Not Skip, but let me take a crack at an explanation.
I agree that NuVoc's wording makes it sound like monadic { doesn't work on
unboxed lists. However...
{1 2 3
┌─────┐
│1 2 3│
└─────┘
{<1 2 3
┌─┬─┬─┐
│1│2│3│
└─┴─┴─┘
So { certainly does *something* to non-boxed arguments. From the documentation,
monadic { takes the *items* of its argument, combining the atoms of each into
some kind of Cartesian product.
So perhaps an imperative programmer might think about monadic { as something
like this:
1) Separate the argument into n=:#y items,
2) Unbox each item, enumerating these as u_1, u_2, ..., u_i, ..., u_n
minding that unboxing a non-boxed noun is just a no-op,
3) Note the rank of each unboxed item, reshaping atomic u_i to 1-lists,
4) Form n-tuples (represented as boxed lists) from the atoms of each u_i, and
5) Combine the n-tuples into an array such that each k-cell that covers all the
atoms of u_i has the same shape as u_i.
So this solves the mystery with the example above---monadic { separates 1 2 3
into 1, 2, and 3, unboxes them (which is a no-op), turns them into 1-lists, and
combines the (single) atoms of these lists into a (single) n-tuple.
So, what about Skip's anwser? Well, consider the rank of {
{ b. 0
1 0 _
the first number of which represents monadic rank. Thus, Skip forms a table,
where each array (row) is what you want boxed. Then, feeding this array to {,
the J engine first splits the table into arrays, feeding each of these arrays
to monadic { individually, which as we showed above effectively acts like < on
the individual unboxed lists. Finally the J engine collects these results
together into a rank n-1 array. Since our original array was rank 2, the result
has rank 2-1 = 1.
Clear as mud?
If you *really* want the nitty gritty details, the jsource repository is fun to
dig around in. Monadic { is implemented under jsrc/vfrom.c as the jtcatalog
function. Keep in mind, though, the source departs wildly from your typical C
idioms, favoring a style which looks and feels a lot more like J.
More generally, when trying to puzzle through the behaviour of J verbs,
conjunctions etc., I find it helpful to ask these questions:
1) What is it's rank?
2) How does it behave for different type/precision arguments?
3) Do complex arguments do something special?
4) If applicable, What do _ and/or __ do?
5) Do boxed arguments have special semantics (e.g. ^:)?
6) What about tolerance and precision?
From my (albeit limited) experience, J verbs go to a lot of effort to be
maximally general, so it pays to really think about corner cases or "wierd"
values and types, e.g. <. on complex numbers.
Cheers!
HH PackRat <hhpack...@gmail.com> wrote:
> On 5/17/20, HH PackRat <hhpack...@gmail.com> wrote:
> > I need to convert the following list of 5 (in reality, far more)
> > 11-character dates: ...
> > to a list of 5 boxed dates: ...
>
> Many THANK YOU's to those who responded! As usual with J, there's
> often more than one way to accomplish something--in this case, 5
> completely different ways that I can choose to add to my personal J
> vocabulary.
>
> The shortest was Skip Cave's response, but, Skip, I have a question
> about how/why yours works. When I read the Vocabulary and NuVoc to
> see how and why this works, I could not figure out how I would even
> think to use "{" (catalogue) based on the information in those two
> sources. The NuVoc definition/description is "Combines ITEMS from the
> ATOMS inside a BOXED LIST to form a catalogue." My original list was
> NOT boxed, so that description would automatically dismiss my example
> from using "{". Can you please clarify why "{" can be used to box an
> unboxed list?
>
> And, by the way, in my programming, I like to do things small step by
> small step so that I can be sure everything works correctly at each
> point. My original goal in this case was to get the unboxed list
> first boxed in the correct manner and then transpose it from a column
> to a row. (It's part of stock market data from a particular vendor
> which is out of "standard" order and format.) Three of the five
> responses did both aspects (box and transpose) in a single
> command--for which I am quite grateful. It showed me again how
> amazing J is!
>
> Harvey
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
