Re^4: Why no comments?

[BrowserUk]

what I thought I'd done was to describe in broad terms what I consider to be the attributes of good commenting,

Your opening line is:

"It is indeed a sad state of affairs when programmers fail to properly comment as they write code."

And that sets the tone for the entire post.

For the rest. When I encounter modules that contain that much verbiage I often just delete it without reading it.

Especially when I encounter comments like

  # Reverse sort the set and partitions and insert sums

To "document":

  @$rs = reverse sort numerically @$rs ;
  @a   = ($sa, reverse sort numerically @a) ;
  @b   = ($sb, reverse sort numerically @b) ;

• Reverse: This keyword precedes all three sorts.

  It is redundant.

• sort: as is this.
• the set: if @$rs was @$set, this word becomes redundant.
• and partitons: Ditto these, if @a and @b where @partA and @partB.

  Especially if the sub were called partitionShuffled() rather than hack_C.

• And insert sums: Ditto if $sa and $sb where $sumA & $sumB.

Three lines of verbiage (including 2 blanks) mostly redundant. And completely redundant if the documentation is placed in the only verifiable content of the file--the code.

And

  # Done -- return partitins in required order

to "document"

  if ($a[0] >= $b[0]) { return (\@a, \@b) ; }
                 else { return (\@b, \@a) ; } ;
} ;

• Done --: The end of the subroutine and your done. No shit Sherlock!
• return: And your going to return something. Cool.

  The fact that you use return in the code (twice) is a pretty clear clue.

• partitins: @partA & partB would be so much better.
• in required order: Required for what? By whom?

  You mean partition A then partition B if the sum of A is greater or equal to the sum of B. Or vice versa otherwise.

• (And what's with that floating semicolon after the subroutine?)

Again, mostly redundant, with the code being far clearer than the "documentation".

And the whole lot would be clearer still (and far more efficient to boot), written as:

    # Sort the callers array in-place
    @$set = sort{ $b<=>$a } @$set;

    return $sumA >= $sumB 
        ? ([ $sumA, sort{ $b<=>$a } @partA ], [ $sumB, sort{ $b<=>$a }
+ @partB ])
        : ([ $sumB, sort{ $b<=>$a } @partB ], [ $sumA, sort{ $b<=>$a }
+ @partA ])
    ;
}

[gone2015]

Thanks for the comments.

It appears there is little common ground on which to base a discussion, but since you picked out two examples, I'll address those briefly, and then not waste any more of the monks' time.

For me, the writing of comments is part of the thinking process. The comments you commented on were (if memory serves) written before the code, as I was sketching out what needed to be done. I agree that comments that do no more than repeat what the code says are not only redundant, but can simply be clutter (except where the code is particularly tricky) -- for me the key is to concentrate on the "why". Though I do find it useful to use comments to point up the stages that a block of code goes through.

The first comment you picked out:

  # Reverse sort the set and partitions and insert sums

served several purposes: (a) it was a stage in the subroutine; (b) the algorithm used depends on stuff being reverse sorted, this is key to understanding what is going on; (c) it's doing something a little tricky: in the resulting array entry 0 is not the same as the other entries, it's the sum of those entries. While I was writing it, I thought about all this... I suppose I could have written a longer comment to emphasise the key points...

The other comment you picked out:

  # Done -- return partitins in required order

(sorry about the spellin error)... well, yes, it's not the most significant of the comments in a 2,800-odd line module. Again, the subroutine was broken into stages, so this also serves as the end of the previous stage. The order in which the partitions are returned is key -- the caller can (and does) depend on this.

As it happens, there are three similar subroutines, with the same general shape, and the comments emphasise that. You've commented on the smallest of the three, where there is less to comment on.

[BrowserUk]

For me, the writing of comments is part of the thinking process.

In effect, you are using prose as a form of pseudo-code.

In the past I've tried or used prose, Weiner-Orr Diagrams, Yourdan, SSADM (and derivitives), Jackson, UML, various flavours of Case, and half a dozen variants of pseudo-code to plan code and/or document it. Some with more or less success than others. But in the end, I've found that the language of implemetation is the best pseudo-code (for me!).

I'll often start by sketching out the skeleton of the algorithm in Perl-glish (or C-glish or whatever-glish), and then slowly refine that into code. The advantage is that even at the early stages, I can use the compiler or interpreter to check my progress. When the compiler stops complaining, I'm usually getting close.

it's not the most significant of the comments in a 2,800-odd line module.

Agreed. And it is perhaps a little unfair to pick out so few lines from so many, but to do more entails a lot of work. I did make a start on attempting to re-write the entire module in what I would consider to be a self-documenting style, but then realised it would be equally unfair, as I would be starting from the position of your working example. I also ran out of awake cycles.

The bottom line has to be that there is no 'one right way' to program. But historically, I've found that people tend to be even worse at writing good comments than they are at writing good code. The difference is that I have the compiler to help me sort out the code, but no assistance is available for comments. (Especially a problem for me if they are in German or Italian or Chinese or Spanglish :)

I've personally found it more efficient and effective to put a little more effort into chosing variable and function names that allow the code to be somewhat self-documenting, than to write everything twice. But I realise that that doesn't do it for everyone.

But I do take exception to the implication that my deliberate and considered preference for minimal commenting is born of laziness or a lack of consideration; as nothing could be further from the truth.