Some other people have said similar things on this topic, but here are my thoughts:
Your way of thinking does need to change in order to use J. It's not an unnatural change, or even a hard change, in the sense that the J way of thinking is just as natural and as easy as the Python (or other imperative language) way of thinking. However, it is a large shift so you will need to put some work into divorcing your thoughts about programming from your knowledge of the imperative style. The first few things you want to do are not ends, but means--they are ways of accomplishing tasks in Python but aren't necessarily important in J. This means you have moved a step forward from your problem statement to a method of solving that problem. In order to learn J, I recommend that you prevent yourself from taking that step and consider what you want to do rather than how you want to do it. For instance: > Passing 3 or more parameters to a verb > Turning a vector into a list of arguments What you really want to do is to work with multiple pieces of data at the same time. This is performed easily enough by placing the data in an array with three or more elements. And why would you want to work with a list of arguments? It's invariable easier to pass a list. If you want to assign a name to each element of an array of a fixed length, use multiple assignment, for example 'x y z' =: 4 ; 'Hello' ; 1 0 0 x 4 y Hello z 1 0 0 . > ...loops? J allows for loops (using similar semantics to Python's for ... in ...) in explicit verbs, but usually general loops are not called for and a specific case suffices. The transition from using for loops to solve every problem to using a number of more specific cases with nicer properties strongly parallels the historic shift from using the general goto construct to structured loops like for and while. In this case, you can add 2.91 to each number in the list, divide by 0.4, and then sum: +/ 0.4 %~ 2.91 + i.10 185.25 . As for interacting with the outside world, a class of verbs called foreigns form J's interface to the outside world: http://www.jsoftware.com/help/dictionary/xmain.htm Many of these have aliases or nicer versions (e.g. fread is a wrapper for 1!:1) in the standard library, but I don't know where all these verbs are documented. Marshall On Sat, Feb 15, 2014 at 12:04:13PM -0500, Lee Fallat wrote: > Thanks for the explanation Raul. > > I think I understand now that string manipulation in J is > non-trivial...I was following right up until the very last J sentence > ( 'hi',:&.(-&(a.i.'*'))&.(a.&i.)'there'), mostly because I still only > know a few verbs. I am starting to think J is either not for me, or my > way of thinking really needs to change for this language. There are so > many things that would seem straight forward in traditional languages > than J, but I will have to see! > > Some examples of what I'm thinking of would be difficult or different in J: > Passing 3 or more parameters to a verb > Turning a vector into a list of arguments > ...loops? From what I've seen loops are built into some verbs (+/ -/ > etc), but what if you wanted to do say: add every number in the > vector, but for every number add 2.91, then divide that by 0.4? > In python: > range() generates a list of numbers from 0 to 9. > for x in range(0,10) > y += (x + 2.91) / 0.4 > One other thing is interacting with the operating system...How do I > check to see if files, or even directories exist? Can you create > directories, or change permissions of files with J? The scope of J > seems to be very constrained to mathematics. It would be nice to use J > as a system programming language! Are there any examples of people > doing these things? > > Anyways, I apologize for being off-topic, but just some things I've > been thinking about. Like I said I still have to get through some more > learning material. > > Regards, > > Lee > > > > On Sat, Feb 15, 2014 at 11:37 AM, Raul Miller <rauldmil...@gmail.com> wrote: > > Conceptually speaking, J has three "atomic data types": > > > > Numeric > > Literal > > Boxed > > > > Numeric types support arithmetic: 1+1 > > Literal types are what might be called character types in other languages. > > (ascii or unicode) > > Boxed types are what might be called reference types in other languages. > > > > J tries to treat numeric types as mathematical arithmetic entities. It's > > not perfect, and involves tradeoffs but it does a pretty good job at being > > useful and educational while still offering decent performance with a > > relatively compact implementation. Also, J booleans are numeric and I > > vastly prefer this approach (which fits my understanding of the history of > > booleans) over the convention of enshrining arbitrary implementation > > limitations. > > > > You cannot perform arithmetic directly on literals but you can, for > > example, compare an ascii 'A' with a unicode 'A' and get a literally > > correct answer. (However, unicode also includes an open ended set of rules > > and extensions and if you want those you will probably need to implement > > them yourself. To avoid open-ended arguments about these issues its perhaps > > better to refer to this type of data as 'literal' rather than 'character'.) > > > > Boxes take an arbitrary array and "puts it in a box" - you get a single > > thing which can be treated something like a literal or a number. > > > > You cannot mix these three types in the same array, but if you put > > something in a box you can put the box in an array of boxes. > > > > Anyways... > > > > For some contexts you might use a list of characters (excuse me: I mean > > literals) to represent a string. For other contexts you might want to put > > that list of characters in a box. If you arrange your strings as a two > > dimensional array they will all be padded (with spaces) to match the length > > of the longest one. > > > > Actually, in some contexts you might want to use a list of numbers to > > represent a string. Sometimes arithmetic is handy. > > > > Put differently, J does not actually have a "string" data type. But you can > > represent them using arrays. > > > > Examples: > > > > 'hello' > > hello > > 'hello';'there' > > ┌─────┬─────┐ > > │hello│there│ > > └─────┴─────┘ > > 'hi',:'there' > > hi > > there > > 'hi',:&.(-&(a.i.'*'))&.(a.&i.)'there' > > hi*** > > there > > > > In that last example, I made the padding character by '*' rather than ' '. > > I did this by combining the two strings as numbers rather than literals. > > The padding value for numbers is zero. But if I had stopped there I would > > have gotten ascii nulls for my padding. So I also combined them under > > subtracting by the numeric value for '*'. > > > > Under performs the reverse transform for the result (of the transform that > > it performed from the arguments), > > http://www.jsoftware.com/jwiki/Essays/Under > > > > I hope this makes sense. > > > > Thanks, > > > > -- > > Raul > > > > On Sat, Feb 15, 2014 at 10:31 AM, Lee Fallat <ircsurfe...@gmail.com> wrote: > > > >> Thank you all for your kind replies! And to those saying how I should > >> really use awk for this job- I know, I was just curious! I am very > >> impressed by the variance in answers, and how different J really is > >> compared to other languages. > >> > >> Thanks again, > >> > >> Lee > >> > >> P.S. As for the input, just numbers was fine. I'm curious though what > >> J does when it encounters strings (and numbers!). I figured reading > >> through the books offered on the wiki will explain this. (+/ "hello" > >> "world" 100 = what? :) > >> > >> On Sat, Feb 15, 2014 at 10:20 AM, Jim Russell <jimsruss...@yahoo.com> > >> wrote: > >> > I love AWK; it (and perl) have saved my bacon many times. If your > >> problem involves processing fields within lines of an I/O stream in a *nix > >> environment, of course you or I should use AWK. Particularly me, since I'd > >> never be given a processing task involving more math than a "gozinta" or > >> takeaway, much less anything involving polynomials, natural logs, verb > >> inverses, factorials, ranks above 3, and a whole bunch of stuff that J > >> would do for me if only I understood what it was. > >> > > >> > (I would also pick AWK if I had only 5 minutes to learn a new language.) > >> > > >> > But had AWK never been invented (shudder), and I needed to write it, > >> would I use J? Well, not me, but I know some folks here that might knock it > >> out using J in an afternoon or two. > >> > > >> >> On Feb 14, 2014, at 9:51 PM, Lee Fallat <ircsurfe...@gmail.com> wrote: > >> >> > >> >> Hey there, > >> >> > >> >> As new user to J (but several years experience with C and Java), I > >> >> find it very, very interesting. The power of its one liners and > >> >> mathematical heritage really have me hooked. I was wondering though > >> >> if it has similar capabilities as awk. What's the equivalent to this > >> >> awk script in J?: > >> >> > >> >> BEGIN { FS=";" } > >> >> { print $1+$2 } > >> >> > >> >> This script sets a FieldSeparator to ;, and then for every "row", add > >> >> the first and second column and prints it. I would like to replace awk > >> >> with J! > >> >> > >> >> Thank you, > >> >> > >> >> Lee > >> >> > >> >> P.S. Excuse me if I've misidentified J sentences. (Sentences -> > >> statements?) > >> >> ---------------------------------------------------------------------- > >> >> 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 > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
