* codegen for threadvars
* two issues for thread local heaps:
  - must prevent to construct a data structure that contains memory 
    from different heaps:  n.next = otherHeapPtr

* add --deadlock_prevention:on|off switch? timeout for locks?
* test the sort implementation again


High priority (version 0.9.0)
=============================

- iterators should not always be destructive!
- warning for implicit openArray -> varargs convention
- implement explicit varargs
- tests: run modules that contain "#RUN_ME", compile the other
  modules; run the GC tests
- fix the streams implementation so that it uses methods
- fix overloading resolution
- wrong co-/contravariance
- make ^ available as operator

Bugs
----
- proc (x: int) is passable to proc (x: var int)  !?
- the parser allows empty object case branches
- pegs: the anchor '^' does not work because many procs use a linear search
  and matchLen()
- BUG: generic assign still buggy
  - Optimization: If we use a temporary for the result anyway the code gen
    should make use of this fact to generate better code...
- bug: invoking a generic iterator twice triggers a code gen bug
- sorting with leads to a strange memory corruption!
  --> system.swap or genericAssign is broken! And indeed, if reference counts
      are not modified and the GC is triggered in between a swap, bad things
      may happen!
  
  proc sort*[A](t: var TCountTable[A]) =
    for i in 0 .. high(t.data)-1:
      var maxIdx = i
      for j in i+1 .. high(t.data):
        if t.data[j].val == 3: echo "touched! ", t.data[j].key
        if t.data[j].val > t.data[maxIdx].val: maxIdx = j
      swap(t.data[maxIdx], t.data[i])


To implement
------------

* distinct types for array/seq indexes
* implement closures for the C code generator
* GC: marker procs for native Nimrod GC and Boehm GC
* built-in serialization


Low priority
------------
- implicit ref/ptr->var conversion; the compiler may store an object
  implicitly on the heap for write barrier efficiency
- resizing of strings/sequences could take into account the memory that
  is allocated
- typeAllowed() for parameters...
- find a way to reintroduce the cleanup() pass for C code generation: this
  is hard because of partial evaluation --> symbol files will fix this as
  a side effect
- floating point checks for EcmaScript
- prefer proc in current module over other procs with same overloading result?
- real types for template results
- generalized case statement (requires better transf)
- tlastmod returns wrong results on BSD (Linux, MacOS X: works)
- nested tuple unpacking
- better error messages for used keywords as identifiers
- case statement branches should support constant sets
- 'nimrod def': does not always work
- test branch coverage
- checked exceptions
- fix implicit generic routines


Library
-------

- specialized bit sets; specialized string sets
- locale support
- conversion between character sets
- bignums
- ftp (and other internet protocols)

- pdcurses bindings
- queues additional to streams: have two positions (read/write) instead of one

- for system:
  proc `@` [T](a: openArray[T]): seq[T] = 
    newSeq(result, a.len)
    for i in 0..a.len-1: result[i] = a[i]
    
  --> ensure @[] calls the array version!


Version 2
---------

- language change: inheritance should only work with reference types, so that
  the ``type`` field is not needed for objects! --> zero overhead aggregation
  BETTER: ``is`` and safe object conversions only work with ref objects. Same
  for multi methods.
  
- explicit nil types?
  * nil seq[int]
  * nil string
  * nil ref int
  * nil ptr THallo
  * nil proc 

- better for backwards compatibility: default nilable, but ``not nil``
  notation:
  
  type
    PWindow = ref TWindow not nil
    
  The problem with ``nil`` is that the language currently relies on it for
  implicit initialization. Initialization is different from assignment. The
  issues can "easily" dealt with by ensuring:
  
    var x = myProc() # checks myProc() initializes every pointer explicitely
    
- the two other parsers


Low priority
------------

- ``when T is int`` for generic code
- ``when validCode( proc ()  )`` for generic code
- macros: ``typecheck`` pragma; this allows transformations based on types!
- find a way for easy constructors and destructors; (destructors are much more
  important than constructors)
- code generated for type information is wasteful


Other ideas
-----------
- startsWith `=^`
- endsWith `=$`
- ignore case `=?` --> `=$?` too?