Gekko 3.2 has been released. Unlike the release of Gekko 3.0, Gekko 3.2 does not include any syntax changes or similar annoyances. In fact, a lot of the 3.1.x series was dedicated to stabilizing version 3.0 as much as possible, as the transition from Gekko 2.4 to 3.0 was heavy surgery regarding the source code, re-coding large parts of Gekko completely.
So, besides stabilization, what’s new in 3.2? The 3.2 version could have been released much sooner, but the intention was to revise the Gekko help system in conjunction with a major release like 3.2, and revising the help system took quite a lot of time. The help system has now been thoroughly revised, resulting in a cleaner and much more consistent experience.
Relative to version 3.0, Gekko 3.2 also includes a long list of changes, cf. the section “Introduction” –> “New features” in the help system. A great deal of effort went into further developing the decomposition of model effects (DECOMP), which now includes support for GAMS models, too. Additionally, a new data-trace system has been implemented, which can trace the origin of individual time-series observations all the way back to their source (e.g., a value from a .csv file or a hard-coded value like x = 100, this is also known as data lineage). More frequencies have been added, and an Excel add-in has been developed. A large number of other incremental improvements and features have also been implemented, too many to list here.
The Road Ahead
Looking forward from Gekko 3.2 onto the 3.3.x development series, what will it be about? This will be decided at the annual Gekko steering committee meetings, but also through more informal discussions between the Gekko editor and the Gekko users. In general, to see how Gekko is progressing, check out the minutes from the steering committee meetings (available here), check out what is happening in the development versions (see here), look at the more detailed change-log, or maybe even check out the source code changes on the Gekko GitHub page.
Regarding the future of the Gekko project, some users believe that Gekko’s core standout features could (also) be offered as a package for a data science language/environment like for instance Python. For instance, Gekko provides list and matrix operations, but so does Python, which also handles printing, plotting, statistics, etc. etc. So perhaps Gekko should focus on its main strengths, such as convenient time-series handling and model decomposition, rather than trying to implement more list or matrix features and so on. Since Gekko is already available as an Excel package (add-in), why not offer it as a package for Python (or R or Julia), too? This topic will be discussed at the upcoming steering committee meetings, and any conclusions will be noted in the minutes.
Recent Comments