Skipping some files when building new gem - ruby

I've never built my own gem before, so this is a totally new experience. I've built a library for my employer, which has to connect to a database of ours in order to run some of my tests. I want to make sure this config.yml file isn't included in the gem when I publish it. I've added it to the .gitignore file, as I know to do that, but is there some other change I have to make to the .gemspec maybe? Or does bundler only include files it knows about when releasing the library to the public? Just trying to be cautious, first time publishing open source code!
Specification after #Oleander replied:
Running git ls-files in the command line doesn't include the file I want to remove, and my spec.require_paths in the .gemspec file reads like this: ["lib"]. The spec directory isn't there. Does that mean the config.yml file won't be included when I release the gem?

Remove config.yml from s.files in your project.gemspec file inside your project and the file wont show up in your .gem-file.
s.files = `git ls-files`.split($/)
s.files -= ["spec/config.yml"]

Related

Struggling to run my ruby cli app

I built a small Cli app in ruby (first ever ruby app), but I'm not able to actually run it.
I started the project off with bundle gem. I have been testing it inside my apps folder by running bundle exec bin/konstruct and everything works fine.
I want to install it locally and test it before I submit it, and so I ran bundle exec rake install after updating all of the info in my gemspec file. It gave successfull output:
konstruct 0.1.1 built to pkg/konstruct-0.1.1.gem.
konstruct (0.1.1) installed.
But when I run the app by entering $ konstruct it gives me an Could not locate Gemfile or .bundle/ directory error, unless I run it from within the root folder of the app.
I have tried to Google but I can't pin the results down to the same issue as I'm having. I've been having this problem even in development (How can I test my Ruby CLI app while still in development?)
I do have a Gemfile and it contains:
source 'https://rubygems.org'
# Specify your gem's dependencies in konstruct.gemspec
gemspec
I don't have a .bundle/ folder though. I'm not sure where/how that folder should be generated. I have run bundle and bundle install many times now.
I have a konstruct.gemspec file which contains: https://github.com/konstruct/konstruct.cli/blob/master/konstruct.gemspec
Most of that file is as it was generated. I just filled it in where I could.
I'm sorry if this is a stupid question, but I am super stuck.
You have the following line in your gemspec:
spec.executables = spec.files.grep(%r{^exe/}) { |f| File.basename(f) }
I believe that means if you move your executable file into an exe directory instead of a bin directory, it will work as you expect.
http://guides.rubygems.org/specification-reference/#executables
It turns out I am a first class idiot. Some time ago I added Konstruct in my .path file, which obviously was now overriding the proper konstruct command. Back then what I did din't work, so I just forgot about it.
So the answer to this question: It's not broken. OP is an idiot. :-/

Missing File in Gem after Build

TL;DR:
Don't run bundle within an existing git repository. Weird things will happen without any error messages.
Original Question:
I've built a gem by adapting the steps in this tutorial:
http://net.tutsplus.com/tutorials/ruby/gem-creation-with-bundler/
As a final step, I've run gem build .gemspec
This succeeds, but when I install the gem I find the critical file, the one which contains my code, isn't in the gem. Another file in the same (lib) directory, "version.rb", does exist in the gem.
I don't know how to start debugging this...how does bundler/gem build decide which files to include in the gem?
Edit:
My workflow is:
gem build <project_name>.gemspec
gem unpack <project_name>
=> confirm file does not exist in <unpacked>/lib/
gem install <project name>
=> confirm file structure in ~/home/stefan/.rvm/... contains gem, but does not contain desired file
Edit 2 / Resolution:
I was finally able to get this working by committing all my code to a remote repository, creating a clean clone, and building the gem. The new gem included all the required files.
A bit of history...I originally created the code and committed it before thinking about making a gem (this is my first gem). I then used bundle inside the original repository, which didn't complain, but was probably the reason for the weirdness.
One of the things bundler did for you is start a local git repo to version-manage your gem code. Check that you have added the file in git
git add lib/gem_name/missing_file.rb
Bundler generated gems use git internally to track "membership" of source files for the gem. You can see this in the .gemspec where it uses backticks to call out to git and generate a file list:
gem.files = `git ls-files`.split($/)
Note this also means you should pay attention to what you list in .gitignore

Why is bundler using the wrong gemspec file?

I've got a custom gem that has been working just fine with regards to bundling, building, distributing, & implementing. The gem is the core of a framework from which other gems are derived. Since most derived gems will have the same basic structure, I want to include a Ruby script in the bin path of the gem that can be used to basically copy files from a template folder into a new folder where the user will develop their own gem.
The problem I'm having is that the template folder has a gemspec file named $name$.gemspec with similarly named classes/modules in the file (e.g.: module $Name$), where the $name$ gets replaced with a name provided by the user.
Unfortunately, when I run bundle install from my gem's top-most path, I get an error:
There was a SyntaxError while evaluating $name$.gemspec:
C:/my_gem/template/$name$.gemspec:8: syntax error, unexpected tGVAR
gem.version = MyGem::$Name$::VERSION
It looks like Bundler is using the wrong Gemfile, even if I explicitly pass the Gemfile or path via one of the following:
bundle install --gemfile=Gemfile
bundle install --path=C:\my_gem
I also tried updating the gemspec line of my Gemfile to no avail:
gemspec name: 'my_gem'
Lastly, I've ensured that the template folder isn't even included in my_gem.gemspec, but that doesn't seem to matter:
gem.files = Dir.glob("lib/**/*") + %w(LICENSE.txt README.md)
Does anyone know why Bundler is trying to read the ./template/$name$.gemspec instead of ./my_gem.gemspec?
Inspecting the Bundler source, I may have spotted the culprit in lib/bundler/source/path.rb. There's GLOB used to find gemspecs in load_spec_files. The default glob is "{,*,*/*}.gemspec". This will find *.gemspec in the root directory of your gem or any directory one descendant from root (which will include your template dir).
If this is indeed the culprit, you could work around this by placing your template directory deeper in your gem's dir hierarchy or changing the name of the template file so it doesn't end in .gemspec. The Bundler::Source::Pathobject looks like it can take a different glob at initilization but I haven't dug deep enough to see if there's a viable way to specify this alternative glob in bundle execution via config or cmdline options.

Rubygem Executable $LOAD_PATH Issues

I'm writing an IDE in Ruby, and I'm stumped on how to get all my files to get "required" when I run the program on the command line, AND when its installed as a Rubygem.
My Rubygem has an executable file named "vr" in it. I need to make this "vr" executable file "require" all the other files from my project.
When I'm developing, its easy to require all my project's files. I simply "require" a relative path to them like this:
require_all Dir.glob(File.expand_path(File.dirname(__FILE__)) + "/../bin/**/*.rb")
The require_all gem will work perfectly. However, I get a big problem when I install this program as a rubygem. When my "vr" executable is installed by rubygems, it copies the "vr" executable to a special directory:
/home/eric/.rvm/gems/ruby-1.9.3-p125/bin
This directory is totally separated from my project's root folder. And so all my project's files are no longer found by the "require" statement.
Rubygems makes this directory for my gem's root:
/home/eric/.rvm/gems/ruby-1.9.3-p125/visualruby-0.0.55
I need to be able to "require" all the files from that directory into my project.
My solution so far, is to make a second file called "visualruby.rb" that resides in my project's lib folder. It has the require_all statement in it to require all the project files. Then I just have to link the executable to it by adding this code to my "vr" executable file:
base_file = File.dirname(__FILE__) + '/lib/visualruby.rb'
if File.file?(base_file)
require base_file #load from project
else
require 'visualruby.rb' #load from gem
end
It is necessary to check if there's a file named "visualruby.rb" relative to the current file because when I'm developing, it will always find the installed gem's version of "visualruby.rb" So when I make a change to a file, it has no effect. I have to force it to load the version from my development project for changes to work.
Also, my IDE creates projects from scratch, so it would be nice to know the general solution to this. I'd like to have a consistent project file system for all projects, but I'm not sure that's possible. I had the general solution of making a file called "requires.rb" for all projects, but I don't think it will work because every project will have the same filename added to the $LOAD_PATH.
Please help me understand how I can make a consistent file structure where I can develop, and make rubygems.
I found the answer to my own question:
The problem was that I was installing my rubygems using the rubygems API:
Gem::Installer.new(file_name)
This created syslinks that messed up my paths. There is an option to make wrappers instead of syslinks and that seems to be the standard way to install:
Gem::Installer.new(file_name, :wrappers => true)
Now the a wrapper is copied to my gem's bin directory and it uses the correct path. Now I can have a universal file that can be made into a gem. And everything runs the same in development and in the gem.
A happy ending...

Why exclude tests from gemspec files?

Bundler's gemspec template provides the following as the default for specifying a gems files:
spec.files = `git ls-files -z`.split("\x0").reject { |f| f.match(%r{^(test|spec|features)/}) }
A lot of other (seemingly irrelevant) files like the .gitignore and .travis.yml file would be included, so why exclude the tests?
I've seen other high-profile gems ignore test files too.
Besides holding arbitrary metadata about a gem, one important thing that the Gemspec does is tell Rubygems which files are needed to build your gem.
I think Bundler recommends excluding tests because they aren't functional code that is needed for your gem to be built or run. Including every test suite for every Gem in the Ruby universe would be expensive. There isn't a compelling reason you would need to access Omniauth's tests from within your app.
The tests are important, and that's why they were exist as part of the repository. They just aren't necessary in production, so they don't get wrapped up in the package you download from Rubygems.
As for .gitignore and .travis.yml you can exclude these on your own. I don't think the Bundler team was trying to think of every scenario, just the most likely ones.

Resources