General:Sequences in
Glee manage hetrogeneous data (e.g. character strings in the same
vector with numeric vectors). It does this by maintaining pointers to the
actual Glee objects. So Glee sequences are vectors
of pointers and are themselves homogeneous. This capability gives
Glee much of its power and flexibility. Any Glee
object may be a member (referent) of a Glee sequence.
Glee uses an internal reference counting mechanism. This allows
it to maintain references to any number of Glee objects with a
storage requirement of only 4 bytes (the size of a 32 bit C++ pointer) per
reference. No copy of the referent is made. If the referent changes, all
references to it implicitly change with it. The referent will not be discarded
as long as there are outstanding references to it. In Glee,
Sequences are themselves Glee objects. Therefore,
Glee Sequences may contain Sequences (which may contain
Sequences, etc.). Glee has a current depth limit of 5. This is
like a Russian egg limited to 5 embedding shells. This limit is totally
arbitrary. It is deemed adequate for real work and is included as protection
for sequences which directly or indirectly reference themselves.
Simple Output:This is an
example of one way Glee can present results. Note: The
parenthesis are not necessary. You will get the same result without them.
Verbose Output: What we
really see in the previous example is a sequence of two elements. The first
element is a character string containing 'answer is '. The
second element is the numeric vector (with one element) containing
(123.45). In quiet mode as in that previous example,
Glee displays sequences by catenating the quiet displays of its
referenced elements. This example is the same as the previous one except I
display in verbose mode. When I turn on verbosity (the $V), I see what
is going on in the previous example. Glee has created a sequence
of two elements. The first is a character string; the second is a numeric
vector. When displayed with verbosity you can see the individual elements.
Sequence Indexing:You can index
Sequences just like you can other Glee vector objects..
Sequence Results: Many
Glee operators, out of necessity, return sequence results. One
such operator is the (``) Indices of operator. In this example,
I generate a vector of ten 5's (5*->10). For each element of that
vector, I create a random number between 1 and 5 using the (v ? 2)
Random operator (the 2 right argument is the random seed assuring you
see what I see when you try this). I assign this result to x and
display it (=> x $;). x is now a vector of 10 random
numbers between 1 and 5. I turn on verbosity ($V;). I perform the
Indices of operation on the contents of x for values of 1 thru
5 (5`) with (x `` (5`)). Our result is a 5 element sequence
vector (one element for each element in the right argument). The result being
hetrogeneous (numeric vectors of different lengths), the sequence is the only
way to capture the result. It reports[1]:1 2 (1 is at index 1 and
2); [2]:NULL (the number 2 doesn't occur);; [3]6 8 10
[4]:4 5 7 9; and [5]:3. A real life application of this
facility might be in analyzing accounts in transaction data.
Count with (#ea
'#'):This is another taste of the power of Glee. Taking
the preceeding example and applying Glee's count operator
(#) to each (#ea) element of the sequence, I get a
sequence of counts: [1]:2 (2 elements); [2]:0 (no elements);
[3]:3; [4]:4; and [5]:1. In any other language, this
might be valuable information for setting up storage areas for rendering
results. In Glee that usually isn't an issue because
Glee is so dynamic.
Another way to do each:
(@&): Since creating the example above, I have created more
powerful ways of dealing with the elements of sequences. The
"@&" operator (read "at each") takes
the operation on the right and applys it to each element of the sequence on the
left. For a simple operator as illustrated here with
"#"count, just the operator may be given. Notice a
sequence is returned. You have to "< " disclose it
to obtain a simple numeric vector. For more complicated requirements, a whole
set of operations can be given within parenthesis. This is shown in a later
example. Finally, if there is a block within the parenthesis, anonymous blocks
of Glee code and namespace protection can be applied to the
elements of the sequence. A simple example of this is also given later below.
Scan counting:
(/#): Counting the elements of sequence elements is such a common
need that I created this count scan operator. As indicated, it is
"count" applied "at-each element" of a sequence and the
results returned element by element as with a normal scanning of elements. In
the examples before, we obtained the same result but as a sequence of counts
which we needed to disclose. Here, we obtain the desired numeric vector
directly. There are many ways to do the same thing in Glee. This
scan counting operator does the operation orders of magnitude faster than the
other examples and is the recommended way of finding the count of elements in
sequence elements. If the elements are simple objects like numeric vectors or
strings, the count is the number of those elements. If the element is a
sequence itself, the count is for the elements in that sequence.
Disclose
( < ) and depth ( ** ): As
a rule, Glee doesn't scrutinize its results. But in this example
we see how the programmer, knowing what to expect, can have Glee
return results more useful to him. From the example before, we see that when we
count the elements in each sequence element, our result is homogeneous. Each
element is a numeric singleton. Glee has the disclose
( < ) operator. It is the Less Than symbol. Used
monadically, to Glee it means disclose . How was the symbol
chosen? Notice that it opens to the right (to Glee, it opens up
or discloses its left argument). There is another symbol, as we'll see in the
next example, that encloses. (guess what symbol I chose for that). If
you ask Glee to disclose a sequence, it scans to see if it is
homogenous (all the same kind of data). If it is homogeneous,
Glee will create a single vector of that type. Otherwise it just
returns the sequence left argument. The programmer can test for this by
checking the depth of the object with the Glee's depth operator
( ** ). This operator returns 0 for simple objects
(like numeric vectors) and 1 for sequence objects (or more than
1 if the sequence contains a sequence). In this example, every element
in the sequence is numeric so Glee is able to create a numeric
vector using the disclose operator. If it couldn't, it would just return the
original sequence. In my example, I take these counts stored in z and
add them up ( \+ ). Sure enough, I've accounted for all 10.
Enclose ( > )
Implicit and Explicit:This is another compound example. First I illustrate
that for character strings, without explicit catenation, Glee
does an implicit enclose of each string returning a sequence.
Glee hates inconsistency, but as I illustrate it doesn't do this
implicit enclose with numeric vectors. The reason is a pragmatic one. In
Glee, when we're working with explicit numeric vectors, we're
usually groping around in our data to see what we have. Implicit enclose would
get in our way. On the other hand, when we're dealing with character strings
(e.g. lines of computer program) we usually want to deal with them as
individual entities so implicit enclosure is what we want. I retain the right
to change my mind on this as I develop real live programs solving real live
problems. You will soon see there are times we need to be able to form
sequences explicitly. That is the purpose of the enclose operator
( > ). Notice how it visually closes down on, or encloses, its
argument to the left.
Multiple Enclosure:Here I
illustrate a reasonably practical example of multiple enclosure. Some
APL dialects called this nested arrays. Notice how Glee
pays no attention to line feeds in instructions. Instructions are delimited by
semicolons. Otherwise, code entry is entirely free form. This example
illustrates how to form multiple enclosures. First, I have two character
strings forming an implicit sequence of 2 elements. I parenthesize to group
these elements. Then I have a character string and numeric vector forming
another implicit sequence of 2 elements. I parenthesize that into a group. Now
I have two groups, each containing two elements. If the groups were standing
alone, there would be no enclosure ... no added depth. However, when two groups
are stranded together as we have here, the groups are implicitly enclosed and
catenated. They become elements of a newly formed sequence. This example also
shows how Glee displays multiple enclosures or nested sequences.
The asterisks ( * ) in the display indicate the level of depth. I admit this
display isn't as pretty as it could be. Cosmetics are not on my critical path.
Grade and Sort:. Sequences can be
graded and sorted just like string and numeric objects. When comparisons are
made between objects of different types, like-types are grouped together and
then sorted. Strings are sorted ignoring case, white space, and punctuation.
Ravel ( , ) and Expose With
Separator( ,, ):Sequences can be created by raveling any object.
If the object is a sequence of depth other than 1, it is flattened to a
sequence of depth one. Raveling a string, number, or other object turns it into
a one element sequence. Separator elements can be inserted during the raveling.
This is called "expose with separator". Any object can be used as a
raveling separator. However, blank is so common, the monadic version of (
,, ) ravels in blanks at the deepest point and newline
characters at lesser depths. Dyadic use of this operator ravels in other
objects given as the right argument. Exposing with separators may be useful for
using sequence building techniques to quickly achieve readable displays of
complicated data objects. Note: Whereas the ravel (,) operator
yields a result of depth 1, the expose with separator (,,) does not
change depth. If depth is zero, expose with separator has no effect.
Disclose and display:
Using a combination of ravel, disclose, and the %* operator you can
manipulate sequence data into displayable form.
Strands:. Strands are formed
by delivering variable names to the interpreter with no intervening operator.
The interpreter makes a sequence with each element being the contents of the
variable. This is a common method of forming argument lists for blocks and
named blocks (programs).
At Each ( @& )
application of monadic: The at (@) each
(&) facilitates the processing of elements in a sequence. Used
niladically, it marks an object for processing element by element. For simple
objects, it converts them into sequences. For sequence objects, it merely marks
them for subsequent "each" processing. This example illustrates at
each followed by a monadic operator. operator serves two purposes. First,
it takes simple objects like numeric or character vectors and returns a
sequence. Each element of the resulting sequence is an element of the source
object. This is illustrated in the first example pair. The most common use
should be for preparing numeric data for entry into fields of record sets.
Secondly, as shown in the second example pair, @& marks sequences
such that the subsequent operator (in this case monadic
"-") processes on the sequence elements individually
rather than on the sequence as a whole. Note, I need to parenthesize the
expression to force this contrived example.
At Each (@&)
application of dyadic:. Illustrated here is the behavior of @&
when used in a dyadic sense. Each element of the sequence is delivered as the
left argument to the operator. The dyadic operation is then performed and the
element result becomes an element in the statement result.
At Each (@& ) and
anonymous blocks:. This example is complicated. First I create a reference
to an anonymous numeric object. I call this reference "p". I then use
"p" to accumulate the elements of a numeric vector in an implicit
"at each" loop. With each iteration, "at each" delivers an
element. In the first example, I capture the element explicitly in
"i". I strand "i" with the reference "p" and
create a namespace (without changing the names). This namespace becomes the
argument to the anonymous block where I do the processing. The self explanatory
processing just accumulates the "i"'s in "p". In the second
example, I do the same thing but I don't explicity name the element delivered
by "at each". It's just there, and it looks like I create the
namespace of two things out of just one thing. Having seen the first example,
you know such is not the case. You know about the phantom first element in the
strand. Be aware that since we are dealing with a block here, any objects
created in the block disappear after this operation. This is because they are
in the block's namespace. The block itself is discarded after the operation and
with it, its namespace. It's for this reason that I pass the accumulator
"p" as a reference in a namespace.