I don't like their indent guidelines, 2 spaces just feels incredibly "cramped" to me with Java. I understand it with an 80 or 100 col line limit (I also tend to 120, but 100's okay), but really, that's the only thing. I personally agree with most of this and we have warnings generated on commit (thanks Sonar!) if people don't follow some of these because they just lead to too many issues down the line that are really easy to accidentally skip over.
I wouldn't imagine it would be too hard to get an IDE auto-formatter to follow all of this (it's more or less what I use with mine, anyways).
If they used tabs instead of spaces, it wouldn't be an issue. People could format code on their machine whatever number of columns they prefer. But no... Spaces. Jesus. If you're looking at someone's HTML and it's got this kind of thing strewn about...
...you know with certainty that you're dealing with either 1) an amateur, or 2) someone constrained by top-down corporate numbnuttery.
The debate over spaces vs. tabs is as old as time. Associating one side of the argument with "top-down corporate numbnuttery" betrays a lack of familiarity with the issue.
The fact is that whatever shop you work at should have an opinion one way or the other, and you should follow it. Programmers will endlessly bikeshed the issue regardless of which style is chosen.
Happily, the Tab-letariat can agree with the Space-eoisie that a separation of the whitespace classes is necessary and good, and any commingling of the two is an affront to God.
Radical idea: lets move code away from the anachronism of flat files. There have been incredible evolution in some areas of development, and other areas time has completely frozen. We should have modern tools from the top-down.
Roughly it would be some sort of structured text, markup that represents the structure of the code rather than having a flat text file which then requires parsing to reconstruct structure. There are a couple of motivations here: structured representation allows editors to much more freely handle custom visual representations. Debates about tabs, spaces, braces, semicolons, parenthesis, etc etc are painfully outdated. These simply become properties of your chosen visual representation. It also allows more meaningful transforms without requiring a compilation pass, as a meaningful intermediate language is already present. This would have a few benefits: semantic text highlighting, straightforward semantic changeset merging, enhanced error highlighting and more semantic error detection (if I insert a line of code somewhere, the error is readily marked as the inserted line as the surrounding code is still semantically valid), sharing the semantic representation between IDE plugins. I'm sure this is just scratching the surface.
I'm reminded of the project I recall reading about from steve yeggie regarding a language agnostic IDE framework (whatever happened to that anyways?). A language agnostic, semantically rich file format would allow entire classes of transforms to apply across all languages that could be marked up by the language agnostic file format.
There are so many possibilities here, and yet a non-trivial amount of programmers are still anchored by the supposed benefits of editing code from the console! Programmers should be the first to be looking forward rather than being tied down to the past.
I thought you were going to suggest something crazy and imaginative issued from Big Data, like introduce some concept of Big Code, where you'd map-reduce your execution path, have non-consistent executions spread over thousands of servers, with the explicit target of singularity ;)
The option you suggest could have been done since long and it hasn't: I'm inclined to wonder why. Probably text-file editing has reached the good-enough level where we've got enough tooling to cope with the drawbacks, and the basics of open source is, it has to be the most urgent solution for the one who takes the initiative.
I used IBM's VisualAge for Java for a while back in the day, which was essentially applying a Smalltalk model of code organization to Java, but I never liked the image system. There's just something wonderful to me about being able to use a bunch of command line tools to perform ad hoc code analysis on files.
It's a radical idea that's been tried several times before and never stuck. See: Interlisp, Smalltalk, VisualAge for Java.
I suspect part of the problem is that there is a huge ecosystem of tools that work on plain text, and it's also easy to write more tools that work on plain text. You lose all of that when the database or parse tree is the primary representation, which means that when someone goes to make an adoption decision and their favorite tool doesn't work, they go back to the old way.
is one of those smells except for when it isn't.
<th>First Name</th>
The above is useful html, but I have had experienced developers "fix" it, thinking it was some copy-paste/dreamweaver accident. Non-Breaking SPaces are useful at times.
You should see how two-space indents look in a proportional font! It's a sorry sight.
That's the real shame of two-space indents. They basically require that all developers use monospaced fonts in their editors.
One of my most core principles for coding style is that the code should be equally readable in a proportional font or a monospaced font.
Why is that a core principle for me? Because I like proportional fonts. I don't use less-readable monospaced fonts when I write; I don't know why I should have to use them when I code.
You can't achieve this with two-space indents. When you view the code in a proportional font, it looks like one-space indentation.
For me, a proportional font is so much more readable, and lets me see so much more code at once, that I'm bummed that so many popular coding styles use conventions that only work monospaced.
To their credit, this coding standard actively discourages "column alignment". Abandoning column alignment is a wonderful thing: it gets you most of the way there toward having code that's perfectly readable in a proportional font.
It's unfortunate that they throw this benefit away by mandating two-space indents.
> Why is that a core principle for me? Because I like proportional fonts. I don't use less-readable monospaced fonts when I write; I don't know why I should have to use them when I code.
I had a phase like that about 15 years ago. In the end I just felt that it wasn't worth it. Whenever I looked at other peoples code through the lens of a proportional font, it would look awkward. Any attempt on the authors part of lining things up would just make things worse.
It turns out that proportional fonts are designed for text, not code.
That 15-year-ago mark is interesting to me, because that was just about the time when I switched to proportional fonts.
I recognize that I'm in a tiny minority: the vast majority of developers code in monospaced fonts.
That's one reason why I value an editor that makes it easy to switch between a proportional font and monospaced, so I can code in the font I like or switch to a monospaced font when necessary. One of the few that does this well is Komodo, where each "theme" includes both proportional and monospaced fonts, and you can easily switch between them. (I just wish IntelliJ IDEA did this!)
It's also why I wouldn't presume to use a coding standard that worked only in a proportional font but looked bad in monospaced.
But to my eyes, a coding standard that works equally well in proportional and monospaced fonts is superior to one that works only in monospaced.
I must respectfully disagree that "It turns out that proportional fonts are designed for text, not code." All of the code I've written in the last 10-15 years is just as readable in a proportional font or monospaced. It really makes no difference at all which kind of font you read the code in, if the coding standard was designed to support both.
Sorry for the slow reply. In case you read this later, my favorite coding font is actually Georgia.
It's not perfect; Georgia's lowercase o looks a lot like 0. But in practice that has never been a problem for me. And unlike many proportional fonts which have monospaced numbers, even the numbers are proportional in Georgia.
But I just like the way it looks and the readability it has - for my eyes.
I spend a lot of time reading code. I read other peoples code every day. To a very large extent, that code is going to use various layout tricks that are based on a monospaced font. Someone might line comments up like this:
// 1. Blah blah, blah
// blah and more blah.
// 2. Second piece of blah.
Arguments to a function might be lined up:
void func(int arg1,
int arg2...
Someone might describe some detail of a network protocol with a diagram (as often seen in RFCs). A large comment block might be adjusted to fill out to a certain column width. I'm probably forgetting a ton of other layout tricks that people use. Almost all of them rely on a monospace font.
Looking at code that wasn't specifically written for proportional fonts is death by a thousand cuts. The readability suffers greatly.
This reminds me of some people who try to do ASCII art in a proportional font in a plaintext email, which then looks completely munged on everyone who uses monospace fonts or a different proportional font. IMHO plain text files should always be read/written in a monospace font, as otherwise you limit the expression of the medium unnecessarily since ASCII art becomes no longer "portable".
I actually write very short functions and blocks of code with very little nesting. So that isn't the problem.
The problem for me is simply that a two-space indent looks like one space in a proportional font, and that makes it pretty hard to follow even one or two levels of indentation.
Indent size is definitely a personal preference; I started out with 8, then shortly went to 4, moved down to 3 for a very short time followed by 2, and have now settled on one space. One nice side-effect is that I can glance at the column counter and see the indent level immediately, and the closing braces look like a diagonal line which makes a missing/extra one quite easy to spot:
}
}
}
The nice thing about C-syntax languages is they can be automatically formatted, so anyone who finds my indents to be too small can increase them, and vice-versa,
It's just because it's stated inside the Android Code Style Guidelines for Contributors. I see a lot of other people, especially Android developers, use that field name conventions too. There is not only mFieldName, there are sFieldName for static variable as well. http://source.android.com/source/code-style.html#follow-fiel...
I really hope they work on the Code Style Guidelines for Contributors. But even the generated source code (like resources in R.java) should adhere to these guidelines.
My favorite rule is 4.1.1; always use braces even where optional. I think a small fraction of the economy may have been damaged by the bugs caused by the silliness of leaving out optional braces and then having another developer come along and, well, you know how that goes :)
As someone who once read the Sun Java Style Coding [1] (which at a quick glance, appears to be a super set of the Google style guidelines) standard years ago and found it exciting and very helpful material, I unfortunately did not get such a kick skimming through this. Perhaps it's because after having read Clean Code [2], I feel that there is much missing that can lend itself to better code.
It is amazing the old-style K&R still persists. I think there was one more reason it was preferred in the early days which is not mentioned in the aforesaid link: storage space which was also expensive. The K&R style saves a new-line char which would be a significant saving in large codebases.
Reading code with K&R braces is almost unbearable to me. I don't really know why, it just seems like a wasted line and visual clutter when reading code.
just looks too good to be wrong. I used Allman style for my first ~10 years of programming, then switched to "compact" style†, and didn't find it a big deal to adjust. Brace concerns melt away just like parens do in Lisp.
† I think K&R does make use of new line braces, for functions.
Interesting how this document focuses on formatting and naming, whereas their C++ style is full of stuff like "don't use exceptions" or "avoid doing too much work in the constructor." I guess Java guides you into a uniform coding style much more than C++.
C++ is a sprawling language that lends itself to many different coding styles. It's also a language with many pitfalls. The guide both encourages a common style so that things look uniform and a style that tries to avoid pitfalls.
That said, the C++ guide has a bunch of peculiar quirks.
A large part of the C++ style guide is basically "don't take advantage of any features Java doesn't have", which obviously doesn't apply to a Java style guide.
I've never understood all of the hate for using tabs instead of spaces while indenting. My thinking has always been that tabs just make it easier to adjust indentation.
As a tab-using heathen, can somebody please explain why spaces would be preferable?
If everyone has to use the same level of indentation (as with spaces), then everyone can adhere to the same width limit, and every source file in your codebase will fit in the same sized window. If some programmer sets their tab width to two spaces and then writes code that fits in 80 characters on their screen, it might not fit in some other programmer's screen.
From a philosophical perspective, tabs just aren't necessary. You need spaces in your source code in places besides indentation, but you don't need tabs. It's simpler to only have a single type of whitespace character.
For me, it's the fact that tabs don't behave like single characters in an editor, so instead of moving backwards a single character-width when you delete one, there's a sudden jump, which I find rather disorienting when I'm editing code (or text files) that uses them.
Each switch statement includes a default statement group, even if it contains no code.
I've never understood this one: adding `default` just prevents the compiler from telling you when you forgot to handle a particular case...in which case (heh), why use a switch statement at all? What are you buying over a series of if-else statements, other than possible code reuse due to the fall-through behavior?
There are two issues here. First, why use switch instead of a series of if-else, apart from compile-time completeness checks? Second, why include an empty default statement in a switch, apart from preventing compile-time completeness checks?
If a series of if-else statements can be rewritten as a switch statement, it will almost always be slightly more concise as a switch statement, and a more semantically correct representation of the logic.
Including an empty default statement come down to a matter of preference, but a pretty good argument for including it is that it favors the explicit over the implicit. It's better to be clear to other programmers that yes, you did intend to have nothing occur if the switched value didn't have a matching case. If you intend to handle all cases, you can put "assert false" or similar in the default case to be explicit about it. I can see that one going either way with enums, though, if you get a compile-time error when forgetting to handle a value (it's been too long since I've worked in Java to remember how it handles that).
I guess they were right when they said that the cobbler's kids wear no shoes. It stumps me how one of the biggest, most powerful, and most resourceful software companies in the world would have a bunch of humans write a document on how to format source code, then have all their programmers read that document and memorize it to the point where they can abide by it every second of their work day. Formatting code is the kind of thing that a computer can do. Write a program that formats the code the way you want it to be formatted, and invoke it in the appropriate place. Preferably, make it invokable in people's IDEs, so that they can go conformant whenever they wish.
> I guess they were right when they said that the cobbler's kids wear no shoes....Formatting code is the kind of thing that a computer can do.
Given that Go does this far better than any other language that I can think of at the moment[0], maybe it's a case of the cobbler's kids making their own shoes.
It's better than many other companies, where nobody bothers to to write or read that document, let alone use a formatter.
Anyway, this problem doesn't exist in Go (gofmt), C++ (clang-format), or at least 2 of the in-house languages. It still exists in Java, Python, and Javascript because of the lack of robust standalone parsers for those languages, but there are editor plugins for vim, emacs, Eclipse, and I think IntelliJ that automatically enforce the styleguide. (Google does not mandate a single editor the way it mandates languages.)
At least if their sponsored open source projects are any indication, they use plenty of sophisticated tools to check and reformat java source code to conform to style guidelines.
That doesn't obviate the need for a style definition in the first place, and I'm happy to see the canonical version of such a document embodied in human readable text rather than in source code or some kind of meta-language monstrosity (a la xsd).
56 comments
[ 1.3 ms ] story [ 130 ms ] threadEdit: Or an eclipse formatter config?
I wouldn't imagine it would be too hard to get an IDE auto-formatter to follow all of this (it's more or less what I use with mine, anyways).
The fact is that whatever shop you work at should have an opinion one way or the other, and you should follow it. Programmers will endlessly bikeshed the issue regardless of which style is chosen.
dons flame retardant suit
Otoh, if you're going to suggest binary structures, I wont be able to approve you if they're not easy to manage in Git.
I'm reminded of the project I recall reading about from steve yeggie regarding a language agnostic IDE framework (whatever happened to that anyways?). A language agnostic, semantically rich file format would allow entire classes of transforms to apply across all languages that could be marked up by the language agnostic file format.
There are so many possibilities here, and yet a non-trivial amount of programmers are still anchored by the supposed benefits of editing code from the console! Programmers should be the first to be looking forward rather than being tied down to the past.
The option you suggest could have been done since long and it hasn't: I'm inclined to wonder why. Probably text-file editing has reached the good-enough level where we've got enough tooling to cope with the drawbacks, and the basics of open source is, it has to be the most urgent solution for the one who takes the initiative.
I suspect part of the problem is that there is a huge ecosystem of tools that work on plain text, and it's also easy to write more tools that work on plain text. You lose all of that when the database or parse tree is the primary representation, which means that when someone goes to make an adoption decision and their favorite tool doesn't work, they go back to the old way.
That's the real shame of two-space indents. They basically require that all developers use monospaced fonts in their editors.
One of my most core principles for coding style is that the code should be equally readable in a proportional font or a monospaced font.
Why is that a core principle for me? Because I like proportional fonts. I don't use less-readable monospaced fonts when I write; I don't know why I should have to use them when I code.
You can't achieve this with two-space indents. When you view the code in a proportional font, it looks like one-space indentation.
For me, a proportional font is so much more readable, and lets me see so much more code at once, that I'm bummed that so many popular coding styles use conventions that only work monospaced.
To their credit, this coding standard actively discourages "column alignment". Abandoning column alignment is a wonderful thing: it gets you most of the way there toward having code that's perfectly readable in a proportional font.
It's unfortunate that they throw this benefit away by mandating two-space indents.
I had a phase like that about 15 years ago. In the end I just felt that it wasn't worth it. Whenever I looked at other peoples code through the lens of a proportional font, it would look awkward. Any attempt on the authors part of lining things up would just make things worse.
It turns out that proportional fonts are designed for text, not code.
I recognize that I'm in a tiny minority: the vast majority of developers code in monospaced fonts.
That's one reason why I value an editor that makes it easy to switch between a proportional font and monospaced, so I can code in the font I like or switch to a monospaced font when necessary. One of the few that does this well is Komodo, where each "theme" includes both proportional and monospaced fonts, and you can easily switch between them. (I just wish IntelliJ IDEA did this!)
It's also why I wouldn't presume to use a coding standard that worked only in a proportional font but looked bad in monospaced.
But to my eyes, a coding standard that works equally well in proportional and monospaced fonts is superior to one that works only in monospaced.
I must respectfully disagree that "It turns out that proportional fonts are designed for text, not code." All of the code I've written in the last 10-15 years is just as readable in a proportional font or monospaced. It really makes no difference at all which kind of font you read the code in, if the coding standard was designed to support both.
Fonts designed for programming typically take extra care of making sure that certain groups of glyphs look different: "O0" and "1lI" for example.
Now I'm curious: what proportional fonts do you use for coding?
It's not perfect; Georgia's lowercase o looks a lot like 0. But in practice that has never been a problem for me. And unlike many proportional fonts which have monospaced numbers, even the numbers are proportional in Georgia.
But I just like the way it looks and the readability it has - for my eyes.
I spend a lot of time reading code. I read other peoples code every day. To a very large extent, that code is going to use various layout tricks that are based on a monospaced font. Someone might line comments up like this:
Arguments to a function might be lined up: Someone might describe some detail of a network protocol with a diagram (as often seen in RFCs). A large comment block might be adjusted to fill out to a certain column width. I'm probably forgetting a ton of other layout tricks that people use. Almost all of them rely on a monospace font.Looking at code that wasn't specifically written for proportional fonts is death by a thousand cuts. The readability suffers greatly.
The problem for me is simply that a two-space indent looks like one space in a proportional font, and that makes it pretty hard to follow even one or two levels of indentation.
(And apologies for not replying sooner!)
As someone who once read the Sun Java Style Coding [1] (which at a quick glance, appears to be a super set of the Google style guidelines) standard years ago and found it exciting and very helpful material, I unfortunately did not get such a kick skimming through this. Perhaps it's because after having read Clean Code [2], I feel that there is much missing that can lend itself to better code.
[1] http://www.oracle.com/technetwork/java/javase/documentation/...
[2] http://www.amazon.com/Clean-Code-Handbook-Software-Craftsman...
http://www.experimentgarden.com/2009/07/facts-behind-code-in...
It is amazing the old-style K&R still persists. I think there was one more reason it was preferred in the early days which is not mentioned in the aforesaid link: storage space which was also expensive. The K&R style saves a new-line char which would be a significant saving in large codebases.
† I think K&R does make use of new line braces, for functions.
That said, the C++ guide has a bunch of peculiar quirks.
As a tab-using heathen, can somebody please explain why spaces would be preferable?
From a philosophical perspective, tabs just aren't necessary. You need spaces in your source code in places besides indentation, but you don't need tabs. It's simpler to only have a single type of whitespace character.
I've never understood this one: adding `default` just prevents the compiler from telling you when you forgot to handle a particular case...in which case (heh), why use a switch statement at all? What are you buying over a series of if-else statements, other than possible code reuse due to the fall-through behavior?
Code clarity. Depending on how and what you're doing in each case a switch statement can be easier to read than a whole series of if .. else blocks.
And sometimes, given that they are equivalent, it's just a style choice on the part of the programmer.
For enums, the default case should be left out, since 'default' actually masks the missing case, as you note.
If a series of if-else statements can be rewritten as a switch statement, it will almost always be slightly more concise as a switch statement, and a more semantically correct representation of the logic.
Including an empty default statement come down to a matter of preference, but a pretty good argument for including it is that it favors the explicit over the implicit. It's better to be clear to other programmers that yes, you did intend to have nothing occur if the switched value didn't have a matching case. If you intend to handle all cases, you can put "assert false" or similar in the default case to be explicit about it. I can see that one going either way with enums, though, if you get a compile-time error when forgetting to handle a value (it's been too long since I've worked in Java to remember how it handles that).
God says... stunning unemployment spirit hey_Mikey_he_likes_it yep biggot I'm_on_a_roll what_do_you_expect rose_colored_glasses King_Midas and_then_what chump_change stunning close_your_eyes you_don't_say geek baffling illogical snap_out_of_it do_you_get_a_cookie guilty far_out_man you're_out_of_your_mind love pow by_the_way dude_such_a_scoffer incoming pride thank_you_very_much bummer peace ice_cream I'm_grieved atrocious Terry resume oh_oh sess_me I_pitty_the_fool little_fish you_don't_say Vegas delicious potentially voodoo what_have_you_done_for_me_lately really dance Hasta uh_huh employer joker
----------
http://businessofsoftware.org/2008/02/the-perfect-cir/
Given that Go does this far better than any other language that I can think of at the moment[0], maybe it's a case of the cobbler's kids making their own shoes.
[0] http://golang.org/cmd/gofmt/
Anyway, this problem doesn't exist in Go (gofmt), C++ (clang-format), or at least 2 of the in-house languages. It still exists in Java, Python, and Javascript because of the lack of robust standalone parsers for those languages, but there are editor plugins for vim, emacs, Eclipse, and I think IntelliJ that automatically enforce the styleguide. (Google does not mandate a single editor the way it mandates languages.)
That doesn't obviate the need for a style definition in the first place, and I'm happy to see the canonical version of such a document embodied in human readable text rather than in source code or some kind of meta-language monstrosity (a la xsd).
1) Don't