Differences between revisions 2 and 3
Revision 2 as of 2008-02-29 16:22:25
Size: 2775
Comment:
Revision 3 as of 2008-03-07 16:00:59
Size: 2727
Comment:
Deletions are marked like this. Additions are marked like this.
Line 2: Line 2:
## page was renamed from ParserTodoList
Line 4: Line 3:

 * create a testsuite
Line 11: Line 12:
 * add conditional compilation

Todo List

  • create a testsuite

Features

  • add an option to model keyword that let use model functions already written for another model in order to save time when the model doesn't change but the computing tasks do. [parser and maybe Matlab code]
  • add option for number of draw in Metropolis subsamples [parser + Matlab code]
  • handling of functions. Either Matlab functions or user functions not in the list known by the derivator. Also in expressions outside of model definition. [parser]
  • allow for expression in many places where numerical constant are currently required [parser]
  • add option for algorithm to solve generalized Sylverster equation [parser + Matlab code]
  • add new function diff(), sum loops and prod loops in modeling language. Should be unrolled. [parser]
  • trigger numerical differentiation when an unknown function is present in the model [see UnknownFunctions]

Error checking

  • add warning for missing variable in current period. This is permissible in some cases, Michel should write a note about it [parser]
  • improve error and warning handling in parser. Find a way that users see Warnings even if scrolling by [parser]

Checks concerning the structure of the mod file

  • allow a single histval, endval, initval block
  • check that a given parameter is never declared two times
  • check that observation trends is used after a varobs

Checks concerning the types of variables allowed in expressions

  • varobs must be declared endogenous
  • shocks must be exogenous
  • objective_planner must be endogenous without leads/lags

Optimizations / Enhancements

  • optimize expression sharing (for example use the fact that A+(B+C) = (A+B)+C); this would also enhance the pre-computing of numerical constants (for the moment, the algorithm can't reduce 1+(x+1))

    • if there is a series of operators with the same precedence, we can collect the constant and reduce them 1+(x+2}=(1+2)+x=3+x

    • but it is possible to factorize more complicated expressions: 2*x+3*x=(2+3)*x=5*x. The two simplifications aren't identical but strongly related. -- MichelJuillard DateTime(2007-05-08T19:53:46Z)

  • for static output file, remove unnecessary temporary terms
  • rewrite outputting system (by using a subclass of ostream containing the output type, and by overloading << operator)

  • improve internal representation of estimated_params* statements

Loose ends

  • add C code for normcdf, both for parser evaluation and for DLLs
  • add normcdf to reference manual
  • add normpdf

DynareWiki: PreProcessorTodoList (last edited 2010-09-27 14:43:27 by SébastienVillemot)