Ifs and &&s and Plan 9's Source Code

http://computationallyendowed.com/blog/2012/12/03/ifs-and-ands-and-plan-9s-source-code.html

by jneen • 13 years ago

190 137 13 years ago

jacobparker2• 13 years ago

C++ has "and"/"or"/"not" etc., so you can do

    if(not A and B)

but I'm not sure I've ever seen anyone do it. http://en.cppreference.com/w/cpp/language/operator_alternati...

To get them in C, see the note about iso646.h in the link.

kostya-kow3• 13 years ago

I don't think it really improves readability. All the programmers are familiar with && and ||, and using something else will just confuse people.

Also, it's not really that hard to #define it yourself, instead of using includes.

jacobparker2• 13 years ago

I definitely wouldn't use them because they are unfamiliar. However, in my head I pronounce "&&" as "and" and "&" as "bitand", so using and/bitand over &&/& would reduce a depressingly not-uncommon typo for me. However, on balance, not worth it to me.

Note that in C++ they are keywords which is superior to macros.

orangeduck• 13 years ago

As an idiom I quite like this form. I often do similar for double iterations:

   for (int x = 0; x < width; x++)
   for (int y = 0; y < height; y++) {
      do_something(x, y);
   }

slg• 13 years ago

This also provides an advantage when debugging. It will become immediately obvious which condition fails when stepping through the code. That isn't always the case with a long string of &&s.

Groxx4• 13 years ago

^ this is a detail which boggles my mind. why are our debuggers still line based? they're clearly not in every respect since you can basically always 'step in'to an && sub-expression, but nothing displays progress as you step through such things, nothing lets you put a breakpoint at some sub-expression, nearly every feature of every tool is delineated by lines as if they're the important part of a program.

darrenkopp• 13 years ago

With visual studio and c# you can put a break point at a certain point in many expressions and while in the debugger you can have it evaluate expressions individually when hovering with the mouse.

Stack traces, however, are still line based and thus if an exception occurs (like null reference) you only get the line, not the statement.

rayiner2• 13 years ago

Because something you do relatively early in the compiler is throw away all of the structure of the source, including flattening nested expressions into a linear IR. Mapping back to line numbers in the debugger is a bit hacky to begin with, and mapping back in an even more fine-grained way would be more complex still.

leoh• 13 years ago

Not all debuggers are line-based. Many of LLVM's tools give descriptive errors ("expressive diagnostics").

Here's an example of gcc versus clang (a "frontend" for LLVM):

  $ gcc-4.2 -fsyntax-only -Wformat format-strings.c
  format-strings.c:91: warning: too few arguments for format
  $ clang -fsyntax-only format-strings.c
  format-strings.c:91:13: warning: '.*' specified field precision is missing a matching 'int' argument
    printf("%.*d");
              ^

LLVM is does some cool stuff :). Some other nice examples are at http://clang.llvm.org/diagnostics.html.

beemoe• 13 years ago

Very interesting point! There is an argument that lines are a unit of human comprehension and complicated expressions should be broken onto different lines already, as a matter of readability. It makes some sense to let the granularity at which you debug reuse the granularity at which you read.

jwatte• 13 years ago

Gdb for g++ may be; msvc for c# is not! Gdb usability and performance is a (THE) major shortcoming of the gnu tool chain.

kabdib• 13 years ago

I've worked with debuggers that allow you to step the short-circuit operations.

That said, a huge cascade of them is something I try to avoid. Two or three, okay. Nine or ten, break it up and make it clear and readily debuggable.

jlgreco2• 13 years ago

This also makes individual conditions slightly easier to comment out.

vy8vWJlco2• 13 years ago

And yank/cut/copy, and re-order, and type-over, and... I have developed similar habits after using line-based editors. This is very easy code to modify. (Notice the return type on a line all its own.)

Groxx4• 13 years ago

I hadn't considered this before... interesting!

RBerenguel• 13 years ago

Thought the same. If I need three && in an if condition, I nest (although I add brackets, I prefer having them). It makes easier debugging, easier reading and in case of need they are a good place to put some printfs

jejones3141• 13 years ago

Ouch. I'm pretty sure that the compiler I worked on 22 years ago could stop at sequence points, e.g. after the evaluation of the LHS of && and ||. (OK, actually it was a superset of those, called "gesornenplatz points" in-house, but that's another story.)

Ignoring the "no space after if/while/for" issue, I'd suspect that code to not have been the author's intent if I just came across it.

JoshTriplett• 13 years ago

If not for the need to get and later free the

    Dir *b;

, and print the message, this condition would get a lot simpler. I'd find it more readable to write a function

    bool samefile(Dir *a, Dir *b)

, have it consist of a single-line return statement, and then have the caller just do if (samefile(a, b)) { ... }.

jamesaguilar3• 13 years ago

Good luck using an auto formatter on a code base that uses this technique.

adrianmsmith2• 13 years ago

I believe that the layout of code is done by humans for humans. An auto code formatter destroys all of that meaning.

I blogged about this more here: http://www.databasesandlife.com/do-not-use-automatic-code-re...

jamesaguilar3• 13 years ago

Yeah, I saw that. I think it's misguided. The amount of time spent on arguing about style far exceeds the time savings of the changes you propose. People can get used to different formats quite easily, but time spent arguing about and maintaining formats can never be recovered.

Also, your formatter looks pretty brain-dead. clang-format formats all of those cases (in C++) as in your green examples with Google style turned on. Even the Eclipse formatter should do a better job than what you've posted with the appropriate settings. When I advocate an auto-formatter, it doesn't make sense for you to respond with a blog post talking about a formatter seemingly designed to make everything worse.

webreac• 13 years ago

When working in a team, human code formatting can cost a lot of time. The rule is generally to have an auto code formatter that runs when we do a commit. The choice of formatting rules is made collectively and can be discussed. The usage of an auto formatting is generally mandated by society policies. I think it is a huge time saver (not using an auto formating tool is way to lose a lot of time).

andkerosine• 13 years ago

For what it's worth, Vim handles it just fine.

calebegg2• 13 years ago

What? Not for me, not with cindent on. It indents each if statement by one level, just like it "should". How would it somehow know to not indent them?

supersillyus• 13 years ago

If I'm not mistaken, Plan9 comes with a code formatter, and this idiom is common in Plan9, so it seems likely that it is supported. http://plan9.bell-labs.com/sources/plan9/sys/src/cmd/cb/

ernesth2• 13 years ago

When you use the fact that && is not the logical conjunction, you'd better use another if.

    if (b != nil)
    if (b->bla == a->bla)

is so much clearer (and language agnostic) than

    if (b != nil && b->bla == a->bla)

piokuc• 13 years ago

I've never heard about writing language agnostic code in my life, lol!

rowborg2• 13 years ago

One downside to non-braced conditionals is that a semicolon accidentally placed after the conditional will cause the block to always run, e.g.:

  if (null != foo);
    bar();

This is valid code in C and Java, and bar() will always run in this case.

Having seen people waste hours on such a semicolon, I always use braces, even in one-liners, because I never know when someone is going to break it out into multiple lines later:

  if (null != foo) { bar(); }

wikwocket• 13 years ago

I was going to say that a compiler should issue a warning for this, as you'd almost never want a semicolon right after an if condition, but to my surprise Eclipse doesn't seem to flag it.

It does however indent the line after the semicolon to the same level as the if, which is at least a red flag that something is up, if you are used to how the auto-indenting normally works.

rowborg2• 13 years ago

Absolutely, indentation helps catch this when writing.

When I've seen this happen, it wasn't because of a semicolon added when the code was written. It was someone accidentally adding a semicolon to a line later, without realizing it. Unless they then went to the next line and hit the "fix indentation" key, they didn't catch it.

comex2• 13 years ago

Both clang and gcc warn on "if(1);", clang by default, gcc with -Wextra. clang also warns on "while(1);" - I think this is somewhat obnoxious, since it can be useful, and would prefer if it only warned if the semicolon was followed by an opening brace, but YMMV.

thedufer2• 13 years ago

I don't think braces solve that particular problem:

    if (null != foo);
    {
      bar();
    }

pavanky• 13 years ago

I agree. This is we do

   if (null != foo) bar();

   if (null != foo) {
       bar();
   }

jbert• 13 years ago

It does if you use the One True Brace Style:

http://en.wikipedia.org/wiki/Indent_style#Variant:_1TBS

since your eyes would flag:

    if (null != foo); {
       bar();
    }

as badness.

[edit: correct } to {, ta]

danbruc2• 13 years ago

I personally have never put a semicolon after a condition and being mainly a C# developer I have written a lot of them. Is this really something that happens more than once or twice in a lifetime?

ansgri• 13 years ago

Yes, but such mistake, once made, is very hard to notice (because you never do this, you don't look for it), so it may well lead to a fatal disaster. Hence 'once in a lifetime'.

SoftwareMaven2• 13 years ago

I know, let's start a language war thread: "That's why I use Python." ;)

More seriously, after more than 10 years of writing C, C++ and Java, I've never run into this problem. I still wrap mine out of habit (to prevent this dreaded occurrence), but I think good testing would obviate any need for this.

jbverschoor• 13 years ago

I often use: if (false) { } else if (something) { } else if (something) { } To get some indentations right.

When constructing UI elements in code, I used to scope copy/pasted blocks: { Button button = xxx; // and some more }

{ Button button = xxx; // and some more but different }

In the rubydays it'd be: Button.new do |b| something end

triztian• 13 years ago

How about using this alternative form:

  if((b != nil)
  &&(b->qid.type==a->qid.type)
  &&(b->qid.path==a->qid.path)
  &&(b->qid.vers==a->qid.vers)
  &&(b->dev==a->dev)
  &&(b->type==a->type)){
      fprint(2, "cp: %s and %s are the same file\n", an, bn);
      ret = 1;
    }

It keeps almost the same visual look and it uses the common convetion, except for the && at the beginning of each line.

phoboslab• 13 years ago

I still prefer this:

  if(
  	b != nil &&
  	b->qid.type == a->qid.type &&
  	b->qid.path == a->qid.path &&
  	b->qid.vers == a->qid.vers &&
  	b->dev == a->dev &&
  	b->type == a->type
  ) {
  	fprint(2, "cp: %s and %s are the same file\n", an, bn);
  	ret = 1;
  }

IvyMike• 13 years ago

We can go deeper...

  if(
  	b != nil &&
  	b->qid.type == a->qid.type &&
  	b->qid.path == a->qid.path &&
  	b->qid.vers == a->qid.vers &&
  	b->dev      == a->dev &&
  	b->type     == a->type
  ) {
  	fprint(2, "cp: %s and %s are the same file\n", an, bn);
  	ret = 1;
  }

(Lined up the "a"s to make it obvious that they're all the same.)

sbmassey• 13 years ago

Putting the &&'s at the beginning of each line makes the overall shape of the logic expression easier to percieve: you can prove they're all one big 'and' expression without having to hunt for the end of each line.

Avshalom• 13 years ago

  if(b != nil){
  if(b->qid.type==a->qid.type){
  if(b->qid.path==a->qid.path){
  if(b->qid.vers==a->qid.vers){
  if(b->dev==a->dev){
  if(b->type==a->type){
    fprint(2, "cp: %s and %s are the same file\n", an, bn);
    ret = 1;
  }}}}}}

I will never understand the allergy to multiple braces on one line.

notaddicted• 13 years ago

I think the aversion is because the way that they line up visually is the opposite of the way that they are matched by the parser. I think everyone needs a dose of LISP to get over any problems they have with any frequency or arrangement of brackets/parens/braces. Anyway, having a function that compares two structs and returns 0 or 1 with a printf out to stderr is just gross in my opinion.

chj• 13 years ago

Using ifs is definitely more readable than &&, words always win. However carefully alignment of && and ( can make things better. Note that the comment after "if" is necessary to keep the shape balanced.

    if( /* b is same as a */
           (b != nil)
        && (b->qid.type == a->qid.type)
        && (b->qid.path == a->qid.path)
        && (b->qid.vers == a->qid.vers)
        && (b->dev == a->dev)
        && (b->type == a->type)
    ){
        fprint(2, "cp: %s and %s are the same file\n", an, bn);
        ret = 1;
    }

dllthomas3• 13 years ago

"words always win"

Add seven to three then multiply the whole thing by twelve.

(7 + 3) * 12

moron4hire3• 13 years ago

only if you treat symbols as a foreign language

kostya-kow3• 13 years ago

It looks weird without tabbing. But if the code was tabbed, then it would take too much space, since they are using huge tabs. That's why I prefer 3-space tabs. Not too small, not too big.

zalew• 13 years ago

3 tabs? is there a language, community or editor that encourages this convention? usually the default for tab is 4 or 2, and 4 is the PEP convention for Python, while 2 is very commonly used around Ruby as far as I have seen.

btw there are no tabs specifically because it's meant to be read as a single statement, not nested conditions.

dllthomas3• 13 years ago

Following typical layout rules, there's a much easier way to get this conjunction without indents:

if(foo == bar

        && baz == qux

        && ! frob

        && whatever()

        && something) {

adrusi• 13 years ago

I use this convention often for loops:

    for (int x = 0; x < width; x++)
    for (int y = 0; y < width; y++)
    if ((x + y) % 3) {
      // ...
    }

The semantics are kind of like using a comprehension.

cnvogel• 13 years ago

I abhor deep indents, and when I have deeply nested loops that just serve to trivially enumerate things, in C I sometimes add a helper function for the iteration.

e.g. instead of:

    for(int i=0;i<N;i++){
        for(int j=0;j<K;j++){
            for(int k=0;k<L;k++){
                  ...

it might become

    i=j=k=0;
    while(all_frob_indices(&i,&j,&k)){
        ....

with all_frob_indices() doing the i++; if(i>N){ i=0; j++ }...

This makes code look more similar to e.g. itertools-constructs in python where you can easily make products, zips, ... from iterables.

barrkel• 13 years ago

Doing this stops most compilers putting the index variables in registers, with knock on effects on array indexing efficiency.

Not necessarily a bad practice, but something to be aware of.

ufo• 13 years ago

The biggest problem I see in doing this is that since C++ does not have yield statements / coroutines / continuations you need to write the iterator code "inside out" and with explicit state, something that usually ends up harder to read than a regular nested-for-loop iteration.

JadeNB• 13 years ago

Notice that your own post illustrates why this is a bad idea; whereas

    for(int i=0;i<N;i++){
    }

correctly steps through `N` `i`-values, a version with your `all_frob_indices` would step through `N + 1` `i`-values (because your termination check is `i > N`). Of course, it's easy to fix this once it's spotted, but you won't spot it with the same ease that you'd spot an off-by-one error in a `for` loop, because your eyes (or at least my eyes) don't know how an `all_frob_indices` function should look as well as they do how a `for` loop should look.

mixedbit• 13 years ago

As a side note:

     if(b->qid.path==a->qid.path)

is interesting. This is not how you compare strings in C. So it is either a bug or dirstat() needs to guarantee that different results pointing to the same file always share the path string.

jedbrown• 13 years ago

This line is comparing integers, not strings.

          Qid is a structure containing path and vers fields:
          path is guaranteed to be unique among all path names cur-
          rently on the file server, and vers changes each time the
          file is modified.  The path is a long long (64 bits, vlong)
          and the vers is an unsigned long (32 bits, ulong).  Thus, if
          two files have the same type, dev, and qid they are the same
          file.

http://man.cat-v.org/plan_9/2/stat

hcarvalhoalves• 13 years ago

Amazing how old C, and still obscure things always pop up here in HN. I had no idea of such idiom.

ebbv2• 13 years ago

Oh god that code is fucking hideous.

If someone who worked with me wrote that I'd talk to them about it and make sure they never did anything like that ever again.

From the terrible argument names to the abuse of the single line if syntax (which should never be used anyway, always use curly braces.)

orangethirty• 13 years ago

Thats hideous? I see worse code everyday.

ebbv2• 13 years ago

Just because there's lots of worse code doesn't make that good.

There's worse songs than "Who Let the Dogs Out" but it's still a really bad song.

cheald• 13 years ago

You would love this:

https://feralbynight.googlecode.com/files/FeralbyNightv3_2_b...

danbruc2• 13 years ago

11k lines in one file - this is the way it should be done. ^^

encoderer2• 13 years ago

Maybe it comes from using Python as my go-to language (and that it's my favorite language) but i personally like to avoid non-essential braces and other minutia.

Yes, of course, I know the argument: a one-line bracketless "if" can set-up a future developer for failure if they need to add an item to the conditional block. And if they for some reason decide not to read the actual conditional. And if they don't test it.

And I don't hate code that uses braces even when they're not strictly needed, but I personally prefer to omit them. And my feeling is that the dogma around braceless-ifs is a little overblown.

Of course when working on a team that has adopted a no-braceless-if policy, I conform. Having consistent code is way, way more important than somebodies own favorite bracing style.

Dylan16807• 13 years ago

Odd, since python is all about using the best way and not variants, that you would want to have some with braces and some without braces.

pmarin• 13 years ago

I find this code pleasantly readable.

How about this one? http://git.savannah.gnu.org/gitweb/?p=coreutils.git;a=blob;f...

dasil0032• 13 years ago

Makes me wonder what the quality of the architecture is on your projects is though. Not that I think coding style is unimportant, but it's all too easy to criticize yesteryear's code with today's style standards and totally miss the forest for the trees. Are you doing work that measures up to the ambition of Plan 9?

pjmlp3• 13 years ago

We all know how well that ambition turned out.

tzs• 13 years ago

What did you find wrong with the argument names?

papsosouid• 13 years ago

It is always the worst developers who fixate on arbitrary rules because they never developed the ability to read code. Go practice reading code, that's the only way to get better at it.

jff• 13 years ago

It's a good thing they didn't use a goto, or this guy might have had an aneurysm.

__david__• 13 years ago

No. Single line if syntax is good. Gratuitous curlies make the code longer and ultimately harder to read (since you can't fit as much on a page). Being able to see your whole function/algorithm at once is more important that almost any other consideration.

vy8vWJlco2• 13 years ago

You may find astyle and indent personally useful if/when you go hoarse (hackers aren't really known for their conformity).

http://man.cx/?page=astyle

http://linux.die.net/man/1/indent

pieguy• 13 years ago

And what's the alternative for if(A && B){ x(); }else{ y(); }? Seems very easy to shoot yourself in the foot with this approach.

anonymous• 13 years ago

Well, you could

    if (one) {
    if (two)
    if (three)
    if (soon) {
        stuff
    }} else {
        else stuff
    }

Personally, I think I won't have any trouble reading that, but I have noticed I'm somewhat more tolerant than others in this regard. Though I am more OCD than others in other ways.

thedufer2• 13 years ago

That doesn't what he's talking about, though. Your else will run anytime `one` is false, not when `one && two && three && soon` is false.

xuhu• 13 years ago

But it's still not the equivalent of: if (one && two && three && soon) {stuff} else {stuff} which I assume the GP meant.

ernesth2• 13 years ago

Here, the x() and y() are assignments, so it's easy:

    ret = 0; //y()
    if (A)
    if (B)
        ret = 1; //x()

If the else statement is more involved, it gets difficult.

danabramov• 13 years ago

There is a similar [common][1] idiom with `using` in C#. When you need to create and later dispose several resources, to avoid excessive nesting you write

    using (var outFile = new StreamReader(outputFile.OpenRead()))
    using (var expFile = new StreamReader(expectedFile.OpenRead())) 
    {
        ///...
    }

[1]: http://stackoverflow.com/a/1329765/458193

OldSchool2• 13 years ago

Perhaps the compiler wasn't relied upon to provide "short circuit" boolean eval? This code will compile that way no matter if it's available in the compiler or not.

Honestly for the sake of being more robust, I'd add if(a != nil) after the first test of b.

if(b != nil) if(a != nil) if ...

SoftwareMaven2• 13 years ago

If the compiler can't do short-circuit evaluation, the compiler can't do C compilation.

glurgh2• 13 years ago

Short circuit is guaranteed by the language standard. On top of that, plan9 comes with its own C compiler so the people who wrote this also wrote the compiler used to compile it.

OldSchool2• 13 years ago

I see, I just have no idea how complete the Plan 9 tools are. Working with in-house domain specific language interpreters or compilers I've twice seen unary minus support missing so I assumed anything was possible when something's not fully commercialized.

endgame• 13 years ago

This isn't a new thing. I remember it being used in ZZT-OOP, where you didn't have boolean operators.

    #if foo #if bar #send obj:dosomething

carlob• 13 years ago

In Mathematica the FullForm of a && b is And[a, b]. You can use this to format long lists of conditions as

    And[
        a,
        b
    ]

You can also rely on the fact that logical expressions are always short-circuited, so False && Print["won't be printed"], doesn't print anything.

Finally, you can put a bunch of conditions in a list and do And @@ list.

klez• 13 years ago

> You can also rely on the fact that logical expressions are always short-circuited, so False && Print["won't be printed"], doesn't print anything.

This is true for any language I can think of.

E.g. in perl

    open(my $fh, "<", "input.txt")
        or die "cannot open < input.txt: $!";

is idiomatic

lnanek2• 13 years ago

Hmm, I'm not a fan of an if without braces. It is just begging for someone to come along and stick multiple statements after it expecting them to run when it fires. Heck, maybe even me tired at the end of the day.

DoubleCluster• 13 years ago

Great, now add an "else".

iajrz• 13 years ago

It's not complex at all. How can it be confusing? I was confused by the author's confusion :-/

ericbb• 13 years ago

The Gish source code does that too.

https://github.com/blinry/gish

stormbrew• 13 years ago

In ruby, the equivalent to a switch (confusingly called case) can take no initial-compare field and basically become a chained-if that lines up nicely (or more nicely than elsifs). I prefer the visual look of it, personally, but it seems that a lot of people find it too confusing.

As an example:

  case
  when (a == 1)
    dosomething
  when (b == 2)
    dosomethingelse
  when (c)
    orthis
  else
    awww
  end

jjoergensen• 13 years ago

Have some fun unit testing that! ha

stox• 13 years ago

I hope everyone realizes that this code was probably written by Ken Thompson or Rob Pike.

insertdisktwo• 13 years ago

My knee-jerk reaction to this post was that this increases the chance of another developer introducing a dangling else manyfold.

http://en.wikipedia.org/wiki/Dangling_else

moron4hire3• 13 years ago

I think the meaning was fairly obvious at first glance, and it read as a sloppy way of conjoining expressions, where && should be preferred, because it doesn't rely on the arbitrary block rules of if expressions.

Imagine for a moment that a stray ";" ends up at the end of one of those if statements. I don't care how, perhaps you just dropped your Warby Parkers on your keyboard or something. Not only is the code now broken, but it still compiles. With &&, it would fail to compile and the error would be caught immediately.

The moral of the story is that syntax is your friend, not your enemy. Use syntax as much as possible to catch errors. Especially with strictly-typed languages, you have an incredible tool for automated verification of certain portions of your program. Use it.

reddiric• 13 years ago

I think the meaning was fairly obvious at first glance, and it read as a fine way of conjoining expressions, where multiple if statements should be preferred, because it doesn't rely on the arbitrary subexpression evaluation ordering rules of binary operators.

Imagine for a moment that a character ends up deleted at the end of one of those && expressions. I don't care how, perhaps you just dropped your Warby Parkers on your keyboard or something. Not only is the code now broken, but it still compiles. With if(...)s, it would fail to compile and the error would be caught immediately.

moron4hire3• 13 years ago

Lazy evaluation is not arbitrary, it's a well established convention that any sane language implements (thus marking the languages that do not use it as completely insane and unworthy of our effort).

And also, your reply is technically incorrect, a typographical error or any kind is more likely to be caught in the && situation, and the style of reply makes you an insufferable geek.

overgard• 13 years ago

I actually quite like that, although usually when I have a large conditional like that I just give it named boolean variables to make it read more like a sentence, IE:

    bool exists = b != nil;
    bool sameType = b->qid.type == a->qid.type;
    bool samePath = b->qid.path == a->qid.path;
    ...
    if(exists && sameType && samePath) {
    }

It's also a lot easier to inspect in a debugger, as all the conditions appear in the locals list.

iso8859-1• 13 years ago

You could also group related conditions, so that nested if's don't bloat the locals so much:

    speedIsSlow = dict(downstreamSlow = downstreamSpeed < 200, 
                       upstreamSlow = upstreamSpeed < 100)
    
    if any(speedIsSlow.values()):
        connectionIsDead = getPingTime() > 100
        if connectionIsDead:
                print("It's dead, Jim")

kbruner• 13 years ago

This just looks like it was written by someone who's done some logic programming.

Ifs and &&s and Plan 9's Source Code

Related Stories