Related:

Roadmap  

[As of January, 2023]

Gekko 3.0 was released as a stable version in the spring of 2019, and new users are advised to use versions in the 3.1.x series, which can be thought of as a “stable” development series (cf. the versions overview page). A stable version 2.4 exists, but in general, the work on the 2.x series is discontinued. Development-wise, at the moment the 3.1.x version is still being stabilized (including help systems etc), but quite a lot of new development is also taking place. It is expected that an official Gekko 3.2 will be released in the spring of 2023, and following this a 3.3.x development series will start up. Syntax changes are not planned for the foreseeable future (except minor tweaks).

Aims regarding the further development of the 3.1.x series (or possibly 3.3.x series):

  • Stabilizing 3.1.x continues.
  • Continued work on better error messages in Gekko 3.1.x.
  • Databank tracing, so that Gekko databanks contain info on the history of series objects (when they were created, which databanks and series they depend upon, etc.). In principle, it should be possible to trace a data value back to its origins, kind of like a DECOMP for databanks.
  • Shemas/templates for Gekko tables, created as Excel sheets. These would complement/supersede the existing xml-based .gtb-files.
  • More advanced seasonal correction (JDemetra+ etc.), perhaps via the R RJDemetra package. Better temporal disaggregation, for instance via the R tempdisagg package.
  • Finetuning Gekcel, among other things implementing some of the Gekcel VBA code that the Central Bank of Denmark uses.
  • Improved solver capabilities. Blocks of equations, equation objects, model objects. More means than goals. Improved tracking when simulations fail. (Handling blocks of equations has been implemented in the 2.5.x series, but regarding the 3.1.x version, this will be built differently, using equation objects etc. Still, the experiences from model blocks in 2.5.x will be useful). Perhaps also static simulation possibilities (in a sense removing lags and solving the model for one period to obtain long-run values).
  • More advanced PLOT windows, vector-based.
  • Implementing some of the “missing” functions/procedures from AREMOS that deal with holes in data, interpolation, extrapolation, or conversion between frequencies.
  • Better Python and R interfaces. Dataframe objects in Gekko would facilitate communication with Python and R. Perhaps use Apache Arrow as an interface for data interchange.