Turn your tutorials into tests and never let your docs be out of date again.
Start off by writing your tutorial in modified-markdown, then execute it with rundoc. If there's a problem with following the directions, then your tutorial will fail to build. When it succeeds, the real world output is embedded in the output markdown file. That means your tutorials will have the EXACT output that your readers will see.
Install the Ruby library:
$ gem install rundoc
Make a rundoc file:
$ mkdir /tmp/rundoc-demo
$ cd /tmp/rundoc-demo
$ cat <<'EOF' > ./RUNDOC.md
```
:::>> $ echo Hello World
```
EOF
Run it:
$ rundoc --on-success-dir=rundoc_output ./RUNDOC.md
View the output
$ cat rundoc_output/README.md
```
$ echo Hello World
Hello World
```
This software is distributed as a Rubygem. Install it manually:
$ gem install rundoc
or add it to your Gemfile:
gem 'rundoc'
Run the rundoc command on any rundoc-flavored markdown file:
$ rundoc <test/fixtures/rails_7/rundoc.md>Note: This command will create and manipulate directories in the working directory of your source markdown file. Best practice is to have your source markdown file in its own empty directory.
This will generate a project folder with your project in it, and a markdown README.md with the parsed output of the markdown docs. See rundoc --help for more configuration options.
- Execute Bash Commands
- Printing
- Chain commands
- Manipulate Files
- Boot background processes such as a local server
- Take screenshots
- Configure RunDOC
- Import and compose documents
RunDOC uses GitHub flavored markdown. This means you write like normal but in your code sections you can add special annotations that when run through RunDOC can generate a project.
All RunDOC commands are prefixed with three colons ::: and are inclosed in a code block a
command such as $ which is an alias for bash commands like this:
```
:::>- $ git init .
```
Nothing before the three colons matters. The space between the colons and the command is optional.
If you don't want the command to output to your markdown document you
can add a minus symbol - to the end to prevent it from being
rendered.
```
:::-- $ git init .
```
Note: If all commands inside of a code block are hidden, the entire codeblock will not be rendered.
If you want the output of the actual command to be rendered to the screen you can use two arrows so that:
```
:::>> $ ls
```
This code block might generate an output something like this to your markdown doc:
```
$ ls
Gemfile README.rdoc app config.ru doc log script tmp
Gemfile.lock Rakefile config db lib public test vendor
```
Any items below the command will be passed into the stdin of the command. For example using a $ command you can effectively pipe contents to stdin:
```
:::>> $ tail -n 2
foo
bar
baz
bahz
```
Would output:
$ tail -n 2
baz
bahz
This STDIN feature could be useful if you are running an interactive command such as play new which requires user input.
Different commands will do different things with this input. For example the rundoc command executes Ruby configuration code:
```
:::-- rundoc
Rundoc.configure do |config|
config.after_build do
puts "you could push to GitHub here"
puts "You could do anything here"
puts "This code will run after the docs are done building"
end
end
```
And the website.visit command allows you to navigate and manipulate a webpage via a Capybara API:
```
:::>> website.visit(name: "localhost", url: "http://localhost:3000", scroll: 100)
session.execute_script "window.scrollBy(0,100)"
session.click("sign up")
```
RunDOC only cares about things that come after a ::: section. If you have a "regular" code section, it will be rendered as as normal:
```
$ echo "I won't run since i'm missing the :::>> at the front"
```
You can mix non-command code and commands, as long as the things that aren't rendering come first. This can be used to "fake" a command, for example:
$ rails new myapp # Not a command since it's missing the ":::>>"
:::-> $ rails new myapp --skip-test --skip-yarn --skip-sprockets
:::>> | $ head -n 5
This will render as:
$ rails new myapp # Not a command since it's missing the ":::>>""
create
create README.md
create Rakefile
create .ruby-version
create config.ru
In this example it looks like the command was run without any flags, but in reality rails new myapp --skip-test --skip-yarn --skip-sprockets | head -n 5 was executed. Though it's more explicit to use a print.text block, see #print.text for more info.
An arrow > is shorthand for "render this" and a dash - is shorthand for skip this section. The two positions are command first and result second.
:::>-(YES command output, not result output):::>>(YES command output, YES result output):::--(not command output, not result output):::->(not command output, YES result output)
Current Commands:
$fail.$
Anything you pass to $ will be run in a shell. If a shell command returns a non-zero exit status an error will be raised. If you expect a non-zero exit status use fail.$ instead:
```
:::>> fail.$ cat /dev/null/foo
```
Even though this command returns a non zero exit status, the contents of the command will be written since we're stating that we don't care if the command fails. This would be the output:
```
$ cat /dev/null/foo
cat: /dev/null/foo: Not a directory
```
Some commands may be custom, for example when running cd you likely want to change the working directory that your script is running in. To do this we need to run Dir.chdir instead of shelling out. So this works as you would expect:
```
:::>> $ cd myapp/config
:::>> $ cat database.yml
```
However this command would fall on its face:
```
:::>> $ cd myapp && cat config/database.yml
:::>> $ rails g scaffold users # <=== This command would be in the wrong directory, not `myapp`
```
These custom commands are kept to a minimum, and for the most part behave as you would expect them to. Write your docs as you normally would and check the output frequently.
Running shell commands like this can be very powerful, you'll likely want more control of how you manipulate files in your project. To do this you can use the file. namespace:
Current commands:
print.textprint.erb
Behaves slightly differently than other commands. The "command" portion of the control character i.e. :::> controls whether the contents will be rendered inside the block or before the block (versus usually this is used to control if the command such as $ cd is shown).
:::>>Print inside the code block:::->Print BEFORE the code block, if multiple calls are made, they will be displayed in order.:::--Nothing will be rendered, can be used to pass data to another rundoc command via the pipe operator.:::>-Same behavior as:::--.
This functionality is present to allow body text to be generated (versus only allowing generated text in code blocks).
Use the print.text keyword followed by what you want to print:
```
:::-> print.text
I will render BEFORE the code block, use :::>> to render in it.
It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness,
it was the epoch of belief, it was the epoch ...
```
Specifying :::-> with print.text will render text without a code block (or before the code block if there are other rundoc commands). If you want to render text with a code block you can do it via :::>>.
To dynamically change the contents of the thing you're printing you can use print.erb:
```
:::-> print.erb
I will render BEFORE the code block, use :::>> to render in it.
What a week!
Captain it's only <%= Time.now.strftime("%A") %>!
```
This will evaluate the context of ERB and write it to the file. Like print.text use :::-> to write the contents without a code block (or before the code block if there are other rundoc commands). If you want to render text with a code block you can do it via :::>>.
ERB commands share a default context. That means you can set a value in one print.erb section and view it from another. If you want to isolate your erb blocks you can provide a custom name via the binding: keyword:
```
:::>> print.erb(binding: "mc_hammer")
I will render IN a code block, use `:::->` to render before.
<%= @stop = true %>
:::>> print.erb(binding: "different")
<% if @stop %>
Hammer time
<% else %>
Can't touch this
<% end %>
```
In this example setting @stop in one print.erb will have no effect on the other.
Current Commands:
file.writefile.appendfile.remove
Use the file.write keyword followed by a filename, on the next line(s) put the contents of the file:
```
:::>- file.write config/routes.rb
Example::Application.routes.draw do
root :to => "pages#index"
namespace :users do
resources :after_signup
end
end
```
If the exact filename is not known you can use a file glob (*).
If you wanted to change users to products you could write to the same file again.
```
:::>- file.write config/routes.rb
Example::Application.routes.draw do
root :to => "pages#index"
namespace :products do
resources :after_signup
end
end
```
To fully delete files use bash $ command such as ::: $ rm foo.rb.
To add contents to a file you can use file.append
```
:::>> file.append myapp/Gemfile
gem 'pg'
gem 'sextant', group: :development
gem 'wicked'
gem 'opro'
```
The contents of the file (in this example a file named Gemfile) will remain unchanged, but the contents of the file.append block will now appear in the bottom of the file. If you want to append the contents to a specific part of the file instead of the end of the file you can specify line number by putting a hash (#) then a number following it.
```
:::>> file.append myapp/Gemfile#22
gem 'rails_12factor'
```
This will add the gem 'rails_12factor' on line 22 of the file myapp/Gemfile. If line 22 has existing contents, they will be bumped down to line 23.
Some times you may want to remove a small amount of text from an existing file. You can do this using file.remove, you pass in the contents you want removed:
```
:::>> file.remove myapp/Gemfile
gem 'sqlite3'
```
When this is run, the file Gemfile will be modified to not include gem 'sqlite3'.
Note: file.remove currently requires a very explicit match so things like double versus single quotes, whitespace, and letter case all matter. Current best practice is to only use it for single line removals.
Commands:
|pipe(aliased|)
Sometimes you need to need to pass data from one command to another. To do this there is a provided pipe command |.
Let's say you want to output the first 23 lines of a file but you don't want to confuse your users with an additional pipe command in your shell line you could write something like this:
:::> $ cat config/database.yml
:::>> | $ head -n 23Anything after the pipe | will generate a new command with the output of the previous command passed to it. The pipe command will only ouput its result, so the user will not know it was even executed.
This command is currently hacked together, and needs a refactor. Use it, but if something does not behave as you would expected open an issue and explain it.
Sometimes you want to start a long lived process like a server in the background. In that case, the background namespace has your, well, back.
To start a process, pass in the command as the first arg, and give it a name (so it can be referenced later):
:::>> background.start("rails server", name: "server")
- Arguments
- name: Identifier of the background process, used in later invocations.
- wait (Optional): A string to wait for before continuing. This is useful to block moving on until an event happend such as a server was booted or web request was received. Also see
background.wait - timeout (Optional): A number of seconds to wait for a given string to be found in the logs.
- out (Optional): A bash redirect. Defaults to
2>&1to merge stderr and stdout together. - allow_fail (Optional): Set to
trueif the command exiting shouldn't halt doc generation. Poorly named, may be renamed in the future.
You can make the background process wait until it receives a certain string in the logs. For instance to make sure that the server is fully booted:
:::>> background.start("rails server", name: "server", wait: "Listening on")
You can stop the process by referencing the name:
:::-- background.stop(name: "server")
- Arguments
- name
You can also get the log contents:
:::>> background.log.read(name: "server")
- Arguments
- name
You can also truncate the logs:
:::>> background.log.clear(name: "server")
- Arguments
- name
You can also wait for a given string to appear in the logs:
:::>> background.wait(name: "server", wait: "method=GET")
- Arguments
- name
- wait: Same as
background.starta string value to look for before continuing. - timeout: Maximum number of seconds to wait
You'll need selenium and chromedriver installed on your system to make screenshots work. On a mac you can run:
$ brew cask install chromedriver
To take a screenshot first "visit" a website. The values you pass in to stdin can be used to further navigate. For more information see the Capybara DSL. Use the keyword session
Once you're on the page you want to capture you can execute website.screenshot:
:::>> website.visit(name: "localhost", url: "http://localhost:3000", scroll: 100)
session.execute_script "window.scrollBy(0,100)"
session.first(:link, "sign up").click
:::>> website.screenshot(name: "localhost")
The result of the screenshot command will be to replace the code section with a markdown link to a relative path of the screenshot.
Once you've visited a website you can further navigate using website.nav or website.navigate:
:::>> website.visit(name: "localhost", url: "http://localhost:3000")
:::>> website.navigate(name: "localhost")
session.execute_script "window.scrollBy(0,100)"
session.first(:link, "sign up").click
:::>> website.screenshot(name: "localhost")
You can specify that you want to upload files to S3 instead of hosting them locally by passing in upload: "s3" to the screenshot command:
:::>> website.visit(name: "localhost", url: "http://localhost:3000", scroll: 100)
:::>> website.screenshot(name: "localhost", upload: "s3")
To authorize, you'll need to set these environment variables:
AWS_ACCESS_KEY_ID
AWS_REGION
AWS_SECRET_ACCESS_KEY
AWS_BUCKET_NAME
The bucketeer addon on Heroku is supported out of the box. To specify project specific environment variables see the "dotenv" section below.
You can also break up your document into smaller components using rundoc.require:
:::>> rundoc.require "../day_one/rundoc.md"
This will prepend the code section with the generated contents of rundoc.require.
If you want to execute another tutorial as a pre-requisite but not embed the results you can use :::--:
:::-- rundoc.require "../day_one/rundoc.md"
If you need to specify project specific environment variables create a file called .env at the same directory as your rundoc.md and it will be imported. Add this file to your .gitignore so you don't accidentally share with the world
You can configure your docs in your docs use the RunDOC command
```
:::-- rundoc.configure
```
Note: Make sure you run this as a hidden command (with -).
After Build
This will eval any code you put under that line (in Ruby) when the build was successful but before the contents are finalized on disk. If you want to run some code after you're done building your docs you could use Rundoc.configure block and call the after_build method like this:
```
:::-- rundoc.configure
Rundoc.configure do |config|
config.after_build do |context|
puts "you could push to GitHub here"
puts "You could do anything here"
puts "This code will run after the docs are done building"
end
end
```
The context object will have details about the structure of the output directory structure. The stable API is:
context.output_dir: A Pathname containing the absolute path to the top level directory where all commands are were executed. If your script runsrails new myappthen this directory would contain another directory namedmyapp. Only modifications to this directory will be persisted to the final--output-dir.context.screenshots_dir: A Pathname containing the absolute path to the directory where screenshots were saved. It is guaranteed to be somewhere within thecontext.output_dircontext.output_markdown_path: A Pathname containing the absolute path to the final markdown file. This is guaranteed to be in thecontext.output_dir
Filter Sensitive Info
Sometimes sensitive info like usernames, email addresses, or passwords may be introduced to the output readme. Let's say that your email address was [email protected] you could filter this out of your final document and replace it with [email protected] instead like this:
```
:::-- rundoc.configure
Rundoc.configure do |config|
config.filter_sensitive("[email protected]" => "[email protected]")
end
```
This command filter_sensitive can be called multiple times with different values. Since the config is in Ruby you could iterate over an array of sensitive data
Rundoc does not have a stable internal command interface. You can define your own commands, but unless it is committed in this repo, it may break on a minor version change.
To add a new command it needs to be parsed and called. Examples of commands being implemented are seen in lib/rundoc/code_command.
A new command needs to be registered:
Rundoc.register_code_command(:lol, Rundoc::CodeCommand::Lol)
They should inherit from Rundoc::CodeCommand:
class Rundoc::CodeCommand::Lol < Rundoc::CodeCommand
def initialize(line)
end
end
The initialize method is called with input from the document. The command is rendered (:::>-) by the output of the def call method. The contents produced by the command (:::->) are rendered by the def to_md method.
The syntax for commands is ruby-ish but it is a custom grammar implemented in lib/peg_parser.rb for more info on manipulating the grammar see this tutorial on how I added keword-like/hash-like syntax https://github.com/schneems/implement_ruby_hash_syntax_with_parslet_example.
Command initialize methods natively support:
- Barewords as a single string input
- Keyword arguments
- A combination of the two
Anything that is passed to the command via "stdin" is available via a method self.contents. The interplay between the input and self.contents is not strongly defined.
All content Copyright Richard Schneeman © 2020