Nokogiri VS Kiba

release planning: v1.16.0 · Issue #2897 · sparklemotion/nokogiri

What's holding me back from going all in with Crystal is I have a lot of pre-existing Ruby code, and porting Ruby code to Crystal can be tricky. For example, Crystal lacks an Enumerator class (aka generators) due to captured block semantics. I also wish the shards ecosystem was a little more mature; for example there's multiple HTML parsing libraries, but none have all of the features that Ruby's Nokogiri has. For new greenfield backend projects, I would totally use Crystal.

Welcome! Ruby isn't exactly "dying", but the hype/popularity is definitely fading. This is primarily because Ruby is no longer "new", most of Ruby's popularity came from Rails, and now Rails is no longer the "new hotness". However, Ruby still has lots of awesome features and lots of awesome other libraries and frameworks, such as the new fancy irb gem that uses reline, nokogiri, chunky_png, the async gems, Dragon Ruby, SciRuby, Ronin, and the new Hanami web framework.

great for hacking at xml

Mike has worked hard over the years to have Nokogiri come with its dependencies. It does come with libxml and all that is required.

From https://nokogiri.org

> These dependencies are met by default by Nokogiri's packaged versions of the libxml2 and libxslt source code, but a configuration option --use-system-libraries is provided to allow specification of alternative library locations.

Some authors work hard to have their tools do the right thing and consistently.

If the page is rendered as html you can use Nokogiri. It has great support and is pretty easy to get started with too.

And even now, pulling in a Java-based HTML5 parser is still probably easier than re-implementing in FFI, which is why I created https://github.com/sparklemotion/nokogiri/issues/2227 and would love to have this the conversation there if possible.

I started https://github.com/thbar/kiba#kiba-etl to scratch my own itch & be able to write properly structured ETL jobs in Ruby. It was a blank-slate rewrite of something larger (activewarehouse-etl) which I could not maintain anymore.

This landed me not strictly a job, but long term consulting gigs with a number of companies in EU, UK & US.

The job was directly related to the project: companies wanted the expertise of data engineering & ETL, often with Kiba directly, but also in general.

This "side project" was totally worth it :-)

Definitely! As a matter of fact, this is the default data structure I use when writing Ruby ETL code (e.g. https://github.com/thbar/kiba/wiki).

Methods like "except" (https://docs.ruby-lang.org/en/3.2/Hash.html#method-i-except) or "fetch" (raising an error on missing key) are very convenient to write defensive data processing code!

Similarly, in Elixir, I use Maps a lot for the same type of jobs (https://hexdocs.pm/elixir/1.15.4/Map.html), with similar properties.

Ruby has a very cool ETL library named Kiba that fits wonderfully with Ruby's strengths.

Though it might be overkill for your problem, but have you had a look at [kiba-etl](https://github.com/thbar/kiba/blob/master/README.md)?

Kiba

If you don’t use Kiba, but work with data, check it out.