Forum: developers
Monitor Forum | 
Start New Thread
| API design [ Reply ] By: Ott Toomet on 2009-04-30 03:24 | [forum:1467] |
|
Actually, I am not quite sure how to design a good API. In c++ I would a number of class public functions. Does it makes sense in R? Something like: val = list(N, getN=function(x) x$N, ...) The advantage is that you separate the structure of the data from API but as the get* functions were part of the API themselves... It simply does not feel any better... I don't see any notable difference here with using S4 either. Another way were to use generic functions + correHi, Just to report the current 'state': I added the printing of the levels of the selection variables. However, as the check crashes at tobitBayes, I will look at the tests when it is fixed. You may want to change the format of the output a little bit, feel free to do that. tobitBayes: it seems like the problem is an API break in plm.sponding methods. But who want to have the memory loaded with the generic functions for every single strange parameter, an esoteric model may spit out? Best solution possibly to implement generic for sufficiently general stuff, and try to keep API fixed. Ideas? |
|
