I just now see this won’t happen. J can’t anticipate how <fold verb>&y will be used. So keep all the primitives but have them semantically work like either reversing x or not. I still think, from the 'control' vs 'data' point of view, the array of controls should go left, the initial data right.
And items of the control should be fed into v on the left, while intermediate processed versions of the initial data should be fet into v on the right. x u F.. v y single “reverses x” x u F.: v y single doesn’t x u F:. v y multiple “reverses x” x u F:: v y multiple doesn’t Maybe for mnemonics, one might want to swap the semantics of trailing . and : in the F primitives but that’s rather an aesthetics/convenience issue. rtl is the standard J-ish way, and ltr is a bastard deviation supported for – well, it doesn’t hurt to have it. A philosophical question? The problem arising with folds is that now the order of the items of x is the order of not space but time: which are the earlier ones to be processed? And here, J has always had a multiple personality disorder: plot y without abscissa values in y takes left-most items to be the /earliest/ ones. That’s wrong. In J, the earliest ones need to be on the right. Hauke Am 24.02.21 um 09:27 schrieb Hauke Rehr: > Maybe it’s better to have custom implementations for the pattern > (|. x) <fold verb> y -- ---------------------- mail written using NEO neo-layout.org ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
