On Fri, Jul 01, 2005 at 11:43:45PM +0300, Martin Vermeer wrote: > OK, I see it now: idx_type should be a struct with two int members, .row > and .col. And suitable operator definitions added. No duplication of > information, and still "idx() = ..." will work. Does that sound like a > plan?
I still don't think so. idx_ is just an integer offset in the cell vector. These cells may be laid out in a grid as in, well, a MathGridInset, or even a MathFracInset, but do not have to (as shown by e.g. MathSqrtInset). This single-number idx member is chosen as primary information as it is directly useful when accessing the cell vector. For grid-like structures it is sometimes convienient to extract row and col from idx. So there are some helper functions. And sometimes it is useful to give the user the impression that row and col are primary because that's what the user expects if he e.g. works in grids. In this situation we just adjust the idx. You found places where this adjsutment was not or not properly done. But that means just that: The adjustment is wrong, not the scheme. You propose to make row and col primary information but that does not fit well into the current scheme. We'd need to compute idx ach time we access the cell vector. For some insets like sqrt it would not even be clear what values for row and col are permissable etc. Andre'
