Language complexity, measured by how many rules GitHub's .gitignore has | Hacker News Reader

Language complexity, measured by how many rules GitHub's .gitignore has

gist.github.com

104 pointsacidflask45 comments

I love these metrics that are 99% meaningless but you still want to know how your language compares. Like average identifier length, source tree directory depth, etc.

Making people think is my favorite part of my job. :)

Many of these categories are not 'languages' making it hard to make any conclusions beside 'people working with Joomla or VS seem to have some gigantic .gitignore template'.

too much code generators?

Typically you'd check in generated code; you'd exclude packages, compiled code, certain kinds of configuration and logs, etc.

Do you typically check in generated code? I think it depends on a lot of things. My understanding is the general rule of thumb is NOT to, except for certain cases (generation takes a long time, for instance)

Well, in the VS world generated code is typically generated once and open to modification post-generation, so yes.

If you are into novelty metrics, you might find the following interesting:

http://imgur.com/a/xhxtm

It shows how actively the gitignore files are modified.

"Novelty Metrics" -- I kind of want to put that on my resume just to see what happens with it. Or, I could imagine an entertaining blog existing solely based on that premise.

You might really enjoy the Strange Maps[1] section of bigthink.com. There are some real wonderful ones in the past entries. I have this in my feed, but look at it far too infrequently for how interesting it looks.

1: http://bigthink.com/articles?blog=strange-maps

A more accurate way imho to see this data is: "Tooling support, measured by how many rules GitHub's .gitignore has"

exactly. tooling and framework config files. for 'pre-compiled' languages also build dirs but that doesn't necessarily mean the language is more complex.

This measures the number of rules in .gitignore files, not language complexity.

"Build tool complexity" seems less reaching

It also probably depends on what development tools people use. If everyone on the project uses a different IDE that's different patterns for each of them, but they all have to be in the ignore file.

Well, that is what a global gitignore file [1] is all about. Entries specific for your working environment (IDE-files) should be defined in your global gitignore, while everything project-specific (object files, build-directory etc.) should be in the project-gitignore.

[1] git config-option: --global core.excludesfile

Not really. That should be put in ~/.gitignore, e.g. I ignore Vim swap files there so I don't have to do it in every single repo.

In Github's .gitignore template, too.

For example, most of the entries in the Python one are utterly inane, and won't apply to your project. You can easily trim that file to 1/5 it's size. It includes the ignores for a bunch of directories I've never see any project ever have, two unit test frameworks, two web frameworks, pip's files that are already better covered by virtualenv ignores, four different ways of naming your virtualenvs…

This isn't measuring language complexity; this is someone trying to pre-cover any potential case generated by any popular-today third party library.

The visual studio one seems to cover just about every way you could use VS, and every plugin made for it. This measurement comments more on the neatness of the IDE(s) than any kind of language complexity.

Agreed. VS supports dozens of languages across web, desktop, mobile, etc. C#, VB, F#, C, C++, JavaScript, T-SQL, Python... and that's just what I can list off the top of my head. Even with single languages in the list it's not apples-to-apples. "Language complexity" should be removed from the link title.

Elaboration: the number of rules in .gitignore files scale pretty linearly with the number of options you have for tooling. If there are lots of tools that work with your language, your .gitignore template will be big. If there is only one or a few tools, your .gitignore template will be small.

Whether or not "language complexity" correlates with "number of options for tooling" is left as an exercise for the reader. Personally, I think there's an argument to be made for popular languages having more tooling options...

Oh my dear... That's so pointless... It's like judging your intelect depending on the amount you dump every day.

Interesting, though especially for the entries in the list that are frameworks, you could conceivably just bury all of the ugly stuff in a `framework` folder, and have only one ignore.

And in a lot of ways, that would be an improvement to randomly dumping files all over the source tree.

Visual Studio seems to be moving in this direction: I've noticed a .vs folder and the .suo seems to live in it for now. I guess more and more will be moved into it as time goes on, but considering the .gitignore contains VS6 ignores we'll probably be stuck with the large file forever.

I'm not at all surprised Magento is number 3. I've never seen anything so simultaneously overly- and poorly-engineered.

It's like someone combined the worst parts of 10-year-old Java (XML configurations over conventions, AbstractEntityBeanFactory-style classes) and PHP (no namespace, autoloading override plugins) into one awful mess. Throw in the use of EAV everywhere, and the "schema" is a nightmare to decipher.

Maybe version 2 fixes some of these issues, but Magento 1.x is awful.

Magento 2 fixes the "no namespace" issue. Everything else is still there, and some parts are taken to eleven.

A while ago, I ran PHPStorm's code inspections against Magento 1.x core code, and the result wasn't very encouraging: https://imgur.com/RMxWEgR

TIL visual studio is a language.

Most developers don't know the difference between IDE, language, library, framework, platform and technology.

You got downvoted for your sarcasm, but you have a point. Some people think that Ruby classes are instantiated using the create method.

Yeah... Qt and Rails are in the list too. And is Symfony a language, too?

Also, Java is an incredibly simple language.

Ruby 25 and Rails 23, really? I'd expected the opposite. Looking at them, half of the files makes sense, the other half is really project and dev env dependent and they don't seem to be made for the same environment. Hardly comparable.

https://github.com/github/gitignore/blob/master/Ruby.gitigno...

https://github.com/github/gitignore/blob/master/Rails.gitign...

> Ruby 25 and Rails 23, really? I'd expected the opposite. Is there really a difference between 25 and 23?

Why no JavaScript ?

There's really no common patterns of just plain JavaScript files being ignored (to some extend, other cases are mostly covered by the `node` file).

Looks like a similar situation with Apache Groovy. Only "grails" and "gradle" are in there.

Here you go:

     *.js

What a useless exercise. It seems to me that we're seeming more and more of these irrelevant posts hitting the front page.

Be the change you want to see in this forum.

How do you propose he prevents inane content from being upvoted by others?

If oneloop doesn't like the submissions that have been made, oneloop (who appears to have submitted a total of "2" links in their posting history) should submit more of the kind of things they feel is appropriate.

Sure, if everyone submitted lots of good stuff it would be less likely that we'd see this garbage on the front page. But for me individually to make a difference I would have to come up with insane amounts of stuff myself. I don't have time for that. If I find good stuff sure, I'll share it. But if you're suggesting that I should spend more of my time looking for good stuff around the web for the sole purpose of helping out the quality of an aggregator website, you got another thing coming.

You seem to be ok with spending more of your time complaining about it.

at least it has a nice power-law distribution. otherwise pretty useless...

Reply on news.ycombinator.com