defmodule MixTemplates do @moduledoc ~S""" > NOTE: This documentation is intended for folks who want to write > their own templates. If you just want to use a template, then > have a look at the README, or try `mix help template` and > `mix help gen`. This is the engine that supports templated directory trees. A template is a trivial mix project that acts as the specification for the projects you want your users to be able to generate. It contains a single source file in `lib` that contains metadata and option parsing. It also contains a top-level directory called `template`. The directories and files underneath `template/` copied to the destination location. The copying function takes a map containing key-value pairs. This is passed to EEx, which is used to expand each individual file. Thus a template file for `mix.exs` may contain: ~~~elixir defmodule <%= @project_name_camel_case %>.Mixfile do use Mix.Project @name :<%= @project_name %> @version "0.1.0" . . . ~~~ The `<%= ... %>` constructs are expanded using the passed-in map. In addition, the template looks for the string `$PROJECT_NAME\$` in the _names_ of files and directories. It replaces each occurrence with the name of the project, taken from `assigns.project_name`. Thus the directory structure for a standard Elixir project might be: template ├── $PROJECT_NAME$ │   ├── README.md │   ├── config │   │   └── config.exs │   ├── lib │   │   └── $PROJECT_NAME$.ex │   ├── mix.exs │   └── test │   ├── $PROJECT_NAME$_test.exs │   └── test_helper.exs └── templates_project.ex ## Write a Template Make sure you have the underlying tools installed: $ mix archive.install hex mix_templates $ mix archive.install hex mix_generator Then install the template for templates (yup :). $ mix template.install hex gen_template_template Now create your template project: $ mix gen template my_template Wander into the directory that is created: $ cd my_template/ $ tree . ├── README.md ├── lib │   └── my_template.ex ├── mix.exs └── template └── $PROJECT_NAME$ └── your_project_tree_goes_here #### Add a Description Your first job is to update the metadata in lib/«whatever».ex: defmodule MyTemplate do @moduledoc File.read!(Path.join([__DIR__, "../README.md"])) use MixTemplates, name: :my_template, short_desc: "Template for ....", source_dir: "../template", based_on: :another_project, options: [ command line options unique to this template ] end For a simple template, the only change you're likely to make to the metadata is to update the short description. This is used to display information about the template when you list the templates you have installed, so you probably want to keep it under 70 characters. If you want to write a template that is based on another, use the `:based_on` option. This causes the parent template to be processed before your local template. This means your template need only implement the changes to the base. #### Add the Files The job of your template is to contain a directory tree that mirrors the tree you want your users to produce locally when they run `mix gen`. * The easiest way to start is with an existing project that uses the same layout. Copy it into your template under `template/$PROJECT_NAME$`. * Remove any files that aren't part of every project. * Look for files and directories whose names include the name of the project. Rename these, replacing the project name with the string $PROJECT_NAME$. For example, if you're following the normal convention for test files, you'll have a file called test/myapp_test.exs Rename this file to test/$PROJECT_NAME$.exs * Now you need to look through the files for content that should be customized to each new project that's generated. Replace this content using EEx substitutions: For example, the top-level application might be an Elixir file: defmodule MyApp do # . . . end Replace this with defmodule <%= project_name_camel_case %> do # . . . end There's a list of the available values in the next section. ### Test Your Template You can use `mix gen` to test your template while you're developing it. Simply give it the path to the directory containing the generator (the top level, with `mix.exs` in it). This path must start with a dot (".") or slash ("/"). $ mix gen ../work/my_generator test_project ### Publish Your Template Wander back to the `mix.exs` file at the top of your project, and update the `@description`, `@maintainers`, and `@github` attributes. Then publish to hex: $ mix hex.publish and wait for the praise. ## Standard Substitutions The following values are available inside EEx substitutions in templates. (Remember that the inside of a `<%= ...%>` is just Elixir code, so you aren't limited to this list. The next section describes how you can extend this set even further in your own templates.) #### Project Information Assuming the template was invoked with a project name of my_app: @project_name my_app @project_name_camel_case MyApp #### Date and Time These examples are from my computer in US Central Daylight Time (GMT-5) @now.utc.date "2017-04-11" @now.utc.time "00:49:37.505034" @now.utc.datetime "2017-04-11T00:49:37.505034Z" @now.local.date "2017-04-10" @now.local.time "19:49:37" @now.local.datetime "2017-04-10 19:49:37" #### The Environment @host_os "os-name" or "os-name (variant)" eg: "unix (darwin)" @original_args the original args passed to mix @elixir_version eg: "1.5.3" @erlang_version eg: "8.2" @otp_release eg: "19" @in_umbrella? true if we're in the apps_path directory of an umbrella project #### Stuff About the Template @template_module the module containing your template metadata @template_name the name of the template (from the metadata) @target_dir the project directory is created in this @target_subdir the project directory is called this ### Handling Command Line Parameters You may need to configure the output of your template depending on the options specified on the command line. For example, the standard `project` template lets you generate basic and supervised apps. To indicate you want the latter, you add a command line flag: $ mix gen project my_app --supervised This option is not handled by the `gen` task. Instead, it passes it to your template module (the file in your top-level `lib/`). You can receive the parameters by defining a callback ~~~ elixir defmodule MyTemplate do @moduledoc File.read!(Path.join([__DIR__, "../README.md"])) use MixTemplates, name: :my_template, short_desc: "Template for ....", source_dir: "../template" options: [ supervised: [ to: :is_supervised?, default: false ], sup: [ same_as: :supervised ], ] end ~~~ `options` is a specification of the command line parameters that your template accepts. In all cases, the key is the parameter as it appears on the command line, and the keyword list that is the value gives information about that option. The simplest option is ~~~ elixir name: [] ~~~ This says that `--name` is a valid option. If you add it to the command line with no value following it, then `:name` will appear in the assigns with the value `true`. It you pass in a value, then that value will appear in the assigns.project_name If you do not specify `--name` on the command line, there will be no entry with the key `:name` in the assigns. If your option takes an argument, you specify its name using `takes:`. ~~~ elixir name: [ takes: "your-name" ] ~~~ The `required` key says that a given parameter _must_ appear on the command line. ~~~ elixir name: [ takes: "your-name", required: true ] ~~~ `default` provides a value to use if the parameter does not appear on the command line: ~~~ elixir name: [ takes: "your-name", default: "nancy" ] ~~~ If a default value is given, the entry will _always_ appear in the assigns.project_name By default the name of the field in the assigns will be the key in the options list. You can override this using `to`. ~~~ elixir name: [ takes: "your-name", to: :basic_id, default: "nancy" ] ~~~ In this example, calling $ mix gen my_template my_app --name walter will create an assigns map that includes `\@basic_id` with a value of “walter.” Finally, you can alias a option using `same_as`. The following will allow both `--sup` and `--supervised` on the command line, and will map either to the key `:is_supervised?` in the assigns. ~~~ elixir options: [ supervised: [ to: :is_supervised?, default: false ], sup: [ same_as: :supervised ], ] ~~~ ### Dealing with optional files and directories Sometimes you need to include a file or directory only if some condition is true. Use these helpers: * `MixTemplates.ignore_file_and_directory_unless(«condition»)` Include this in a template, and the template and it's immediate directory will not be generated in the output unless the condition is true. For example, in a new mix project, we only generate `lib/«name»/application.ex` if we're creating a supervised app. The `application.ex` template includes the following: ~~~ elixir <% # ------------------------------------------------------------ MixTemplates.ignore_file_and_directory_unless \@is_supervisor? # ------------------------------------------------------------ %> defmodule <%= \@project_name_camel_case %>.Application do # ... end ~~~ Sometimes you just need to skip a single file if some condition is true. Use this helper: * `MixTemplates.ignore_file_unless(«condition»)` ### Binary Files By default, binary files are ignored if they exist in a template. To copy over binary files into generated projects, specify the `just_files` option, which functions as a whitelist of files that should be copied over directly. As an example: ~~~ elixir use MixTemplates, name: :my_template, just_files: [".png", /assets/*.gif"] ~~~ The above example will copy over all files with an extension of `png` as well as all files with the `gif` extension that are in the assets folder. It will not copy any files that do not match, they will simply be ignored. To whitelist all binary files, simply set a `just_files` value of `["*"]`. ### Cleaning Up In most cases your work is done once the template is copied into the project. There are times, however, where you may want to do some manual adjustments at the end. For that, add a `clean_up/1` function to your template module. ~~~ elixir def clean_up(assigns) do # ... end ~~~ The cleanup function is invoked in the directory where the project is created (and not inside the project itself). Thus if you invoke mix gen my_template chat_server in the directory `/Projects` (which will create `/Projects/chat_server`), the `clean_up` function's cwd will be `/Projects`. ### Deriving from Another Template Sometimes you want to create a template which is similar to another. Perhaps some files' contents are different, new files are added or others taken away. Use the `based_on: «template»` option to facilitate this: ~~~ elixir defmodule MyTemplate do \@moduledoc File.read!(Path.join([__DIR__, "../README.md"])) use MixTemplates, name: :my_template, short_desc: "Template for ....", source_dir: "../template", based_on: :project def populate_assigns(assigns, options) do # ... end end ~~~ The value of `based_on` is the name or the path to a template. When people create a project based on your template, the generator will run twice. The first time, it creates the `based_on` project. It then runs again with your template. Any files or directories in your template will overwrite the corresponding files in the based-on template. It isn't necessary to have a full tree under the `template` directory in your template. Just populate the parts you want to override in the base template. If you want to remove files generated by the base template, you can add code to the `clean_up/1` hook. Remember that the cleanup hook is invoked in the directory that contains the target project, so you'll need to descend down into the project itself. Obviously, this is something you'll want to test carefully before releasing :) ~~~ elixir def clean_up(assigns) do Path.join([assigns.target_subdir, "lib", "#{assigns.project_name}.ex"])) |> File.rm end ~~~ """ use Private alias Mix.Generator, as: MG alias MixTemplates.Cache defmacro __using__(opts) do name = mandatory_option(opts[:name], "template must include\n\n\tname: \"template_name\"\n\n") override_source_dir = Keyword.get(opts, :source_dir) quote do @doc """ Return the name of this template as an atom. This is the name passed to the gen command. """ def name do unquote(name) end @doc """ Return the short description of this template, or nil. """ def short_desc do unquote(opts[:short_desc]) end @doc """ Return a map where the keys are filename extensions and the keys are either the :all atom or a list of allowed files with that extension. """ def just_files do unquote(opts[:just_files]) end @doc """ Return the absolute path to the tree that is to be copied when instantiating this template. This top-level dir will typically just contain a directory called `$APP_NAME$`. """ def source_dir do cond do unquote(override_source_dir) -> Path.absname(unquote(override_source_dir), __DIR__) true -> __DIR__ end |> Path.join("$PROJECT_NAME$") end @doc """ Return the name or path of a template that this template is based upon. That template will be processed first, and then this one will be executed. """ def based_on do unquote(opts[:based_on]) end @doc """ Return the list of options supported by this template. """ def options do unquote(opts[:options] || []) end @doc """ Override this function to do any cleanup after your template has been copied into the user project. One use of this is to remove unwanted files created by a template upon which this template is based. """ def clean_up(_assigns) do nil end defoverridable clean_up: 1 end end def find(name) when is_binary(name) do name |> String.to_atom |> find end def find(name) when is_atom(name) do Cache.find(name) end def generate(template, assigns) do kws = [ assigns: assigns |> Map.to_list ] check_existence_of(assigns.target_dir, assigns.target_subdir) |> create_or_merge(template, kws) end # called from within a template to cause it not to generate either this # file or anything in this file's directory def ignore_file_and_directory_unless(flag) when flag do flag && nil # bypass unused variable warning end def ignore_file_and_directory_unless(_) do throw :ignore_file_and_directory end # called from within a template to cause it not to generate this file def ignore_file_unless(flag) when flag do flag && nil # bypass unused variable warning end def ignore_file_unless(_) do throw :ignore_file end private do defp check_existence_of(dir, name) do path = Path.join(dir, name) cond do !File.exists?(dir) -> { :error, "target directory #{dir} does not exist" } !File.dir?(dir) -> { :error, "'#{dir}' is not a directory" } !File.exists?(path) -> { :need_to_create, path } !File.dir?(path) -> { :error, "'#{path}' exists but is not a directory" } true -> { :maybe_update, path } end end defp create_or_merge({:error, reason}, _, _), do: {:error, reason} defp create_or_merge({:need_to_create, dest_dir}, template, assigns) do source_dir = template.source_dir copy_tree_with_expansions(source_dir, dest_dir, assigns) end defp create_or_merge({:maybe_update, dest}, template, assigns) do if assigns[:assigns][:force] do copy_tree_with_expansions(template.source_dir, dest, assigns) else { :error, "Updating an existing project is not yet supported" } end end defp copy_tree_with_expansions(source, dest, assigns) do if File.dir?(source) do if !String.ends_with?(source, "_build") do copy_dir(source, dest, assigns) end else copy_and_expand(source, dest, assigns) end end defp copy_dir(source, dest, assigns) do maybe_create_directory(dest, assigns[:assigns][:force]) try do File.ls!(source) |> Enum.each(fn name -> s = Path.join(source, name) d = Path.join(dest, dest_file_name(name, assigns)) copy_tree_with_expansions(s, d, assigns) end) catch :ignore_file_and_directory -> File.rm_rf!(dest) Mix.shell.info([:green, "- deleting", :reset, " #{dest} ", :faint, :cyan, "(it isn't needed)"]) end end defp copy_and_expand(source, dest, assigns) do try do case generate_new_file(source, dest, assigns) do {:error, message} -> {:error, message} new_file -> MG.create_file(dest, new_file) mode = File.stat!(source).mode File.chmod!(dest, mode) end catch :ignore_file -> Mix.shell.info([:green, "- ignoring", :reset, " #{dest} ", :faint, :cyan, "(it isn't needed)"]) end end defp generate_new_file(source, dest, assigns) do just_files = assigns[:assigns][:just_files] if filename_matches_just_files(source, just_files) do File.read!(source) else evaluate_eex(source, dest, assigns) end end defp filename_matches_just_files(filename, just_files) when is_binary(just_files) do filename_matches_wildcard?(filename, just_files) end defp filename_matches_just_files(filename, just_files) when is_list(just_files) do just_files |> Enum.any?(fn exp -> filename_matches_wildcard?(filename, exp) end) end defp filename_matches_just_files(_filename, _), do: false defp filename_matches_wildcard?(filename, expression) do expression |> wildcard_to_regex |> Regex.match?(filename) end defp wildcard_to_regex(expression) do {:ok, regex} = expression |> String.split("*") |> Enum.map(&Regex.escape(&1)) |> Enum.join(".+") |> Regex.compile regex end defp evaluate_eex(source, dest, assigns) do try do EEx.eval_file(source, assigns, [ trim: true ]) rescue UnicodeConversionError -> Mix.shell.error([:red, "- ignoring", :reset, " #{dest} ", :faint, :red, "(unexpected binary file. ", "See the \"just_files\" option)"]) {:error, :unexpected_binary} _ -> Mix.shell.error([:red, "- ignoring", :reset, " #{dest} ", :faint, :red, "(unhandled error generating file)"]) {:error, :unhandled} end end defp mandatory_option(nil, msg), do: raise(CompileError, description: msg) defp mandatory_option(value, _msg), do: value # You can escape the project name by doubling the $ characters, # so $$PROJECT_NAME$$ becomes $PROJECT_NAME$ defp dest_file_name(name, assigns) do if name =~ ~r{\$\$PROJECT_NAME\$\$} do String.replace(name,"$$PROJECT_NAME$$", "$PROJECT_NAME$") else String.replace(name, "$PROJECT_NAME$", assigns[:assigns][:project_name]) end end defp maybe_create_directory(path, force) when not force do MG.create_directory(path) end defp maybe_create_directory(path, _force) do if File.exists?(path) do Mix.shell.info([:green, "• existing #{path}"]) else maybe_create_directory(path, false) end end end end