Commands

• |inputs|action|
• |task, +|add task if necessary|
• |task:, :task... +|fork, add if necessary|
• |task:..., :task, +|merge, 2x for sync_merge, add if necessary|
• |task, -|remove task (blank input, hide)|
• |task, up|updates configurations using name|
• |up|update page -- rebuild condensed workflow|

Currently the system uses:

  nodes[index][0]    # tasc
  nodes[index][1]%w  # args
  nodes[index][2]    # configs
  nodes[index][3]    # options

Args, configs, and options are separated in this way, to ensure that the final results are ordered consistently. Configs and options don't require a specific order, but the tasc specification always needs to be first.

(Note that the keys are provided as a hash, and so the representational order is scrambled by the time the cgi params reach the parser.)

Multiple entries to each sub index (0,1,2,3) do preserve representation order; all are concatenated into an argh:

nodes: 
  "0": 
    "0": 
    - tap:dump
    "1": []

    "2": 
    - --backup-dir
    - backup
    - --filter
    - ""
    - --audit
    - --date-format
    - "%Y-%m-%d %H:%M:%S"
    - --rollback-on-error
    - --date
    - --name
    - tap/tasks/dump

Which is then parsed by Parser into a schema:

- - name
  - args
  - configs
  - options

Cgi scripts could use a lazy attribute to specify non-default routing. Default route:

  [cgi/path/to/file.rb]

  # ::route :cgi (path/to/cgi)
  # ::route :cgi/:action (path/to/cgi/action)

Then you can also have a CGIController class to work like your standard controller. Action is a method. Call the method, use the instance variables in a template. Template named like cgi/action.

This will give compatibility with more established frameworks, and an easy port if desired. Additionally this will allow basic cgis, alternate template engines, testing, etc.

At this point:

  ::manifest declares task
  ::generator declares generator
  ::desc shortand for rap tasks
  ::route for cgi routes

Not too bad.

Reorganize:

Tap::Server is a type of task.

• call is call
• process takes a url and pretends it was sent to the server (use rack request)
• setup env default... server should require env (as it isn't required by tasks normally) and take the instance for app.root
• task can be run by a handler for a true server

Implement a route/proxy method. Update the parsing scheme:

      # /                      # => app
      # /non-existant          # => /app/non-existant
      # /env/key:              # => env/key:app
      # /env/key:non-existant  # => /env/key:app/non-existant

Cache key at earliest moment to the controller. Clear cache if in development mode. If people want something fancier they have to implement it themselves.

Update template engine so that:

  render thing, options

is the syntax. Determine the extension dynamically, and determine rendering engine from there.

Alternate implementation. Route everything as:

  /key/argv?query

Then make Controller a kind of Task. Every call the url is parsed and the resulting task processed.

  tasc = env.tasks.search(key)
  task, args = tasc.parse(argv)
  task.reconfigure query
  task.process(*args)

On the plus side this is very natural to write and suggests a subtle difference between this and the command line; no schema just a single task. It also has the benefit of being predictable.

On the down side, --help and other options (not configs) must be in the argv (a break in the consistency). --help itself will cause trouble as it exits by default.

For Rack 1.0:

<form method="post" action="<%= uri(:index) %>">
  <input type="hidden" name="argv[][id]" value="task" />
  <input type="hidden" name="argv[][config][one]" value="c1" />
  <input type="hidden" name="argv[][config][two]" value="c2" />
  <input type="hidden" name="argv[][config][three]" value="c3" />
  <input type="hidden" name="argv[][argv][]" value="v1" />
  <input type="hidden" name="argv[][argv][]" value="v2" />
  <input type="hidden" name="argv[][argv][]" value="v3" />
  
  <!-- add nonsense break to ensure division between nodes -->
  <input type="hidden" name="argv[]" value="--" />
  
  <input type="hidden" name="argv[][argv][]" value="v1" />
  <input type="hidden" name="argv[][argv][]" value="v2" />
  <input type="hidden" name="argv[][argv][]" value="v3" />
  <input type="hidden" name="argv[][id]" value="task" />
  <input type="hidden" name="argv[][config][one]" value="c1" />
  <input type="hidden" name="argv[][config][two]" value="c2" />
  <input type="hidden" name="argv[][config][three]" value="c3" />
  
  <!-- additional breaks ok anywhere -->
  <input type="hidden" name="argv[]" value="1" />
  <input type="hidden" name="argv[]" value="2" />
  <input type="hidden" name="argv[]" value="3" />
  
  <input type="submit" value="Update" />
</form>