I've never seen a language's style guide recommend avoiding comments before

Comments are useful to describe __why__ you're doing something, often when you are not able to change the unexpected behaviour. Whenever I build an API library, my code is littered with comments like "Acme Corp API requires this happens before that" with a link to that bit of the API documentation.

Here's a C++ example about "documenting surpises" (taken from Steve McConnell's Code Complete:

    for ( element = 0; element < elementCount; element++ ) {
        // Use right shift to divide by two. Substituting the
        // right-shift operation cuts the loop time by 75%.
        elementList[ element ] = elementList[ element ] >> 1;
    }

    /* The following code is necessary to work around an error in
       WriteData() that appears only when the third parameter
       equals 500. '500' has been replaced with a named constant
       for clarity. */

    if ( blockSize == WRITEDATA_BROKEN_SIZE ) {
         blockSize = WRITEDATA_WORKAROUND_SIZE;
    }
    WriteData ( file, data, blockSize );

Unfortunately, having been doing this trade for 30+ years, I've found most people write crappy code in a hurry to try to hit some deadline based on incomplete requirements and confusing business rules. A few precious comments stuck in there can help the next guy, months or years later, figure out what the heck your original intent was or why a block of code exists at all.

As I get older, I find it helps me remember what I was doing.

One of my software developer friends was fond of saying: "the worst code I ever saw was my own!"

YES if your code is clean and elegant and well named and clear you don't need to explain anything in common language.

YES you should strive for such.

However, the REALITY is your code sucks and no one is going to want to have to figure out what the heck you were doing. A few comments would really help.

The next reality is that the typical developer may be literate in a dozen or more languages and the language du jour that you coded so elegantly in has fallen out of favor and no one remembers those dusty corners you so beautifully exploited to make something work.

Comments would help even more if you learn to use some basic grammar and spelling when you create your comments. (It doesn't have to be literature, but try to make your comments as readable as you think your code is - please!) Nothing will turn another developer off to trying to decipher your code than a few comments that make you look like a moron.

Style guides that eschew comments, IMO, are counterproductive. They feed on the developer's ego and disregard reality.

Comments cost essentially nothing to add to your code and can save it from an early death and complete refactoring by the next guy who comes along.

- The proper use of comments is to compensate for our failure to express ourself in code. Comments are always failures. We must have them because we cannot always figure out how to express ourselves without them, but their use is not a cause for celebration. So when you find yourself in a position where you need to write a comment, think it through and see whether there isn’t some way to turn the tables and express yourself in code.

- The older a comment is, and the farther away it is from the code it describes, the more likely it is to be just plain wrong. The reason is simple. Programmers can’t realistically maintain them.

- Comments Do Not Make Up for Bad Code! One of the more common motivations for writing comments is bad code. We write a module and we know it is confusing and disorganized. We know it’s a mess. So we say to ourselves, “Ooh, I’d better comment that!” No! You’d better clean it! Clear and expressive code with few comments is far superior to cluttered and complex code with lots of comments. Rather than spend your time writing the comments that explain the mess you’ve made, spend it cleaning that mess.

``` // Bad:

    // Check to see if the employee is eligible for full benefits
        if ((employee.flags & HOURLY_FLAG) &&
        (employee.age > 65))

    // Good:
    if (employee.isEligibleForFullBenefits())

    /**
     * Frobnicates a foobar
     * 
     * @param Foobar $foobar The foobar to be frobnicated
     * @param int $intensity The intensity with which the foobar will
     *    be frobnicated (defaults to 4)
     * @return mixed The result of frobnicating a foobar
     */
    function foobar_frobnicate(Foobar $foobar, $intensity=5)
    {
        // frobnicates the foobar
        return $foobar->frobnicate($intensity);
    }

Pretty sure I've been guilty of this in the past, too. As I recall, the documentor tools make a lot of noise if you don't supply wasteful and irrelevant values for every single little thing even if it's blindingly obvious from the symbol, context or idiom what it means and what it does.

http://www.haskell.org/haskellwiki/index.php?title=Commentin...

It isn't official in any sense.

One of the best cases for comments IMO is documenting an unexpected behaviour on the part of a third-party API. But even then, correct exception / error-handling code can obviate the need for comments in many cases.

If I'm reading code (from an experienced programmer) and I see a comment, I immediately pay attention, because Here Be Dragons.

    shouldPickWorkUnit :: (ImportId, WorkUnit) -> STM Bool
    shouldPickWorkUnit (k, u) =
      case hostNameFor url of
        Nothing       -> return True -- Invalid URLs are fast to process.
        Just hostName ->
          takeWorkThat'sAlreadyDone <*>
            (don'tTakeSomeoneElse'sWork <*>
              don'tExceedTheRateLimitFor hostName)

In fact, the remaining comment can be factored away too:

    shouldPickWorkUnit :: (ImportId, WorkUnit) -> STM Bool
    shouldPickWorkUnit (k, u) =
      takeWorkWithInvalidUrl <*>
        (takeWorkThat'sAlreadyDone <*>
          (don'tTakeSomeoneElse'sWork <*>
            don'tExceedTheRateLimitFor hostName))

I don't want to run my internal compiler in my head when i'm reading your code, so you better make sure there's at least a docblock above every function that describes in 2 sentences what it does so I can get a global overview of what the heck this file is doing.

People that suggest that 'the code is the documentation' are always forgetting that reading code is way more taxing on the brain than reading english.

Code comments definitely fall into this category. I've worked with developers who I greatly respect who are obsessive about code comments. I've even been told that the comments are more important than the code and in that specific context it made sense.

But my own experience and biases make me think code comments are a problem. I like to refer to them as future lies. There is virtually no back pressure on comments to keep them in sync with the code. There is no automated way to verify them and refactoring tools on comments are rudimentary at best. To put it simply, I no longer trust comments and will usually ignore them in order to verify the code itself. I can't count the number of times I've found comments that directly contradicted the code it was commenting. It isn't even uncommon to find comments that are incorrect when they are written!

As to the folks recommending comments that document the "why" of a piece of code, I'd counter that if you have a "why" you have a specification. If you have a specification it should be verified in a systematic way. So performance improvements, or specific client requirements should be encoded in tests so that they don't regress. Comments do not provide that safety.

That's not to say I never comment my code. Just that it always feels like a failure when I do. It is usually because it is cheaper to comment than to provide cleaner code or better verified specifications.

Look at the Backbone.js annotated source [1] (the stuff on the left is just the comments pulled out from the original source JS). The comments make it much, much quicker to grasp what's going on, even though many of them just state exactly what the corresponding code does, which according to the Haskell docs' advice is pointless and to be avoided.

[1] http://backbonejs.org/docs/backbone.html

This is golden!

    ...thus they (comments) tend to diverge from actual implementation.

It allows you to document the intended inputs and outputs of the function and state its purpose. This increases maintainability and reusability.

Functions themselves should be short and written as a sequence of logical steps.

I'm also a big fan of doing things right rather than just hacking until it works, which seems to put me in a minority.

"In this book we don't use many comments; we try to make our programs self-documenting by using descriptive names."

http://mitpress.mit.edu/sicp/full-text/book/book-Z-H-15.html...

  PendingResult<MessageApi.SendMessageResult> pending = Wearable.MessageApi.sendMessage(mGoogleApiClient, mWearableNode.getId(), path, data);
  pending.setResultCallback(new ResultCallback<MessageApi.SendMessageResult>() {
      @Override
      public void onResult(MessageApi.SendMessageResult result) {
          ....
      }
  }

  wearable.sendmessage(mywearable, path, data)
  
  @wearable.onresult
  def onresult(result):
    ...

It's crazy how much time I spend on them and sometimes they are confusing because the specs change and I keep forgetting to add/remove/update something in a docblock when I rewrite some part of the code it documents (I'd say between 10-20% of my commits are docblock updates).

I believe brief or even no documentation may be the best approach until you are on the verge of releasing a stable version.

While you are on developing and testing mode, it seems a better idea to forget about comments and focus on modularity.

After all, Why would I need comments if all the rest of my team sees is an interface satisfying a previously established and well defined contract? That's why docblocks should be only documenting public interfaces, some kind of dump taken from the part(s) of the contract they implement.

I'd consider instead other top priorites on those phases:

- Keeping an homogeneous codebase in regard of design and coding guidelines and conventions.

- Re-factoring before it becomes a problem

- Writing neat unit and integration tests

And, the most important:

- Keeping a channel open with the client, constantly feeding guided demos and prototypes showing your progress to make sure you are on the right track and you didn't get it backwards, updating specs and being realistic about what can be done and what not in which time frames with the provided resources.

An official style guide? Maybe. But it is one of common philosophies. See for example:

"If you need to comment something to make it understandable it should probably be rewritten."

http://kotaku.com/5975610/the-exceptional-beauty-of-doom-3s-...

Comments often go out-of-sync with the code, so I think it makes a lot of sense to prefer writing comprehensible code instead of trying to explain with comments something totally incomprehensible.

I wrote more about this here (it's Lua code, but it applies to any language):

http://kiki.to/blog/2012/03/16/small-functions-are-good-for-...

    -- swap the elements of a pair
    swap :: (a,b) -> (b,a)

    let b=a+1   -- add one to 'a'

Does it mean that every piece of code can be expressed as clearly as in a one-line comment in natural language? I don't think so.

What the article omits is the suggestion of commenting the right way, i.e. adding reasoning or the description of the high level logic behind the code.

Comments may be a code smell when it is necessary to explain what your code is doing.

Anywhere where you have some freedom to solve a problem one way or the other it may clarify why the implemented approach was chosen.

eg.

// The reason we use a custom swap function instead of

// the one that is shipped with the framework is because

// of an edge-case that occurs quite frequently in our

// scenario

// Refer to Change Request 345.

void Swap (Foo a, Foo b) ....

// get the user $user = $this->getUser();

Times that by the thousands of lines in a project and you have one big headache!

It may be a good idea to comment "what" the code does, if it isn't clear (the code itself is "how" it is done, but "what" does it do may be hard to read, e.g. sometimes you use a clever hack for performance reasons).

As always, handle with care :-)

I know that's absurd in practice, we don't always have the time but I've always used comments as a last resort.

If I need to comment code to make it understandable at a glance, so be it but I'd rather avoid them all together and rewrite until it's clear enough without them.

Let's say you are parsing a standard tab delimited file. You find that the tab delimited file has some non-standard features, so you have to write some extra lines of code to handle it. For people who thinks the code just parses a standard tab delimited file, these lines will be confusing, so you comment these lines and say why you included them.

I do favor putting a short comment on some methods/functions/procedures.

Assume someone who can't write clear code can write clear comments. Also search and replace clear with "readable" "literate" "concise" and last but not least, "correct"

Assume a programmer has full authority over all 3rd party, supplier and customer APIs, interdepartmental processes, and all business logic, management selected fad technologies, such that its logically impossible to be unable to always factor out weird confusing stuff resulting in clear code / clear comments. My program is the world and none have dominion over any of the rest of it and any other conception of reality is wrong. (And edited to add I've gotten involved in some weird "EE" stuff and like it or not, the world itself is plain old weird and illogical sometimes and if you don't like that, a computer programmer can't fix it, only a physicist, or maybe a diety. This isn't a big problem in the world of CRUD apps but it does happen)

Assume comments only exist as a inspirational descriptional prose tool. Sometimes I use them as placeholders for something I know belongs there but either I or the business are not ready. Sometimes I use them as a cheatsheet because I'm personally really uncomfortable. Sometimes I use them as an outline more like names on a map to orient myself than a travelogue.

Assume all programmers fit the management ideal of identical replacable cogs. "How could someone work here without knowing by heart how to convert dBmW into volts or the difference between S21 and S12 microwave scattering parameters, so I have no need to comment this, but I've never actually used this corner of matrix math while employed before so I'll make one of those laughable comments that is a simple linear translation just to help me keep my head on straight.

Assume comments go thru the same code review process as code. If a comment in file A tangentially relates to function Q in file B, and you modify function Q, your code review process will probably examine file B and the comments in it, but how do you ensure file A gets modified? This is especially bad with those "because" style comments. (edited to add, at least date your comments?)

Assume no metrics exist WRT comments to be gamed. Your continued employment and possible promotion exist because of a content free meaningless metric number, perhaps lines of comments. Ask a professional to generate a number, you'll get a nice number, but unprofessional work. Ask a professional to do professional work, and you get professional results and who cares what the number is. That requires a high caliber of management, usually unavailable. Even worse a low caliber of management, the kind most likely to demand adherence to meaningless metrics, is also exactly the type least likely to successfully evaluate the professionalism of the code so they don't end up with good code. So you get meaningless metrics resulting in meaningless comments right next to bad code, if you enforce metrics.

Assume there exists a silver bullet for comments, just like this months silver bullet fad for code also fixes all problems.

(edited to add) Assume there's one human language. I worked at a place where outsourcing and H1B took complete control over corporate IT such that code comments and even some internal documents were no longer written in English. This makes comments rather hard to follow when engineering tries to cooperate with IT. So... I'd love to follow your detailed internal process for dynamic DNS for my spectrum analyzer, but you guys don't use English and we don't use your India language, so...

    swap :: (a, b) -> (b, a)
    swap (x, y) = (y, x)

    map :: (a -> b) -> [a] -> [b]
    map f [] = []
    map f (x:xs) = (f x):(map f xs)

On the other hand, at 10,000 lines of code, a lot of that being parochial business logic, I'm going to want high-level documentation of why all this code exists. My emotional impulse is going to be to throw out all this shit code (in the business world, all code is shit) so please tell me why that is a bad idea. (I know it is, and I'm not going to do it, but please tell me why I'm not going to do it.) I'm going to want an entry point. I can't count the number of days of life I've lost just looking for entry-points in gigantic enterprise codeballs. Like, what actually runs?

Actually, 10,000-line single-programs should be rare-- Big Software is almost always a mistake, see here: http://michaelochurch.wordpress.com/2012/04/13/java-shop-pol... but that's another rant.