Ken Iverson's introductory text is <i>Math for the Layman</i>. It suggests that "learning J" is a bit like "learning math" and that proficiency should probably be measured more in terms of how one applies J to particular problems rather than general knowledge of syntax and recall of library functions as is common when talking about computer languages.<p><a href="http://www.cs.trinity.edu/About/The_Courses/cs301/math-for-the-layman/" rel="nofollow">http://www.cs.trinity.edu/About/The_Courses/cs301/math-for-t...</a><p>Or to put it another way, a person can know J in the same sort of way a 7th grader knows math...or a 3rd grader or a Master's candidate. J is an exploratory language and will meet a person where they are. It is useful whether the person is only capable of basic arithmetic or convex hulls.
I've been slowly learning J on and off for the past year or so (well, I wanted to learn it ever since I saw this [1]). It still feels pretty write-only to me, and I wouldn't make any large project in it, but as a brainstorming, research and experimentation tool, it's gold. This especially applies if you're working with matrices or things that can be well represented by arrays of any dimension.<p>The jqt environment that comes with the standard package is polished, and encourages an interactive workflow where you progressively build your sentences while looking at the output. This way also greatly eases the learning curve.<p>Another useful tool to help learning is the cheat sheet [2]. I'm not really into dead trees but I still printed it on paper because I'm not surrounded by monitors, and it really helps to have a quick reference to the very rich and concise J vocabulary. Applying unexplored verbs to problems that interest you is a fun exercise.<p>One thing that struck me it that it's difficult to look up examples to help you as you go, in part because the one-letter name is not search engine-friendly, but more importantly because of the small size of the J corpus out there on the web. This is good because it force you to find solutions yourself, but it's bad because you never know if you're limiting yourself to a sub-optimal solution.<p>[1]: <a href="https://news.ycombinator.com/item?id=1041500" rel="nofollow">https://news.ycombinator.com/item?id=1041500</a><p>[2]: <a href="http://www.jsoftware.com/jwiki/HenryRich?action=AttachFile&do=view&target=J602_RefCard_color_letter_current.pdf" rel="nofollow">http://www.jsoftware.com/jwiki/HenryRich?action=AttachFile&d...</a>
Nice to see an article that explains a bit about J's background. And definite sympathy points for the language supporting complex numbers. But I'm still wondering what the language's “high performance” is based upon. I'm an assembler hog and any language that doesn't compile to native code is just a toy to me. Yet the article only mentions an interpreter. I think, by now there must be something better than an interpreted write-only language.
if you put<p><pre><code> s =: ({. , }. /: 12"_ o. }. - {.) @ /:~
</code></pre>
in my code base, you are going to get smacked.
Expressability of programming languages has a cognitive limit and J is flirting with it.