TODO.md
In hughjonesd/betr: betr: Behavioral Experiments Toolkit in R

[] RookServer not respecting port under RStudio (workaround; needs actual fix)
[] Server doesn't stop after rm(expt). Still there?
[] RookServer creates sink() a lot
[] FormStage doesn't like not getting params...?
[] how to move on manually from TextStages? autorefresh?
[] ReplayServer gets put in .oldserver
[] documentation doesn't show default values e.g. in experiment(...)

[x] timeout stage to wrap other stages
would be nice to do this for "wait" also but hard to "redisplay the page"
[] use makeActiveBinding to "watch" data frames etc.... or define an on_request hook?
[] easy equivalent to zTree Participate=0, i.e. make some subjects skip a stage
[x] general framework for HTML pages
[] how to deal with 'errors'?
[] check function errors should be translatable
[] is_email, is_date
[] HTML form elements in separate package
checker functions
javascript attribute?
using Twitter Bootstrap?
[] vignette: tips/tricks/bugs

[] chat box with Rook/websockets?
[] media server with Rook?
[] JSON server with Rook?
[] dynamics with Shiny?
[] package for updatable views in R?
[] admin interface with Rook or gtk?
[] include the subjects data frame in a standard location on a per period basis
integrate existing subjects df; also Profit, Group?
subject(ids, periods) and subject(ids, periods, ...) functions
similarly group(gids, periods) and subjects(periods)
all with defaults taken from current id, period? This requires messing with environments, see below. Problem: approach won't work reliably when called from other functions.... Seems extra hassle compared to just typing subject(i, p, ...) or group(group_of(i,p), p, ...).

  tmp <- setRefClass("tmp", 
    fields=list(id="ANY", period="ANY", handler="ANY", mydf="data.frame"), 
    methods=list(
      callMe = function() {
        assign(".id", id, pos=environment(handler)); 
        assign(".period", period, pos=environment(handler)); 
        assign("subject", function(id=.id, period=.period, a){
              mydf[mydf$id==id & mydf$period==period,"a"] <<- a}, 
              pos=environment(handler)); 
        handler(id, period)
      }
    ))

[] MTurk payment integration?

Require Shiny server to be installed....
SES knows about ShinyServer (or even is a subclass?)
server.R includes shinyServer(func) where func is part of SES
or you can use runApp... but NB, does not return! parallel?
HTML output goes into a central page
No more Rook objects: timeouts set as attribute, detected by Server
reactiveTimer() could be the implementation
NEXT, WAIT and next_stage immediately pushed to client
new stages which presumably return a basicPage object
Pointless to pass these to RookServer; give them shiny_ prefix
existing stages get wrapped in HTML() and pushed to client by SES
header() and footer() delegated to Server?
SES registers UI handler via shinyUI
need to make sure that anything that changes experiment state is handled as a request. maybe a function that wraps the various UI elements and says "this one generates a request"?
some new shiny objects showing per-user data?