On Sun, Jun 23, 2019 at 4:41 PM Eli Zaretskii wrote: > > From: Pip Cet > > Date: Sun, 23 Jun 2019 16:26:16 +0000 > > Cc: triska@metalevel.at, 36337@debbugs.gnu.org > > > > My suggestion would be to expand `substring' to work on bool vectors, > > then building a vector of bool vectors and using the existing code for > > that case. Less code in image.c, plus a new utility function that > > might be generally useful. > > Or maybe we should have a variant of make-bool-vector that accepts 2 > dimension s instead of just one? I don't really see how that would be generally useful, to be honest. In fact, I just played around with removing bool vector support entirely. > > (However, do we want to encourage people to use bool vectors?) > Why not? We seem to lack even very basic functions for interacting with bool vectors, and hardly anyone appears to be using them. Even the :stipple face property doesn't. Emacs starts up fine with bool vector support removed. We can use vectors of nil/t (in most cases) or unibyte strings or bignums (which have arbitrary size limits now, but bigbignums would be just a few lines of code, I think). And people _think_ bool vectors have a natural presentation as bytes, but they don't, because some people start with the most significant bit. So I just don't see where bool vectors fit in. > Evidently, it's convenient in this particular use case. Is the convenience worth a thousand lines of code (much of it C) and documentation?