Talk:Loops/With multiple ranges
Further explanation needed
I think that it needs of a more specific description in the task part. At start I have got so much confusion because I have had no other experience of multiple conditional DO. The real difficult was about how the several conditions act on the do loop and if they interact among themselves. In a first time I have understood that all the conditions are in parallel/ cohoperation so I have suddenly asked to myselfç are the conditions in AND or in OR chain? Or in other words: all the condition must be true (to have a value in the range written) to run the DO inner code or at least one condition must be true (to have a value in the range written)? Also reading all explanations in the Page of Task and in this Discussion page I have got no clear definition about this question. So I had taken a look at other language codes and so I have got that the conditions of DO loop are in sequence and not in parallel mode. So you run the DO loop using separately range of index value (J) in your PL/I code. Moreover are prod and sum sequential or must they be initializated at each range/condition? I have choosen the first option, but it is not clear for me what is that of PL/I code!--TempodiBasic
You can't expect RC coders to understand your programming language. You shold convert to some form of pseudo-code and explanation in English.
It would also help if you went step-by-step through a single, two and maybe three ranges should produce - in a mix of pseudo-code and English. Thanks. --Paddy3118 (talk) 19:05, 15 September 2018 (UTC)
- I thought that a do loop would be self explanatory (even with such an old language as PL/I, conceived in 1963 or earlier as a possible extension to FORTRAN, announced in 1964, delivered in 1966) with the same degree that when a for loop is used in pseudo-code, but is never described. An astonishing number of pseudo-code on Wikipedia and Rosetta Code uses (a) C (or C-type) language, but I have yet to see a for loop explained/described in the form of pseudo-code. One reason that multiple ranges for the do loop were chosen for this task (along with expressions in the various ranges) was to ensure that short-cuts weren't chosen to bypass the logic and expressiveness of the original code (for maintainability and range expanding, for example), and also to have some method of verification that the various ranges were indeed, executed correctly. One such concern was:
do j=1 to 100, 200 to 300; say j; end;
- ─── being executed as:
do j=1 to 300 if j>100 & j<200 then iterate say j end
- ─── or:
do j=1 to 300 if j<=100 | j>=200 then say j end
- (both of which are slightly harder to comprehend as compared to the original case code, and harder to modify if the original ranges are changed or their "step" (incrementation) values are either modified or added.) Also of note is that in the later two cases, the variable j will have more values assigned to it over the execution of the do loop.
- The main intent was to show how various languages would or could utilize multiple ranges of do loops, not to show how a do loop works. I had assumed that almost everyone knew how do loops worked, it was the use of multiple ranges that was the point of this task. I had assumed that once other popular languages were entered (most of which would probably use a for construction or something similar), that other people would understand how a do loop worked. I had carefully chosen the ranges to not use a range that some languages would behave differently (which would make a nice task addition to Rosetta Code, by the way ─── this is one area that is seldom explored when comparing languages here on Rosetta Code).
- For instance:
n=75 do j=100 to n say j end
- ─── where some languages will not produce any output (not testing the range before execution of the do loop at least once), other languages (such as PL/I and REXX) will test at the start (head) of the do loop.
- However, I'll take a stab at adding pseudo-code to explain do loops as used by the PL/I language. By the way, PL/I isn't my language, it's been years since I even had access to a PL/I compiler, that is the only reason that I didn't enter a PL/I example (here and in other Rosetta Code tasks). If I had access to a PL/I compiler, I'd be entering numerous PL/I examples; PL/I syntax is very similar to REXX's, and indeed, REXX borrowed a lot of PL/I's syntax and PL/I's BIFs. -- Gerard Schildberger (talk) 23:26, 15 September 2018 (UTC)
Hi Gerard, I would first like to say that after first reading the task description:
- I really liked what seemed to me the central idea of the task: looping over multiple ranges with only the single loop body.
- I didn't like the initial reliance on the PL/I code; although PL/I served the crucial requirement of making you come up with the idea, I thought that its variable initialisation and explicit language syntax didn't add to the description.
I was also lazy and didn't try and give an example for discusion of my ideas.
- Here's more of what I was thinking of
Some languages allow for the easy looping around a single block of code using a variable that iterates through several ranges of values in succession. In pseudocode:
LOOP variable J over range(-3 to 9 by 3) then range(30 to 24 by -2) then range(startfunc(x) to stopfunc(x) by stepfunc(x)) BEGIN /* do_loop_body */ print_variable_followed_by_space(J) END; /* LOOP */ print_newline()
The PL/I code could be rewritten in such pseudocode, or the task itself altered to ask for not much more than what is stated above. (I would need to add that ranges are inclusive of endpoints and what the three funcs do when computing a ranges limits).
- First, I used the initializations of several variables to show that values (and expressions) should be able to be used within the constructs shown. This by-passed the static or hard-coding of values within the iteration ranges. -- Gerard Schildberger (talk) 09:37, 16 September 2018 (UTC)
- I definitely wasn't thinking of an easy way to loop around a block (or blocks) of code, if fact, this is exactly what I was trying to avoid, that is, avoiding logic which avoids going around code (to avoid execution of some code ─── this seems/appears to an anathema to structured programming) by some kind of artificial construct, or even using logic to chose which block to execute (or not execute). Specifically, I wanted a "pure" do construct (or for construct) if possible. I was only interested in methods that directly stated what was intended (that is, if it was possible to write code in that manner). I had, in fact, originally wrote the REXX version to do exactly what the PL/I language used, but it involved writing a function to essentially parse (scan) for the appropriate clauses (to, by, which, for, until, ... in any order) and execute the do index variable accordingly ─── but the amount of code seemed to be a tad bit bulky and had to deal with too many (error/omission) issues, and it made the program hard to follow and understand, defeating the purpose of a (clean-looking) do loop with multiple (iteration) ranges. I may end up including it as an alternative solution with the usual "your mileage may vary" caveat. I have often come across the need for a multiple-range do loop structure, and once you've used it, it comes very natural to think (and code) in that manner, --- but in any case, it's easy to program around any deficiency or short-coming of a language, but I wanted to see how other programming languages dealt with such an animal, or failing that, how programmers dealt with that type of logic to perform the same function/need. One thing that I definitely wanted was to use expressions instead of hard coded start/stops/increments (or however one wanted to express such a range construct). I think using functions (as one form of expressions or within expressions) really muddies up the waters (but it covers the ground!) when writing pseudo-code, even in a general manner such as this. I had assumed that programmers could understand the use of functions instead or embedded within an expressions, but I didn't think that would make it any clearer. Also note that I didn't assume that many languages supported such a structure, so part of the Rosetta Code's task requirement stated that other "work-arounds" may be used if do loops with multiple ranges weren't possible. I also had to deal with several types (kinds) of do loop (incrementing/decrementing) ranges, and also the "step" or by values, and to make sure that expressions could be used instead of hard-coded values, and to also ensure that every computer programming solution did the exact index enumerations, so I had to Jerry-rig a method to validate the do index (in this case, j). I had to add a somewhat complicated if statement so that the product wasn't zero, and also so that the product didn't exceed the limit(s) of an integer for most programming languages. I was going to have the programming examples just display all the values of the j index, but it takes a lot of eyeballing to verify a list of numbers (even a horizontal list), whereas observing just two values makes it easy to verify that the do loop index was incremented and/or decremented correctly. I threw in the commatizing the two (output) results as a more-or-less optional requirement for programmers that wanted to walk an extra few paces ─── I always appreciated those Rosetta Code tasks that had an optional or "extra" credit thingy. -- Gerard Schildberger (talk) 09:37, 16 September 2018 (UTC)
other DO loop cases
Another good tasks for Rosetta Code would be the case of how a language treats the following code snippets: <lang>k= 5 L= 2
do j=k to L say j end</lang>
and show what values (if any) are displayed.
Or, another case: <lang>k= 5 L= 2
do j=L to k by -2 say j end</lang>
Or, yet another case: <lang>k= 5 L= 2
do j=L to k by -2 say j k= 9 end</lang>
- You could add several more.
a DO with no TO a DO that increments not one variable, but a list of variables a DO with no BY a DO that stops and stops on the same value (it is within range), but when is the index checked?