@Raul, > I think it's difficult to understand, and I am uncomfortable with that.
So am I. I've merely copied the info from the J Dict as a holding operation. No added value in that. The comma idea is a good one. No reason not to do it. The improved rank info which Henry and I are putting together will be on each primitive-page, not on the portal - at least not to start with. Henry's proposal is mocked-up here: http://www.jsoftware.com/jwiki/Vocabulary/eq_iac2 Clicking the link (in this instance:) "Operates on y as a whole" links to the relevant section in: http://www.jsoftware.com/jwiki/Vocabulary/RankInfo ...but currently this instance just points to a stub. Look elsewhere on the RankInfo page to see the wording Henry's drafted already for other rank instances. Note that monads and dyads will be treated separately, so there won't be as many sections as the 22 distinct "ranks" (=rank triples?) you've identified. Once completed, Henry is going to try it out with his students. Meanwhile I'm going to hold-off making improvements to the NuVoc portal itself. Then, it won't surprise me if the final material isn't adaptable to the portal too. Maybe I'm missing something, but isn't this in essence the same thing as you're proposing? On Sat, Jan 25, 2014 at 2:21 PM, Raul Miller <[email protected]> wrote: > I doubt many will see it, the way you currently have rank information > on NuVoc. But for people who do stumble over it (possibly if referred > by others), I think it's difficult to understand, and I am > uncomfortable with that. > > Here's an example page, though: > > ---------------------------------------------------------------- > "scalar" rank 0, 0 0 > > "scalar functions" include things like add and subtract. These operate > independently on individual numbers (or individual characters or > individual boxes), except that when combining arguments missing > trailing dimensions are achieved through repetition. In other words: > > 1 + 2 3 4 > 3 4 5 > gives the same result as > 1 1 1 + 2 3 4 > 3 4 5 > > In the first example, here, we have an array with three elements and > another array with no dimension at all (single numbers are different > from lists of numbers, if you want a list of just one number do > something like (,1)). > > Similarly, > 10 100 1000 - i. 3 2 > 10 9 > 98 97 > 996 995 > > Here, the list on the left does not have a second dimension, so when > matched up against a two column argument you get the effect of having > two columns by repeating each number. > > ---------------------------------------------------------------- > > This is slightly repetitive, but I think that the repetition can be > worth it, for some readers. Similarly, I think separating the monadic > rank from the dyadic rank with a comma would be a good thing. > > But if you have a problem with this approach, I would love to understand > it. > > Thanks, > > -- > Raul > > > On Fri, Jan 24, 2014 at 11:36 PM, Ian Clark <[email protected]> wrote: > > @Raul -- I think what you're proposing is in line with Henry's proposals > > (see earlier thread), which I'm in-process of implementing. But these > will > > show on individual pages, not the NuVoc portal page. > > > > Trouble is, I don't seriously believe in rank information on the NuVoc > > portal -- much less the "mu mv mu" -stuff for adverbs and conjunctions (I > > originally had: ? ? ?). I just did it as a service to people who prize > it. > > > > @Joe - I've given up on J beginners with an APL background. They've > either > > all ditched APL for J by now, or they never will. :-~ > > > > > > On Fri, Jan 24, 2014 at 3:44 PM, Joe Bogner <[email protected]> wrote: > > > >> Connecting two threads -- Ian wrote this here: > >> http://jsoftware.com/pipermail/programming/2014-January/034769.html > >> > >> > Thus, in the first year of learning J, I unconsciously conjectured > Rank > >> > didn't matter -- for most purposes. You could intuit most problems, > ones > >> > that don't emerge much for rank 3 arrays or less. In a lifetime of APL > >> I'd > >> > got away without bothering my head about verb rank. > >> > >> This has been my experience as well in my first year. > >> > >> On Fri, Jan 24, 2014 at 9:55 AM, Raul Miller <[email protected]> > >> wrote: > >> > Looking at your current nuvoc page, I see 22 distinct "ranks". The > >> > most common rank entries are _ _ _ (27 of those) and 0 0 0 (26 of > >> > those). Actual numeric values are limited to 0, 1, 2 and _. and there > >> > are also a variety of abstract values (mv, mu, lv, rv, lu, ru). > >> > > >> > So one observation is that the casual user will not understand that > >> > 'mv' means "the rank of the monadic definition of verb argument v". > >> > >> I think I understand the simple case of rank (at the wikipedia page > >> level[1]). I like the suggestion on somehow clarifying the differences > >> between the 22 possibilities. > >> > >> I also wonder if something really direct "verb rank matters most in > >> the following situations..." would be helpful. If it's mostly rank 4+ > >> arrays, then the complexity of the different cases may not be > >> encountered too frequently. This quote comes to mind: > >> > >> http://www.jsoftware.com/papers/perlis78.htm > >> In APL, what flashes through your mind is a cascade of operations: > >> chasing data through arrays, out of the other end of which come — > >> limping and bruised, you know — seven numbers. After having built up > >> arrays of rank eight and coming perilously close to a workspace full > >> out from the other end comes these seven numbers — and they’re pulled > >> out almost painfully — and you say to yourself, “My God, that’s > >> wonderful! That’s a mechanism!” > >> > >> If that's what those 22 cases are intended to support, I and many > >> other beginners to APL may have a long journey to go before it > >> matters. Beginners to J with an APL background are likely a different > >> story. Is NuVoc targeted more so at either one? > >> > >> > >> [1] - http://en.wikipedia.org/wiki/Rank_(J_programming_language) > >> ---------------------------------------------------------------------- > >> For information about J forums see http://www.jsoftware.com/forums.htm > >> > > ---------------------------------------------------------------------- > > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
