron. It does an excellent job wrapping the syntax of system cron making it Ruby-nice, globbing Rails commands in it making sure they work, and making the adding and removal of recurral logic a breeze. Sidekiq-cron and Sidekiq-scheduler both relies on Sidekiq. It’s a heavy dependency as you’ll need to run Sidekiq workers, Redis and add Sidekiq logic to your application. Sidekiq-cron Readme does a poor job getting you setup — default config just doesn’t work — and default polling 30s is also pretty not aggressive compared to Sidekiq-Scheduler work in both areas. However Sidekiq-cron does have the best UI. It has more details and is actually fun to play with. And as we’ve defined earlier, toolings around monitoring, logging, and quick actions are usually what should matter for long term recurring logic central to your business.</p><p>[Edit 05/03/2021] Following this article, Mike, creator of Sidekiq, generously offered us a temporary Enterprise license for Sidekiq to try out their official solution to this. Thank you!</p><h4>Sidekiq Enterprise Periodic Jobs (Official Sidekiq solution)</h4><p><a href="https://github.com/mperham/sidekiq/wiki/Ent-Periodic-Jobs">Documentation page</a></p><p>After Sidekiq Enterprise installation, config is super simple. Directly inside initializers/sidekiq.rb file:</p><pre>Sidekiq.configure_server do |config|<br>  config.periodic do |periodic|<br>    periodic.register(&quot;* * * * *&quot;, &quot;Users::RenewCreditsWorker&quot;)<br>  end<br>end</pre><p>And it immediately works:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ob-uRvqcUqYlNg_2blW3MA.png" /></figure><p>Notice the multiple repetition of “17:59:27” and “18:13:28” jobs. It was probably from my MacBook going to sleep or going idle. I would have a preference for Sidekiq Enterprise Periodic Jobs to ignore and just skip jobs that haven’t be run on time for whatever reason. The previous 3 solutions behaved like this when my MacBook went to sleep and it seemed more reasonable. Indeed you wouldn’t want to run backups 10x at the same time to catch up or reset customer credit 10x. In these scenarii, It’s not useful and might mess things up even further. Notice though this solution seems to be most precise when it does run — Cron jobs like this anyway shouldn’t be run on laptop that can go to sleep to be fair! — with runs happening exactly on the second.</p><p>UI is also nice and look like Sidekiq-cron with an overview page:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ReDuVtWgCJi6TLI_afvsaA.png" /></figure><p>And a separated page for each periodic job:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dJYaQ8YbHuYWEH7t0bhJIg.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b229d7ca5256" width="1" height="1" alt=""><hr><p><a href="https://medium.com/serpapi/ruby-schedulers-whenever-vs-sidekiq-cron-vs-sidekiq-scheduler-b229d7ca5256">Ruby Schedulers: Whenever vs Sidekiq Cron vs Sidekiq Scheduler</a> was originally published in <a href="https://medium.com/serpapi">SerpApi</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[SerpApi vs Zenserp vs Serpstack vs SerpWow vs ScaleSerp Benchmarks]]></title>
            <link>https://medium.com/@hartator/serpapi-vs-zenserp-vs-serpstack-vs-serpwow-vs-scaleserp-benchmarks-6ef5fd467f7d?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ef5fd467f7d</guid>
            <category><![CDATA[big-data]]></category>
            <category><![CDATA[benchmark]]></category>
            <category><![CDATA[performance]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[seo]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Mon, 27 Jul 2020 21:54:36 GMT</pubDate>
            <atom:updated>2020-07-27T21:54:36.938Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Disclaimer: I am the CEO and founder of SerpApi, LLC. However I’ll aim to be as fair and unbiased as possible in running these benchmarks.</em></p><p>We’ll be comparing SerpApi [1] vs Zenserp [2] vs Serpstack [3] vs SerpWow [4][6] vs ScaleSerp [5][6] in 3 categories: Response times, service availabilities, and quantity of data extracted to JSON.</p><h4>Response times</h4><figure><img alt="Average response times: SerpApi vs Zenserp vs Serpstack vs SerpWow vs ScaleSerp" src="https://cdn-images-1.medium.com/max/1024/1*KsrsUir7tcXZj51hDns4fQ.png" /></figure><p>We get the average response times by running a batch of searches against each service. We use the exact same queries on each service on regular Google searches. Fast and consistent response times are particularly important if your application depends on real-time performance.</p><h4>Service availabilities</h4><figure><img alt="Service availabilities: SerpApi vs Zenserp vs Serpstack vs SerpWow vs ScaleSerp" src="https://cdn-images-1.medium.com/max/1024/1*DXgCHbOzwgpgsG7gJkjsGA.png" /><figcaption>Services Availabilities (green: no downtimes, red: at least one period of downtime, black: unknow)</figcaption></figure><p>We get measure service availabilities using a third party service named SimpleOps. We did the measurement during a 1-week period and any downtime is reported here.</p><h4>Quantity of data extracted</h4><figure><img alt="Average number of fields extracted to JSON: SerpApi vs Zenserp vs Serpstack vs SerpWow vs ScaleSerp" src="https://cdn-images-1.medium.com/max/1024/1*-PFU3WfE7iCQAlJoM_mHdw.png" /></figure><p>We get the average number of fields extracted to JSON by running a batch of the exact same queries against each service. This give us an idea how much data each service is capable to parse from raw HTMLs, make sense of it, and export it to JSON.</p><p>Notes:<br>[1] https://serpapi.com/<br>[2] https://zenserp.com/<br>[3] https://serpstack.com/<br>[4] https://serpwow.com/<br>[5] https://scaleserp.com/<br>[6] SerpWow and ScaleSerp are the same company, hence the similar results.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ef5fd467f7d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Estimation of excess suicides due to covid-19 lockdowns]]></title>
            <link>https://medium.com/@hartator/estimation-of-excess-suicides-due-to-covid-19-lockdowns-947507927c4b?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/947507927c4b</guid>
            <category><![CDATA[coronavirus]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[estimations]]></category>
            <category><![CDATA[suicide]]></category>
            <category><![CDATA[mathematics]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Thu, 30 Apr 2020 04:04:24 GMT</pubDate>
            <atom:updated>2020-05-17T18:10:30.269Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*plnddEOU9FBBvO9DCdAo3w.png" /></figure><p>Plenty of work have been done trying to estimate the number of deaths from covid-19 in different scenarii. However little has been done to estimate the number of suicides due to the lockdowns.</p><p>Earlier covid-19 death models may have erred too far on the side of caution. On March 16th, the Imperial College published a paper estimating the US deaths due to covid-19 to be 2.2M. [1] This study and others have triggered a series of lockdowns across the nation. As of today, April 29th, CDC forecasts “only” 72k deaths due to covid-19. [2] As 47k people die from committing suicide every year in the US [3], the rise of number of suicides due to covid-19 lockdowns should become a major concern.</p><p>Covid-19 lockdowns can and will trigger more suicides for various reasons but we will focus on the two main ones:<br>- Unemployment due to the covid-19 recession<br>- Isolation due to the covid-19 lockdowns</p><p>During the Great Depression, suicide rates increased by 22.8%. [4] Indeed, being unemployed has been found to increase the odds of committing suicide by 246% to 263% varying on gender and age. [5] Covid-19 job losses are expected to reach 47M people. [6] This alone will translate to an additional 9.6k to 10.7k suicides just in 2020. [10]</p><p>Moreover social isolation is as deadly as unemployment. Being strictly isolated raise all causes of mortality by 60% to 134%. [7] With 85% of the US population under stay-at-home order [8], depending if the lockdowns are 3 months long or up to next year, and 36.48% of the population living alone in 2019 [9], this will translate to another 2.2k to 20.1k additional suicides in 2020. [10]</p><p>In total in 2020, we should expect between 11.9k to 30.8k additional suicides just due to covid-19 lockdowns in the United States.</p><p>References:<br>[1] <a href="https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf">https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf</a><br>[2] <a href="https://covid19.healthdata.org/united-states-of-america">https://covid19.healthdata.org/united-states-of-america</a><br>[3] <a href="https://www.cdc.gov/nchs/data/databriefs/db355_tables-508.pdf">https://www.cdc.gov/nchs/data/databriefs/db355_tables-508.pdf</a><br>[4] <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3093269/">https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3093269/</a><br>[5] <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1732539/pdf/v057p00594.pdf">https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1732539/</a><br>[6] <a href="https://www.cnbc.com/2020/03/30/coronavirus-job-losses-could-total-47-million-unemployment-rate-of-32percent-fed-says.html">https://www.cnbc.com/2020/03/30/coronavirus-job-losses-could-total-47-million-unemployment-rate-of-32percent-fed-says.html</a><br>[7] <a href="https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/aje/188/1/10.1093_aje_kwy231/1/kwy231_alcaraz_web_material_final.pdf?Expires=1589189099&amp;Signature=SXg0HztTCzfZfwivy8SdNgTZa9YwHrEWTW8QnMYl83lEWyCD2YEA5Ld1hFZNwmFOnqkzkDKLU8baZesyFnX7rFO98FRoTNWiirOK~WSQRiMiejs7XIvD6rZk3SgQ978Er~-sOyY4eJ1XnpucgvyC8h4mNTQyVAi0e5FNjuCRBSd5f8WO~kQgLJxheR1fEC9KkAG4vVGAda2TuUn2PMHdxoNt~LTAyGxySto2WtbOoaX4BeDf1RnYAauYwbKX9Uq14RmKi-ar06qCRhTGAggv78rU4LOQXnyJHFzgJ9DG3glgb0MbnexhNfkIIt7IgGKZpB3NMgbWwO6~zGGdMFeZzQ__&amp;Key-Pair-Id=APKAIE5G5CRDK6RD3PGA">https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/aje/188/1/10.1093_aje_kwy231/1/kwy231_alcaraz_web_material_final.pdf?Expires=1589189099&amp;Signature=SXg0HztTCzfZfwivy8SdNgTZa9YwHrEWTW8QnMYl83lEWyCD2YEA5Ld1hFZNwmFOnqkzkDKLU8baZesyFnX7rFO98FRoTNWiirOK~WSQRiMiejs7XIvD6rZk3SgQ978Er~-sOyY4eJ1XnpucgvyC8h4mNTQyVAi0e5FNjuCRBSd5f8WO~kQgLJxheR1fEC9KkAG4vVGAda2TuUn2PMHdxoNt~LTAyGxySto2WtbOoaX4BeDf1RnYAauYwbKX9Uq14RmKi-ar06qCRhTGAggv78rU4LOQXnyJHFzgJ9DG3glgb0MbnexhNfkIIt7IgGKZpB3NMgbWwO6~zGGdMFeZzQ__&amp;Key-Pair-Id=APKAIE5G5CRDK6RD3PGA</a><br>[8] <a href="https://www.businessinsider.com/us-map-stay-at-home-orders-lockdowns-2020-3">https://www.businessinsider.com/us-map-stay-at-home-orders-lockdowns-2020-3</a><br>[9] <a href="https://www.statista.com/statistics/242022/number-of-single-person-households-in-the-us/">https://www.statista.com/statistics/242022/number-of-single-person-households-in-the-us/</a><br>[10] <a href="https://docs.google.com/spreadsheets/d/1GCnHpCwiKXtqavANugHH6f1pC4y2M70EaDXRUoJ42Ec/edit?usp=sharing">https://docs.google.com/spreadsheets/d/1GCnHpCwiKXtqavANugHH6f1pC4y2M70EaDXRUoJ42Ec/edit?usp=sharing</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=947507927c4b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Covid-19 still has the same exponential growth despite France strict mass quarantine]]></title>
            <link>https://medium.com/@hartator/covid-19-still-has-the-same-exponential-growth-despite-france-strict-mass-quarantine-115c751dc9bc?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/115c751dc9bc</guid>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[france]]></category>
            <category><![CDATA[quarantine]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Wed, 08 Apr 2020 04:21:45 GMT</pubDate>
            <atom:updated>2020-04-08T04:21:45.625Z</atom:updated>
            <content:encoded><![CDATA[<p>France started strict mass quarantine on March 17th. Strict defined by needing a printed permit to got out to get grocery from a 1-kilometer radius and for 1 hour maximum backed up by police enforcement with actual fines and even jail times. [1]</p><p>Successful measures should have shown statistically significant effects before 5.1 days which is covid-19 median incubation time. [2] However after 21 days of strict mass quarantine, new covid-19 cases [3] still follows the same exponential growth:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RzClfSmgvp6uJYuzLTNXzQ.png" /></figure><p>Strict mass quarantine doesn’t seem any effect on the rate of spread of the coronavirus. There is no “flatten the curve” due to it.</p><p>Strict mass quarantine does affect negatively our capacity to re-allocate resources effectively both on a private and public level. It’s hindering our capacity to fight the coronavirus and address patients. It will also come with its own death tolls: suicides going up [4], stress induced deaths by heart diseases [5], and stress induced deaths by progressing cancers.[6]</p><p>We should abandon these kind of measures urgently and focus our energy elsewhere.</p><p>References:<br>[1] <a href="https://fr.wikipedia.org/wiki/Confinement_de_2020_en_France">https://fr.wikipedia.org/wiki/Confinement_de_2020_en_France</a><br>[2] <a href="https://www.ncbi.nlm.nih.gov/pubmed/32150748">https://www.ncbi.nlm.nih.gov/pubmed/32150748</a><br>[3] <a href="https://aatishb.com/covidtrends/?location=France">https://aatishb.com/covidtrends/?location=France</a><br>[4] <a href="https://www.bmj.com/content/347/bmj.f5239">https://www.bmj.com/content/347/bmj.f5239</a><br>[5] <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2633295/">https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2633295/</a><br>[6] <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3037818/">https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3037818/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=115c751dc9bc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Force Google to Show “Total Results” When Scraping]]></title>
            <link>https://medium.com/serpapi/how-to-force-google-to-show-total-results-when-scraping-16ba8b706474?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/16ba8b706474</guid>
            <category><![CDATA[scraping-google-images]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[scraping]]></category>
            <category><![CDATA[google-scraping]]></category>
            <category><![CDATA[scrape]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Wed, 18 Sep 2019 17:03:17 GMT</pubDate>
            <atom:updated>2019-09-18T17:06:10.251Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to Force Google to Show “total_results” When Scraping</h3><p>Users of our Google scraping API have noticed that certain queries don’t return the total_results property that you would normally see on the top left side of the Google interface.</p><figure><img alt="Google total results property" src="https://cdn-images-1.medium.com/max/833/1*R3JzmPXIGaupAawtV1HXQQ.png" /><figcaption>Google results showing “total_results” value</figcaption></figure><p>The total_results property is a number returned by Google that represents the total number of results in the result set of that particular query. That number is just an approximation and may not represent the exact number of results, nor will Google allow you to retrieve all those results — you can read more on that in our documentation regarding <a href="https://serpapi.com/pagination">pagination</a>.</p><p>Our API allows a user to scrape Google for the total_results property and will return a number value in JSON format, as shown in the example below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/841/1*Pv4NTzyMNAXJZYfzGI-KGQ.png" /><figcaption>API returned JSON format with “total_results” property and value</figcaption></figure><p>What happens when Google doesn’t return the total_results value for a particular search? An example of a query that won’t return the “total_results” property is “SQL Entwickler,” which we will be using for our example. You can follow along in this example by using our <a href="https://serpapi.com/playground">playground</a>.</p><p>Notice when searching “SQL Entwickler,” Google fails to return a value for the total_results property.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/892/1*ALSaDzlph4gnpVUsdRMlQA.png" /><figcaption>Google search results indicating missing “total_results” value</figcaption></figure><p>You will also notice that our API did not return a value for the total_results property.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/734/1*rzefIWafO63qC8HIINXAtw.png" /><figcaption>API returned JSON format <strong>without</strong> “total_results” property and value</figcaption></figure><p>To override this issue, we will navigate to our playground and set the pagination parameter to a value of “1.” Set your search query to the same value of “SQL Entwickler.” Hit the search button when all your parameters are set.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dj0P4mlgq-6Aof5bJQnS3Q.png" /><figcaption>SerpApi Playground showing “pagination” value</figcaption></figure><p>Navigate over to the top right corner of the page in the playground and click “Export to Code,” which will provide you with the results of your query. As you will notice, you are provided with URLs for results in both HTML and JSON format. Navigate to either URL and you will notice that the total_results value is displayed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/957/1*PCNkFz8PsdvJGgPMC5qD7g.png" /><figcaption>Links to API results</figcaption></figure><p>The returned results now include the total_results value displayed in its usual place, as expected.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*snL9x2FlP1NQajkciMpW2w.png" /><figcaption>API results in HTML format</figcaption></figure><p>Our API returned also returned the missing total_results value in JSON as well.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/1*RolVRUzcalmCwL7jWsSYzA.png" /><figcaption>API results in JSON format</figcaption></figure><p>What appears to have happened is that either the first or second SERP result may have triggered some other informational property to show up, and somehow replace the total_results property on Google. Nonetheless, this fix works perfectly fine.</p><p>Visit <a href="http://serpapi.com/">SerpAPI</a> for additional information on our app and try different configurations and searches utilizing our <a href="http://serpapi.com/playground">playground</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=16ba8b706474" width="1" height="1" alt=""><hr><p><a href="https://medium.com/serpapi/how-to-force-google-to-show-total-results-when-scraping-16ba8b706474">How to Force Google to Show “Total Results” When Scraping</a> was originally published in <a href="https://medium.com/serpapi">SerpApi</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[MongoDB 4.2 vs 4.0, 3.6, 3.4, and 3.2 Benchmarks]]></title>
            <link>https://medium.com/@hartator/mongodb-4-2-vs-4-0-3-6-3-4-and-3-2-benchmarks-ee96a09ef231?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/ee96a09ef231</guid>
            <category><![CDATA[performance]]></category>
            <category><![CDATA[ruby]]></category>
            <category><![CDATA[mongodb]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[benchmark]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Sat, 14 Sep 2019 22:14:28 GMT</pubDate>
            <atom:updated>2019-09-14T22:26:19.433Z</atom:updated>
            <content:encoded><![CDATA[<p>I couldn’t find any benchmark measuring MongoDB newer versions against each other. I’ve decided to make my own.</p><p>We are going to do very very basic insertions, reads, and deletes. Regular defaults are kept for both the server and the Ruby client. Apart disabling logging on the client. Run on Ubuntu 18.04 DigitalOcean droplet with 2 CPUs and 4GB of Ram.</p><p>Overall results:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WLzPgonY-Cm0tatiga3WoQ.png" /></figure><p>It’s a little disappointing. Performance seems to have degraded since the release of 3.6 (in yellow above). MongoDB 3.2 and 3.4 seem to be in the same ballpark of performance whereas MongoDB 3.6, 4.0, and 4.2 seem to have introduced changes that led to a degradation of performance. We are seeing this for inserts, reads, and deletes. For example, inserts in 3.4 were around 53% faster than in 3.6.</p><p>It will be very interesting to know how come it wasn’t possible to be back to this level of performance in more recent versions of MongoDB. Maybe some features or regressions will be worth checking into.</p><p>Full Ruby script to run the benchmark:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/b0ef0d7d20c0c39c2b3151938d66fcd3/href">https://medium.com/media/b0ef0d7d20c0c39c2b3151938d66fcd3/href</a></iframe><p>Raw results:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/6c7f2c25c5850a7a5dd71400d4413f56/href">https://medium.com/media/6c7f2c25c5850a7a5dd71400d4413f56/href</a></iframe><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/d431923fd9de52d3d45d4548caaf1fe8/href">https://medium.com/media/d431923fd9de52d3d45d4548caaf1fe8/href</a></iframe><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/a807360f9ab204c8bce9cf0f3724bd65/href">https://medium.com/media/a807360f9ab204c8bce9cf0f3724bd65/href</a></iframe><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/1c85b9e022f8c5e6c89caa72be9f160d/href">https://medium.com/media/1c85b9e022f8c5e6c89caa72be9f160d/href</a></iframe><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/b9fb222dc0b590610b455e2f2d47ccd7/href">https://medium.com/media/b9fb222dc0b590610b455e2f2d47ccd7/href</a></iframe><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/12472ad07f9f96ef3c2e08d8ab62c90c/href">https://medium.com/media/12472ad07f9f96ef3c2e08d8ab62c90c/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ee96a09ef231" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[SerpApi Dashboard Redesign]]></title>
            <link>https://medium.com/serpapi/serpapi-dashboard-redesign-c13cddbcb9da?source=rss-d009ce9b7027------2</link>
            <guid isPermaLink="false">https://medium.com/p/c13cddbcb9da</guid>
            <category><![CDATA[design]]></category>
            <dc:creator><![CDATA[hartator]]></dc:creator>
            <pubDate>Wed, 02 Jan 2019 23:48:25 GMT</pubDate>
            <atom:updated>2019-01-03T00:28:17.003Z</atom:updated>
            <content:encoded><![CDATA[<p>We’ve decided to undertake a redesign of our dashboard. The submenu can now be found on the left. It will help users to navigate our website much smoother after the submenu options began to grow:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ChBKIjnGV6pWwy3WcvRjmg.png" /><figcaption>Old submenu — too much clutter! We wanted to keep clickable elements to a maximum of five.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X6qd1yk5w-xWMM5Coi4mBw.png" /><figcaption>Old submenu — it was even worse for the submenu of the submenu!</figcaption></figure><p>Our redesign entails a left-to-right navigation system that is easily readable for users. The new color palette contains minimal soft colors to optimize mobile rendering. Here are a few before and after screenshots.</p><p>Search API documentation page:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2UlrxDNJk7GYuVKcUTPvdA.png" /><figcaption>Before</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6OkKnFqN9HWN4CaaPUlOjg.png" /><figcaption>After</figcaption></figure><p>Search API archive documentation page:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GKbJqcoFS6WZP_kb3SdPig.png" /><figcaption>Before</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gtxm7uP31f3L6susMJ0HfQ.png" /><figcaption>After</figcaption></figure><p>Search list page:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wk-oPMXyu1Z74nnPj5-2LQ.png" /><figcaption>Before</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tremUy1q391MoFmsPIjhFA.png" /><figcaption>After</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c13cddbcb9da" width="1" height="1" alt=""><hr><p><a href="https://medium.com/serpapi/serpapi-dashboard-redesign-c13cddbcb9da">SerpApi Dashboard Redesign</a> was originally published in <a href="https://medium.com/serpapi">SerpApi</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>