monkrb 0.20.0 → 0.21.0

data/lib/monk/scaffold.rb

@@ -1,1063 +0,0 @@
-require "fileutils"
-
-require_relative "errors"
-
-module Monk
-  # Writes a new Monk project's skeleton to disk. Templates are static
-  # files, copied verbatim -- nothing here needs the project's own name
-  # substituted in, so there's no templating engine involved. See
-  # docs/history/plan-init.md for the full design.
-  class Scaffold
-    TEMPLATES_DIR = File.expand_path("templates", __dir__)
-
-    # The role directories under app/, in config/load.rb's load order.
-    APP_ROLES = %w[models presenters helpers mailers broadcasts jobs].freeze
-
-    BASE_FILES = {
-      "Gemfile" => "base/Gemfile",
-      "config.ru" => "base/config.ru",
-      "config/settings.rb" => "base/config/settings.rb",
-      "config/load.rb" => "base/config/load.rb",
-      "app/app.rb" => "base/app/app.rb",
-      ".ruby-version" => "base/.ruby-version",
-      ".gitignore" => "base/.gitignore",
-      ".dockerignore" => "base/.dockerignore",
-      "Dockerfile" => "base/Dockerfile",
-      "bin/server" => "base/bin/server",
-      "bin/websocket_server" => "base/bin/websocket_server",
-      "app/views/layouts/app.erb" => "base/app/views/layouts/app.erb",
-      "app/views/index.erb" => "base/app/views/index.erb",
-      # Every role directory, whatever the flags (docs/adr/0015): the tree
-      # shows where each kind of code goes. Flags add files next to these.
-      **APP_ROLES.to_h { |role| ["app/#{role}/.keep", "base/app/#{role}/.keep"] },
-      "public/css/app.css" => "base/public/css/app.css",
-      "public/js/app.js" => "base/public/js/app.js",
-    }.freeze
-
-    POSTGRES_FILES = {
-      "config/persistence.rb" => "postgres/config/persistence.rb",
-      "bin/console" => "postgres/bin/console",
-      "bin/setup_db" => "postgres/bin/setup_db",
-      "bin/migrate" => "postgres/bin/migrate",
-      # Kept by git when empty, like app/'s role directories: a fresh clone
-      # still shows where migrations go. The migrator reads only *.sql.
-      "db/migrate/.keep" => "postgres/db/migrate/.keep",
-    }.freeze
-
-    # The pg gem's native extension needs libpq -- the base Dockerfile has
-    # no system packages at all (kino is a precompiled platform gem), so
-    # --postgres/--auth swap in a Dockerfile that adds libpq-dev/libpq5
-    # instead of patching one image for every combination.
-    POSTGRES_OVERRIDES = {
-      "Dockerfile" => "postgres/Dockerfile",
-    }.freeze
-
-    AUTH_FILES = {
-      "config/auth.rb" => "auth/config/auth.rb",
-      "db/migrate/00000000000001_create_auth_tables.up.sql" => "auth/db/migrate/00000000000001_create_auth_tables.up.sql",
-      "db/migrate/00000000000001_create_auth_tables.down.sql" => "auth/db/migrate/00000000000001_create_auth_tables.down.sql",
-      # The magic link's sender, Monk::Auth's deliver: hook -- config/auth.rb
-      # requires it -- and the HTML part it renders and sends through
-      # Monk::Mail (MAIL_FILES, implied by --auth).
-      "app/mailers/app_mailer.rb" => "auth/app/mailers/app_mailer.rb",
-      "app/views/mail/magic_link.erb" => "auth/app/views/mail/magic_link.erb",
-    }.freeze
-
-    # --mail, or implied by --auth: Monk::Mail's config. Its Gemfile line
-    # (net-smtp) and MAIL_* env vars are added by #write! / #write_env_files!.
-    MAIL_FILES = {
-      "config/mail.rb" => "mail/config/mail.rb",
-    }.freeze
-
-    # --jobs: Monk::Jobs. The migration is the canonical one under
-    # templates/jobs/db/migrate -- the same file Monk's own tests apply --
-    # numbered after auth's so the two sort in order when both are present.
-    JOBS_FILES = {
-      "config/jobs.rb" => "jobs/config/jobs.rb",
-      "app/jobs/hello_job.rb" => "jobs/app/jobs/hello_job.rb",
-      "bin/jobs" => "jobs/bin/jobs",
-      "db/migrate/00000000000002_create_jobs_tables.up.sql" => "jobs/db/migrate/00000000000002_create_jobs_tables.up.sql",
-      "db/migrate/00000000000002_create_jobs_tables.down.sql" =>
-        "jobs/db/migrate/00000000000002_create_jobs_tables.down.sql",
-    }.freeze
-
-    # --auth --jobs: the magic link is created and sent inside a job, so the
-    # raw token is never stored (docs/adr/0014).
-    JOBS_AUTH_FILES = {
-      "app/jobs/send_login_link.rb" => "jobs/app/jobs/send_login_link.rb",
-    }.freeze
-
-    # --mail --jobs (or --auth --jobs): Monk::Mail.deliver_later, loaded in
-    # config/jobs.rb right after Monk::Jobs itself.
-    JOBS_REQUIRE_ANCHOR = %(require "monk/jobs"\n).freeze
-    JOBS_MAIL_REQUIRE = %(require "monk/mail/later"\n).freeze
-
-    # --jobs's demo route, added right after this line in app/app.rb (the
-    # base and the --live one both end their routes with it), so the
-    # round trip -- enqueue from a request, run in bin/jobs -- works out of
-    # the box.
-    JOBS_ROUTE_ANCHOR = %(  get("/api/hello") { json(message: "hello from monk") }\n).freeze
-    JOBS_ROUTE = <<~RUBY.gsub(/^(?!$)/, "  ").freeze
-
-      # Enqueues the demo job (app/jobs/hello_job.rb) for bin/jobs to run:
-      #   curl -X POST "http://localhost:9292/jobs/hello?name=Ann"
-      post("/jobs/hello") { json(enqueued: HelloJob.enqueue(params[:name] || "world")) }
-    RUBY
-
-    # --live: Monk::Live's demo (a counter whose open tabs update when
-    # another request changes it). These replace three base files outright
-    # rather than patching them -- a live app and index are different
-    # files, not one line of diff -- and add two. config.ru stays the base
-    # one: config/load.rb is where --live's config gets wired in.
-    LIVE_OVERRIDES = {
-      "app/app.rb" => "live/app/app.rb",
-      "bin/websocket_server" => "live/bin/websocket_server",
-      "app/views/index.erb" => "live/app/views/index.erb",
-    }.freeze
-
-    # config/live.rb itself isn't here -- #write_live! picks between
-    # LIVE_CONFIG_TEMPLATES[@live_transport] instead, since which one gets
-    # written depends on --redis vs --postgres, not a fixed mapping.
-    LIVE_FILES = {
-      "app/views/live/_hits.erb" => "live/app/views/live/_hits.erb",
-    }.freeze
-
-    LIVE_CONFIG_TEMPLATES = {
-      redis: "live/config/live.rb",
-      postgres: "live/config/live_pg.rb",
-    }.freeze
-
-    # Where the client runtime lands in the app's public root (an app serves
-    # it like any other static file; its files import each other by relative
-    # path, so they stay together in one directory).
-    LIVE_CLIENT_DIR = "public/js/monk_live".freeze
-
-    LIVE_HEAD = <<~HTML.gsub(/^/, "    ").freeze
-      <meta name="monk-live-url" content="<%= settings[:live_ws_url] %>">
-      <script type="module" src="<%= asset_path "/js/monk_live/monk_live.js" %>"></script>
-    HTML
-
-    EXECUTABLE_FILES = %w[bin/server bin/websocket_server bin/console bin/setup_db bin/migrate bin/jobs].freeze
-
-    # auth: implies postgres -- Monk::Auth is Postgres-only (lib/monk/auth.rb
-    # subclasses Monk::Persistence::Pg::Model), so there's no combination
-    # where an app gets Auth without also getting the persistence scaffold.
-    #
-    # redis: is unrelated to both -- it doesn't imply, and isn't implied by,
-    # postgres or auth. Unlike them it adds no files of its own, just a
-    # Gemfile line: bin/websocket_server (always present, see BASE_FILES
-    # above -- WebSocket needs no external service, so unlike Postgres/Redis
-    # it isn't gated behind a flag at all) checks ENV["REDIS_URL"] at boot
-    # and only then requires "redis", so the gem has to already be in the
-    # bundle for that to work.
-    #
-    # live: bin/server and bin/websocket_server are always separate
-    # processes, so *something* has to carry a publish between them --
-    # --redis (Monk::WebSocket::RedisFanout) or --postgres
-    # (Monk::WebSocket::PgFanout, reusing --postgres's own DB_* settings,
-    # no Redis to run). --live no longer silently defaults to one: with
-    # neither flag, there's no way to tell whether this app has Postgres
-    # available at all, so guessing would be as likely to hand back a
-    # template that can't connect as one that can. --live --redis wins
-    # over --postgres if both are passed (an explicit ask beats an
-    # implied default); --live --postgres alone (or via --auth, which
-    # implies --postgres) picks PgFanout; --live with neither raises.
-    # See docs/design/live-pg-fanout.md / docs/history/plan-live-pg-fanout.md
-    # for the design this decision comes from.
-    #
-    # mail: Monk::Mail on its own, no Postgres needed. auth: implies it, the
-    # same way it implies postgres: a magic link has to reach someone.
-    #
-    # jobs: implies postgres, since the queue lives there, and nothing else.
-    def initialize(dir, postgres: false, auth: false, redis: false, live: false, mail: false, jobs: false)
-      @dir = dir
-      # As passed, before anything is implied -- #summary reports the difference.
-      @requested = { postgres: postgres, auth: auth, mail: mail, jobs: jobs, redis: redis, live: live }
-      @auth = auth
-      @jobs = jobs
-      @postgres = postgres || auth || jobs
-      @mail = mail || auth
-      @live = live
-
-      if @live
-        @redis = redis || false
-        @live_transport = if redis
-          :redis
-        elsif @postgres
-          :postgres
-        else
-          raise Monk::AmbiguousLiveTransportError,
-            "monk new --live needs --redis or --postgres to pick Monk::Live's cross-process " \
-            "transport (bin/server and bin/websocket_server are always separate processes) -- " \
-            "pass one explicitly, e.g. `--live --postgres` (docs/guides/live.md's " \
-            "\"Without Redis\" section has the tradeoffs)"
-        end
-      else
-        @redis = redis
-      end
-    end
-
-    # The flags as resolved, for `monk new` to print: which flags were
-    # given, which ones they implied and why, and which combinations change
-    # what gets generated. The rule behind it, stated in `monk --help`: a
-    # flag is implied when there's only one right answer (--auth can only
-    # use Postgres), and required when there's a real choice (--live's
-    # transport).
-    def summary
-      given = FLAG_ORDER.select { |flag| @requested[flag] }
-      return ["Flags: none (the base skeleton)."] if given.empty?
-
-      implied = implied_flags
-      flags = given.map { |flag| "--#{flag}" }.join(" ")
-      lines = implied.empty? ? ["Flags: #{flags}."] : ["Flags: #{flags}, which also turned on:"]
-      implied.each do |flag, sources|
-        lines << "  #{"--#{flag}".ljust(12)}(needed by #{sources.map { |source| "--#{source}" }.join(", ")})"
-      end
-
-      together = combinations
-      lines << "Together they also generate:" unless together.empty?
-      width = together.map { |pair, _| pair.length }.max
-      lines.concat(together.map { |pair, what| "  #{pair.ljust(width)}  #{what}" })
-    end
-
-    def write!
-      raise Monk::ScaffoldExistsError, "#{@dir} already exists" if File.exist?(@dir)
-
-      FileUtils.mkdir_p(@dir)
-      base_files = BASE_FILES
-      base_files = base_files.merge(LIVE_OVERRIDES) if @live
-      base_files = base_files.merge(POSTGRES_OVERRIDES) if @postgres
-      base_files.each { |relative, template| write_file(relative, template, executable: EXECUTABLE_FILES.include?(relative)) }
-      write_live! if @live
-
-      if @postgres
-        POSTGRES_FILES.each { |relative, template| write_file(relative, template, executable: EXECUTABLE_FILES.include?(relative)) }
-        append_gemfile_extra("postgres/Gemfile.extra")
-      end
-
-      if @mail
-        MAIL_FILES.each { |relative, template| write_file(relative, template) }
-        append_gemfile_extra("mail/Gemfile.extra")
-      end
-
-      if @jobs
-        JOBS_FILES.each { |relative, template| write_file(relative, template, executable: EXECUTABLE_FILES.include?(relative)) }
-        JOBS_AUTH_FILES.each { |relative, template| write_file(relative, template) } if @auth
-        add_jobs_route!
-        add_deliver_later! if @mail
-      end
-
-      wire_load! if @postgres || @mail || @live
-      append_gemfile_extra("redis/Gemfile.extra") if @redis
-
-      # --postgres, --redis and --mail are the flags with anything worth
-      # putting in an env file (a database, a REDIS_URL, MAIL_*) -- the
-      # base skeleton alone has nothing to configure this way.
-      if @postgres || @redis || @mail
-        write_env_files!
-        uncomment_dotenv!
-      end
-
-      AUTH_FILES.each { |relative, template| write_file(relative, template) } if @auth
-
-      # Every combination of flags gets a SETUP.md, not just --postgres --
-      # even the base skeleton has a dev step (bin/server) and an
-      # unscaffolded test framework to set up.
-      write_setup_md!
-    end
-
-    FLAG_ORDER = %i[postgres auth mail jobs redis live].freeze
-
-    # Only these two imply anything, and only one flag each can need.
-    IMPLIED_BY = { postgres: %i[auth jobs], mail: %i[auth] }.freeze
-
-    private
-
-    def implied_flags
-      resolved = { postgres: @postgres, mail: @mail }
-      IMPLIED_BY.filter_map do |flag, sources|
-        next if @requested[flag] || !resolved[flag]
-
-        [flag, sources.select { |source| @requested[source] }]
-      end
-    end
-
-    def combinations
-      pairs = []
-      pairs << ["--auth + --jobs", "app/jobs/send_login_link.rb: login links are sent from a job"] if @auth && @jobs
-      if @mail && @jobs
-        pairs << ["--mail + --jobs",
-                  "config/jobs.rb loads Monk::Mail.deliver_later; JOBS_QUEUES serves mailers first",]
-      end
-      pairs << live_combination if @live
-      pairs
-    end
-
-    def live_combination
-      if @live_transport == :redis
-        both = @postgres ? "; --redis wins over --postgres" : ""
-        ["--live + --redis", "config/live.rb fans out over Redis (Monk::WebSocket::RedisFanout#{both})"]
-      else
-        ["--live + --postgres", "config/live.rb fans out over Postgres (Monk::WebSocket::PgFanout)"]
-      end
-    end
-
-    def write_live!
-      write_file("config/live.rb", LIVE_CONFIG_TEMPLATES.fetch(@live_transport))
-      LIVE_FILES.each { |relative, template| write_file(relative, template) }
-      copy_live_client!
-      add_live_to_layout!
-    end
-
-    # The runtime ships inside the gem (Monk::Live.client_dir); copied, not
-    # duplicated under templates/, so there is one source of truth.
-    def copy_live_client!
-      source = File.join(__dir__, "live/client")
-      target = File.join(@dir, LIVE_CLIENT_DIR)
-      FileUtils.mkdir_p(target)
-      FileUtils.cp(Dir.children(source).map { |name| File.join(source, name) }, target)
-    end
-
-    def add_live_to_layout!
-      path = File.join(@dir, "app/views/layouts/app.erb")
-      content = File.read(path)
-      raise "layout wiring failed: </head> not found" unless content.include?("  </head>\n")
-
-      File.write(path, content.sub("  </head>\n", "#{LIVE_HEAD}  </head>\n"))
-    end
-
-    def append_gemfile_extra(template_path)
-      extra = File.read(File.join(TEMPLATES_DIR, template_path))
-      File.write(File.join(@dir, "Gemfile"), "\n#{extra}", mode: "a")
-    end
-
-    # --postgres/--redis write a real .env (see write_env_files!) -- if the
-    # dotenv gem stays commented out, as it does in the base skeleton,
-    # config/settings.rb's `require "dotenv/load"` never runs, so nothing
-    # ever actually loads them: bin/setup_db (or bin/websocket_server's
-    # REDIS_URL check) silently falls back to config/persistence.rb's own
-    # ENV.fetch defaults, or no REDIS_URL at all, instead of the
-    # app-specific values .env was written to provide.
-    DOTENV_COMMENTED_LINE = %(# gem "dotenv" # uncomment to load a local .env file (config/settings.rb)\n).freeze
-    DOTENV_LINE = %(gem "dotenv" # loads .env/.env.test -- see config/settings.rb\n).freeze
-
-    def uncomment_dotenv!
-      path = File.join(@dir, "Gemfile")
-      content = File.read(path)
-      raise "Gemfile wiring failed: #{DOTENV_COMMENTED_LINE.inspect} not found" unless content.include?(DOTENV_COMMENTED_LINE)
-
-      File.write(path, content.sub(DOTENV_COMMENTED_LINE, DOTENV_LINE))
-    end
-
-    # config/load.rb ships in BASE_FILES unconditionally (config.ru and the
-    # tests require it), so the flags can't gate whether it exists, only
-    # which module configs it requires. Unlike BASE_FILES/POSTGRES_FILES/
-    # AUTH_FILES, this is a post-write edit rather than a verbatim copy --
-    # the alternative (a second, fuller load.rb template per flag
-    # combination) would duplicate the whole file for a few lines of diff.
-    def wire_load!
-      path = File.join(@dir, "config/load.rb")
-      settings_require = %(require_relative "settings"\n)
-      requires = +""
-      if @postgres
-        target = @auth ? "auth" : "persistence" # config/auth.rb itself require_relative "persistence"
-        requires << "require_relative \"#{target}\"\n"
-      end
-      # config/mail.rb from config/load.rb only, never from config/auth.rb:
-      # bin/websocket_server loads config/auth.rb too and never sends mail,
-      # so it shouldn't need MAIL_URL to boot.
-      requires << "require_relative \"mail\"\n" if @mail
-      requires << "require_relative \"jobs\"\n" if @jobs
-      requires << "require_relative \"live\"\n" if @live
-
-      content = File.read(path)
-      raise "config/load.rb wiring failed: #{settings_require.inspect} not found" unless content.include?(settings_require)
-
-      File.write(path, content.sub(settings_require, "#{settings_require}#{requires}"))
-    end
-
-    def add_jobs_route!
-      path = File.join(@dir, "app/app.rb")
-      content = File.read(path)
-      raise "app/app.rb wiring failed: #{JOBS_ROUTE_ANCHOR.inspect} not found" unless content.include?(JOBS_ROUTE_ANCHOR)
-
-      File.write(path, content.sub(JOBS_ROUTE_ANCHOR, "#{JOBS_ROUTE_ANCHOR}#{JOBS_ROUTE}"))
-    end
-
-    def add_deliver_later!
-      path = File.join(@dir, "config/jobs.rb")
-      content = File.read(path)
-      raise "config/jobs.rb wiring failed: #{JOBS_REQUIRE_ANCHOR.inspect} not found" unless content.include?(JOBS_REQUIRE_ANCHOR)
-
-      File.write(path, content.sub(JOBS_REQUIRE_ANCHOR, "#{JOBS_REQUIRE_ANCHOR}#{JOBS_MAIL_REQUIRE}"))
-    end
-
-    # .env/.env.test/.env.example are the other deliberate exception to
-    # "templates are static files, copied verbatim" (see the class comment
-    # above): DB_NAME needs this app's own directory name in it -- the one
-    # piece of scaffold-wide content that's actually per-project -- so
-    # these are composed here instead of read from disk. .env/.env.test are
-    # gitignored (see base/.gitignore); .env.example is the tracked
-    # placeholder that file explicitly carves out.
-    def write_env_files!
-      app_name = File.basename(@dir)
-
-      dev = @postgres ? pg_env_lines("#{app_name}_development") : []
-      test = @postgres ? pg_env_lines("#{app_name}_test") : []
-      example = @postgres ? pg_env_lines("#{app_name}_development") : []
-
-      if @auth
-        dev << "AUTH_SECRET=change-me-dev-secret"
-        test << "AUTH_SECRET=change-me-test-secret"
-        example << "AUTH_SECRET=change-me"
-      end
-
-      if @mail
-        # MAIL_URL is left out of .env on purpose: unset in development
-        # means log://, the message printed to the console. Tests boot
-        # outside development, where an unset one raises, hence log:// there.
-        dev << %(MAIL_FROM="#{app_name} <no-reply@localhost>")
-        test << "MAIL_URL=log://"
-        test << %(MAIL_FROM="#{app_name} <no-reply@localhost>")
-        example << "MAIL_URL=smtp://user:password@smtp.example.com:587"
-        example << %(MAIL_FROM="#{app_name} <no-reply@example.com>")
-      end
-
-      # bin/jobs's settings. Not in .env.test: tests run jobs with
-      # Monk::Jobs.drain!, never a job process.
-      # With mail, the mailers queue is served first, so a backlog of other
-      # work never delays a login email (docs/adr/0014).
-      if @jobs
-        queues = @mail ? "mailers,default" : "default"
-        [dev, example].each { |lines| lines.push("JOBS_WORKERS=2", "JOBS_QUEUES=#{queues}") }
-      end
-
-      # REDIS_URL is absent from .env.test for --redis alone -- only a test
-      # that actually exercises RedisFanout needs it. --live --redis is
-      # different: config/live.rb raises without it, and the test helper
-      # loads config/live.rb through config/load.rb. Building the fanout
-      # connects to nothing, so tests still need no Redis running unless
-      # one publishes.
-      if @redis
-        dev << "REDIS_URL=redis://localhost:6379/0"
-        example << "REDIS_URL=redis://localhost:6379/0"
-        test << "REDIS_URL=redis://localhost:6379/0" if live_over_redis?
-      end
-
-      write_lines(".env", dev)
-      write_lines(".env.example", example)
-      # --redis alone (no --postgres) has nothing to put in .env.test --
-      # skip it rather than write an empty, pointless file.
-      write_lines(".env.test", test) unless test.empty?
-    end
-
-    def pg_env_lines(dbname)
-      [
-        "DB_HOST=127.0.0.1",
-        "DB_PORT=5432",
-        "DB_USER=postgres",
-        "DB_PASSWORD=postgres",
-        "DB_NAME=#{dbname}",
-      ]
-    end
-
-    def write_lines(relative_path, lines)
-      File.write(File.join(@dir, relative_path), "#{lines.join("\n")}\n")
-    end
-
-    def write_setup_md!
-      File.write(File.join(@dir, "SETUP.md"), setup_md_content)
-      File.write(File.join(@dir, "SETUP.md"), live_setup_md_content, mode: "a") if @live
-    end
-
-    def live_setup_md_content
-      @live_transport == :postgres ? live_setup_md_content_postgres : live_setup_md_content_redis
-    end
-
-    def live_setup_md_content_redis
-      <<~MARKDOWN
-
-        ## Live updates (Monk::Live)
-
-        `monk new --live --redis` wired a demo: a counter whose open tabs
-        update by themselves. It needs three things running, each in its own
-        terminal:
-
-        ```bash
-        docker run --rm -d -p 6379:6379 --name #{File.basename(@dir)}_redis redis:7   # skip if one is already up
-        bin/server              # the HTTP app on :9292, which publishes updates
-        bin/websocket_server    # the sockets on :9293, which deliver them
-        ```
-
-        Open http://localhost:9292 in two tabs and press the button in one.
-
-        `bin/server` and `bin/websocket_server` are separate processes, so Redis
-        (`REDIS_URL`, already in `.env`) is what carries an update from one to
-        the other. Where things are:
-
-        - `config/live.rb` -- the Redis wiring (`Monk::WebSocket::RedisFanout`)
-          and the subscribe rules (nothing is allowed unless a rule says so).
-        - `app/app.rb` -- `POST /hit` changes state and calls `Monk::Live.patch`.
-        - `app/views/index.erb` -- `live_topic "hits"` marks what to subscribe to.
-        - `app/views/live/_hits.erb` -- the fragment that gets pushed. Partials
-          used this way see only their locals (`locals[:hits]`), never
-          `params` or the session.
-        - `bin/websocket_server` -- calls `Monk::Live.listen!` at boot: the
-          only process that listens for updates. `bin/server`#{@jobs ? " and `bin/jobs`" : ""} only
-          publish.
-        - `public/js/monk_live/` -- the browser runtime, copied from the gem.
-          The layout points at it and at `LIVE_WS_URL` (default
-          `ws://localhost:9293`; use `wss://` in production).
-
-        `WS_ALLOWED_ORIGINS` (default `http://localhost:9292`) must list the
-        origin your pages are served from, or the browser's socket is refused.
-        #{live_auth_setup_paragraph}
-      MARKDOWN
-    end
-
-    def live_setup_md_content_postgres
-      <<~MARKDOWN
-
-        ## Live updates (Monk::Live)
-
-        `monk new --live --postgres` wired a demo: a counter whose open tabs
-        update by themselves. It needs two things running, each in its own
-        terminal -- no Redis, since Postgres `LISTEN`/`NOTIFY` carries the
-        update instead:
-
-        ```bash
-        bin/server              # the HTTP app on :9292, which publishes updates
-        bin/websocket_server    # the sockets on :9293, which deliver them
-        ```
-
-        Open http://localhost:9292 in two tabs and press the button in one.
-
-        `bin/server` and `bin/websocket_server` are separate processes, so
-        Postgres `LISTEN`/`NOTIFY` (the same `DB_*` settings already in `.env`)
-        is what carries an update from one to the other. Where things are:
-
-        - `config/live.rb` -- the Postgres wiring (`Monk::WebSocket::PgFanout`)
-          and the subscribe rules (nothing is allowed unless a rule says so).
-        - `app/app.rb` -- `POST /hit` changes state and calls `Monk::Live.patch`.
-        - `app/views/index.erb` -- `live_topic "hits"` marks what to subscribe to.
-        - `app/views/live/_hits.erb` -- the fragment that gets pushed. Partials
-          used this way see only their locals (`locals[:hits]`), never
-          `params` or the session.
-        - `bin/websocket_server` -- calls `Monk::Live.listen!` at boot: the
-          only process that listens for updates. `bin/server`#{@jobs ? " and `bin/jobs`" : ""} only
-          publish.
-        - `public/js/monk_live/` -- the browser runtime, copied from the gem.
-          The layout points at it and at `LIVE_WS_URL` (default
-          `ws://localhost:9293`; use `wss://` in production).
-
-        `WS_ALLOWED_ORIGINS` (default `http://localhost:9292`) must list the
-        origin your pages are served from, or the browser's socket is refused.
-        #{live_auth_setup_paragraph}
-        A single broadcast is capped at just under 8000 bytes
-        (`Monk::WebSocket::PgFanout::MAX_NOTIFY_PAYLOAD_BYTES`, Postgres's own
-        `NOTIFY` limit) -- far more than this demo ever sends, but worth
-        knowing before pushing much larger partials this way. Regenerate with
-        `--redis` instead of `--postgres` if that becomes a real constraint.
-      MARKDOWN
-    end
-
-    def setup_md_content
-      @postgres ? postgres_setup_md_content : base_setup_md_content
-    end
-
-    # No Postgres, so no database/containers/migrations -- but still a
-    # real dev-then-test walkthrough: bin/server needs nothing external,
-    # and "no test framework scaffolded" is just as true here as it is
-    # with --postgres.
-    def base_setup_md_content
-      app_name = File.basename(@dir)
-
-      <<~MARKDOWN
-        # Setting up #{app_name} (dev, then test)
-
-        #{base_setup_md_intro(app_name)}
-
-        ## Dev environment, first run
-
-        ```bash
-        bundle install
-        bin/server              # HTTP app on :9292 -> http://localhost:9292/hello
-        bin/websocket_server    # WS chat process on :9293, in another terminal
-        ```
-        #{redis_only_note}#{mail_setup_note}
-        ## Test environment
-
-        `monk new` scaffolds no test framework at all -- this is the minimum to
-        get `bundle exec rake test` working, using Minitest (matches `monk`'s
-        own suite):
-
-        **Gemfile** -- add:
-
-        ```ruby
-        group :test do
-          gem "minitest"
-          gem "rake"
-        end
-        ```
-
-        **test/test_helper.rb**:
-
-        ```ruby
-        $LOAD_PATH.unshift(File.expand_path("..", __dir__))
-
-        ENV["MONK_ENV"] ||= "test"
-        #{base_test_dotenv_lines}
-        #{test_helper_app_lines}
-        ```
-
-        **Rakefile**:
-
-        ```ruby
-        require "rake/testtask"
-
-        Rake::TestTask.new do |t|
-          t.libs << "test"
-          t.pattern = "test/**/*_test.rb"
-        end
-
-        task default: :test
-        ```
-
-        #{app_test_sample}
-
-        Then:
-
-        ```bash
-        bundle install
-        bundle exec rake test
-        ```
-      MARKDOWN
-    end
-
-    def base_setup_md_intro(app_name)
-      unless @redis
-        return "No external services are scaffolded here (`monk new #{app_name}`, no " \
-          "`--postgres`/`--auth`/`--redis`) -- `bin/server`/`bin/websocket_server` need " \
-          "nothing external to run."
-      end
-
-      "No database is scaffolded here (`monk new #{app_name} --redis`, no `--postgres`/" \
-        "`--auth`) -- `bin/server` needs nothing external, but `bin/websocket_server`'s " \
-        "cross-process fan-out needs a reachable Redis once `REDIS_URL` (already set in " \
-        "the generated `.env`) is visible to it."
-    end
-
-    def redis_only_note
-      return "" unless @redis
-
-      app_name = File.basename(@dir)
-      "\n`.env` already sets `REDIS_URL=redis://localhost:6379/0`, turning on " \
-        "`bin/websocket_server`'s cross-process fan-out -- unset (or missing entirely), it " \
-        "runs in-process only. Check `docker ps` first in case a Redis container is already " \
-        "running from another project; otherwise start one: `docker run --rm -d -p 6379:6379 " \
-        "--name #{app_name}_redis redis:7`.\n"
-    end
-
-    def postgres_setup_md_content
-      app_name = File.basename(@dir)
-
-      <<~MARKDOWN
-        # Setting up #{app_name} (dev, then test)
-
-        `config/load.rb`, `.env`, and `.env.test` are already wired up by `monk new`
-        -- `config/load.rb` requires `config/#{@auth ? "auth" : "persistence"}` (and every other config
-        you enabled), then loads `app/`, and
-        `.env`/`.env.test` are pre-filled with a database name derived from
-        this project's directory (`#{app_name}_development` / `#{app_name}_test`), not the
-        generic `app_development` fallback baked into `config/persistence.rb`
-        itself. Adjust `DB_HOST`/`DB_USER`/`DB_PASSWORD` in both files if your
-        local Postgres doesn't use the `postgres`/`postgres` defaults.
-
-        `MONK_ENV` and the database name are two separate, unlinked knobs --
-        setting `MONK_ENV=test` does not by itself change which database you
-        connect to. Only `DB_NAME` (here, via `.env`/`.env.test`) does that.
-
-        ## Dev environment, first run
-
-        ```bash
-        bundle install
-        ```
-
-        ### 1. Start Postgres#{@redis ? " and Redis" : ""}
-
-        Check first whether a Postgres#{@redis ? "/Redis" : ""} container is already running from another
-        project -- starting a second one bound to the same host port fails with
-        `port is already allocated`:
-
-        ```bash
-        docker ps
-        ```
-
-        If something's already listening on 5432#{@redis ? "/6379" : ""}, reuse it instead of starting a new
-        one -- point `.env`/`.env.test`'s `DB_HOST`/`DB_PORT`#{@redis ? "/`REDIS_URL`" : ""} at it, and use its
-        actual `DB_USER`/`DB_PASSWORD` rather than the generated defaults.
-        Otherwise, start #{@redis ? "fresh containers" : "a fresh container"}:
-
-        ```bash
-        docker run --rm -d -p 5432:5432 -e POSTGRES_PASSWORD=postgres --name #{app_name}_pg postgres:16
-        #{"docker run --rm -d -p 6379:6379 --name #{app_name}_redis redis:7\n" if @redis}```
-
-        ### 2. Create the dev database
-
-        Plain `createdb` connects over the local Unix socket by default, which a
-        Docker container never provides -- use `-h`/`-p` to force a TCP
-        connection, or run `createdb` from inside the container instead to avoid
-        passing a password on the command line:
-
-        ```bash
-        # via TCP (adjust -h/-p/-U to match whatever's actually running):
-        PGPASSWORD=postgres createdb -h 127.0.0.1 -p 5432 -U postgres #{app_name}_development
-
-        # or, from inside an existing Postgres container, as the postgres user:
-        docker exec -u postgres <container_name> createdb #{app_name}_development
-        ```
-
-        ### 3. Run migrations
-
-        ```bash
-        bin/setup_db
-        ```
-
-        ### 4. Run it
-
-        ```bash
-        bin/server              # HTTP app on :9292
-        bin/websocket_server    # WS chat process on :9293, in another terminal
-        #{"bin/jobs                # background jobs (app/jobs/), in a third terminal\n" if @jobs}```
-        #{auth_or_redis_confirmation_note}#{mail_setup_note}#{jobs_setup_note}
-        ## Test environment
-
-        ### 1. Create the test database
-
-        ```bash
-        PGPASSWORD=postgres createdb -h 127.0.0.1 -p 5432 -U postgres #{app_name}_test
-
-        # or, from inside an existing Postgres container, as the postgres user:
-        docker exec -u postgres <container_name> createdb #{app_name}_test
-        ```
-
-        ### 2. Migrate the test database
-
-        ```bash
-        DB_NAME=#{app_name}_test bin/setup_db
-        ```
-
-        ### 3. Add a test framework and run it
-
-        `monk new` scaffolds no test framework at all -- this is the minimum to
-        get `bundle exec rake test` working, using Minitest (matches `monk`'s
-        own suite):
-
-        **Gemfile** -- add:
-
-        ```ruby
-        group :test do
-          gem "minitest"
-          gem "rake"
-        end
-        ```
-
-        **test/test_helper.rb** -- loads `.env.test` explicitly (not the default
-        dotenv-in-`config/settings.rb` path, which only loads plain `.env`), then
-        loads and boots the app the same way `config.ru` does:
-
-        ```ruby
-        $LOAD_PATH.unshift(File.expand_path("..", __dir__))
-
-        ENV["MONK_ENV"] ||= "test"
-
-        require "dotenv"
-        Dotenv.load(File.expand_path(".env.test", __dir__ + "/.."))
-
-        #{test_helper_app_lines}
-        ```
-
-        **Rakefile**:
-
-        ```ruby
-        require "rake/testtask"
-
-        Rake::TestTask.new do |t|
-          t.libs << "test"
-          t.pattern = "test/**/*_test.rb"
-        end
-
-        task default: :test
-        ```
-
-        **test/persistence_test.rb** -- a real smoke test, not just a
-        connectivity check:
-
-        ```ruby
-        require_relative "test_helper"
-
-        class PersistenceTest < Minitest::Test
-        #{sample_test_body}
-        end
-        ```
-
-        #{app_test_sample}
-        #{jobs_test_sample}
-        Then:
-
-        ```bash
-        bundle install
-        bundle exec rake test
-        ```
-
-        `test/test_helper.rb` sets `MONK_ENV=test` itself (so `Monk.env.test?`
-        reads correctly if the app ever branches on it) and loads `.env.test` for
-        the actual connection details -- but per the note at the top, those are
-        two separate knobs: `MONK_ENV` doesn't affect `DB_NAME` on its own,
-        `.env.test`'s `DB_NAME=#{app_name}_test` is what actually points tests at
-        the right database.
-      MARKDOWN
-    end
-
-    def auth_or_redis_confirmation_note
-      return live_auth_confirmation_note if @live
-      return "" unless @auth || @redis
-
-      flags = [("authenticate: true" if @auth), ("redis fan-out: on" if @redis)].compact.join(", ")
-      visible = if @auth && @redis
-        "both `AUTH_SECRET` and `REDIS_URL` are"
-      else
-        @auth ? "`AUTH_SECRET` is" : "`REDIS_URL` is"
-      end
-
-      "\n`bin/websocket_server`'s startup line should print `#{flags}` once " \
-        "#{visible} visible to it -- that confirms everything's actually wired up.\n"
-    end
-
-    # Who gets which topics once there are users. Interpolated into a
-    # squiggly heredoc, so no indentation of its own.
-    def live_auth_setup_paragraph
-      return "" unless @auth
-
-      <<~MARKDOWN
-
-        **Visitors and logged-in users.** A logged-in user's socket carries
-        their identity (the session cookie reaches `:9293` too); a visitor's
-        comes in anonymous. Each topic's rule in `config/live.rb` decides:
-        `anonymous: true` opens it to visitors (the demo's `hits`), and a
-        rule without it is for logged-in users only, e.g. one topic per user.
-        `monk new` writes no login routes: see `docs/guides/auth.md` in the
-        monk repo, and `docs/guides/live.md`, "Who may subscribe".
-      MARKDOWN
-    end
-
-    # The --live bin/websocket_server prints only `authenticate:`, and with
-    # --auth it's :optional (a visitor's socket is anonymous, not refused).
-    def live_auth_confirmation_note
-      return "" unless @auth
-
-      "\n`bin/websocket_server`'s startup line should print `authenticate: optional` once " \
-        "`AUTH_SECRET` is visible to it: a logged-in user's socket carries their identity, and a " \
-        "visitor's comes in anonymous, so it gets only the topics `config/live.rb` opens with " \
-        "`anonymous: true`.\n"
-    end
-
-    # Interpolated into a squiggly heredoc, so no leading indentation of
-    # its own (interpolated text isn't dedented).
-    def mail_setup_note
-      return "" unless @mail
-
-      <<~MARKDOWN
-
-        ### Email
-
-        `config/mail.rb` configures `Monk::Mail`; send with
-        `Monk::Mail.deliver(to:, subject:, text:, html:)`.#{auth_mail_sentence} In
-        development `MAIL_URL` is unset, so messages are printed to the
-        console instead of sent. Outside
-        development set `MAIL_URL` (e.g. `smtp://user:pass@smtp.provider.com:587`)
-        and `MAIL_FROM` to a sender on a domain you've verified with your
-        provider -- an unset `MAIL_URL` fails the boot there. Provider
-        examples: `docs/guides/mail.md` in the monk repo.
-      MARKDOWN
-    end
-
-    # Interpolated into a squiggly heredoc, so no indentation of its own.
-    def jobs_setup_note
-      return "" unless @jobs
-
-      <<~MARKDOWN
-
-        ### Background jobs
-
-        `config/jobs.rb` configures `Monk::Jobs` on the app's own database --
-        its tables come from `db/migrate/00000000000002_create_jobs_tables`,
-        which `bin/setup_db` already applied. The job classes live in
-        `app/jobs/`, which `config/load.rb` loads. With `bin/server` and
-        `bin/jobs` both running, enqueue the demo job and watch it run:
-
-        ```bash
-        curl -X POST "http://localhost:9292/jobs/hello?name=Ann"
-        tail -f log/development.log   # ... INFO Hello, Ann, from a background job
-        ```
-
-        `JOBS_WORKERS` and `JOBS_QUEUES` (in `.env`) set how many worker Ractors
-        `bin/jobs` runs and which queues they take jobs from, in order. A job
-        may run more than once (its process was killed mid-job, say), so make
-        each one safe to repeat. `TERM` or Ctrl-C lets the jobs in hand finish
-        before `bin/jobs` exits.#{jobs_mail_note}#{jobs_auth_note}
-      MARKDOWN
-    end
-
-    # Interpolated into a squiggly heredoc, so no indentation of its own.
-    def jobs_mail_note
-      return "" unless @mail
-
-      <<~MARKDOWN.chomp
-
-        To send an email from a job instead of the request, use
-        `Monk::Mail.deliver_later` -- the same arguments as `deliver` (render
-        HTML first with `Monk::Mail.render`), plus `wait:`/`at:`, or `conn:`
-        to enqueue inside your own transaction. It goes on the `mailers`
-        queue, which `JOBS_QUEUES` serves first. Temporary failures are
-        retried; a refusal the mail server made final is not.
-      MARKDOWN
-    end
-
-    # Interpolated into a squiggly heredoc, so no indentation of its own.
-    def jobs_auth_note
-      return "" unless @auth
-
-      <<~MARKDOWN.chomp
-
-        Magic links go through a job too, `app/jobs/send_login_link.rb`: it
-        creates the token and sends the link inside the job, so the raw
-        token is never stored, not even in the queue. Your login route,
-        after its own per-email rate limit, just enqueues it:
-
-        ```ruby
-        post("/auth/request") do
-          SendLoginLink.enqueue(params[:email])
-          json(sent: true)
-        end
-        ```
-
-        Don't point `config/auth.rb`'s `deliver:` at `deliver_later`: that
-        would store the link, token included, in the queue until it's sent.
-      MARKDOWN
-    end
-
-    # Interpolated into a squiggly heredoc, so no indentation of its own.
-    def jobs_test_sample
-      return "" unless @jobs
-
-      <<~MARKDOWN
-
-        **test/jobs_test.rb** -- `Monk::Jobs.drain!` runs every enqueued job
-        right in the test (each in its own Ractor, as `bin/jobs` would, so a
-        job that only works outside one fails here too), and
-        `Monk::Jobs.clear!` empties the queue between tests:
-
-        ```ruby
-        require_relative "test_helper"
-
-        class JobsTest < Minitest::Test
-          def teardown
-            Monk::Jobs.clear!
-          end
-
-          def test_the_demo_job_runs
-            HelloJob.enqueue("test")
-
-            assert_equal 1, Monk::Jobs.drain!
-          end
-        end
-        ```
-      MARKDOWN
-    end
-
-    # The no-Postgres test helper normally has nothing to load from
-    # .env.test -- but with --mail it has MAIL_URL=log://, which tests need
-    # (they boot outside development, where an unset MAIL_URL raises), and
-    # with --live --redis it has REDIS_URL, which config/live.rb needs.
-    # Interpolated into a squiggly heredoc, so no indentation of its own.
-    def base_test_dotenv_lines
-      return "" unless @mail || live_over_redis?
-
-      %(\nrequire "dotenv"\nDotenv.load(File.expand_path(".env.test", __dir__ + "/.."))\n)
-    end
-
-    def auth_mail_sentence
-      return "" unless @auth
-
-      " `AppMailer::MAGIC_LINK` (`app/mailers/app_mailer.rb`) sends each login link through it " \
-        "(HTML part: `app/views/mail/magic_link.erb`)."
-    end
-
-    def sample_test_body
-      if @auth
-        <<~RUBY.chomp.gsub(/^/, "  ")
-          def test_connects_to_the_test_database_and_sees_the_auth_tables
-            Monk::Persistence::Pg.checkout(:primary) do |conn|
-              result = conn.exec("SELECT to_regclass('login_tokens') IS NOT NULL AS present")
-              assert_equal true, result[0]["present"]
-            end
-          end
-        RUBY
-      else
-        <<~RUBY.chomp.gsub(/^/, "  ")
-          def test_connects_to_the_test_database
-            Monk::Persistence::Pg.checkout(:primary) do |conn|
-              assert_equal 1, conn.exec("SELECT 1").getvalue(0, 0)
-            end
-          end
-        RUBY
-      end
-    end
-
-    def live_over_redis? = @live && @live_transport == :redis
-
-    # The end of every SETUP.md test helper: the app loaded the way
-    # config.ru loads it (config/load.rb, then app/app.rb), booted once.
-    # Interpolated into a squiggly heredoc, so no indentation of its own.
-    def test_helper_app_lines
-      <<~RUBY.chomp
-        require "minitest/autorun"
-        require_relative "../config/load"
-        require_relative "../app/app"
-
-        # Booted once, as config.ru does: request tests call APP.
-        APP = Monk.boot(App)
-      RUBY
-    end
-
-    # A request through the booted app. Every generated app, --live's
-    # included, has GET /hello.
-    def app_test_sample
-      <<~MARKDOWN.chomp
-        **test/app_test.rb** -- a request through the booted app:
-
-        ```ruby
-        require_relative "test_helper"
-        require "rack/mock_request"
-
-        class AppTest < Minitest::Test
-          def test_hello
-            status, _headers, body = APP.call(Rack::MockRequest.env_for("/hello"))
-
-            assert_equal 200, status
-            assert_equal "hello from monk", body.join
-          end
-        end
-        ```
-      MARKDOWN
-    end
-
-    def write_file(relative_path, template_path, executable: false)
-      destination = File.join(@dir, relative_path)
-      FileUtils.mkdir_p(File.dirname(destination))
-      FileUtils.cp(File.join(TEMPLATES_DIR, template_path), destination)
-      File.chmod(0o755, destination) if executable
-    end
-  end
-end