No major systems language has emerged in over a decade, but over that time the computing landscape has changed tremendously. There are several trends:
We believe it's worth trying again with a new language, a concurrent, garbage-collected language with fast compilation. Regarding the points above:
A much more expansive answer to this question is available in the article, Go at Google: Language Design in the Service of Software Engineering.
Go became a public open source project on November 10, 2009. After a couple of years of very active design and development, stability was called for and Go 1 was released on March 28, 2012. Go 1, which includes a language specification, standard libraries, and custom tools, provides a stable foundation for creating reliable products, projects, and publications.
With that stability established, we are using Go to develop programs, products, and tools rather than actively changing the language and libraries. In fact, the purpose of Go 1 is to provide long-term stability. Backwards-incompatible changes will not be made to any Go 1 point release. We want to use what we have to learn how a future version of Go might look, rather than to play with the language underfoot.
Of course, development will continue on Go itself, but the focus will be on performance, reliability, portability and the addition of new functionality such as improved support for internationalization.
There may well be a Go 2 one day, but not for a few years and it will be influenced by what we learn using Go 1 as it is today.
The mascot and logo were designed by Renée French, who also designed Glenda, the Plan 9 bunny. The gopher is derived from one she used for an WFMU T-shirt design some years ago. The logo and mascot are covered by the Creative Commons Attribution 3.0 license.
Robert Griesemer, Rob Pike and Ken Thompson started sketching the goals for a new language on the white board on September 21, 2007. Within a few days the goals had settled into a plan to do something and a fair idea of what it would be. Design continued part-time in parallel with unrelated work. By January 2008, Ken had started work on a compiler with which to explore ideas; it generated C code as its output. By mid-year the language had become a full-time project and had settled enough to attempt a production compiler. In May 2008, Ian Taylor independently started on a GCC front end for Go using the draft specification. Russ Cox joined in late 2008 and helped move the language and libraries from prototype to reality.
Go became a public open source project on November 10, 2009. Many people from the community have contributed ideas, discussions, and code.
Go was born out of frustration with existing languages and environments for systems programming. Programming had become too difficult and the choice of languages was partly to blame. One had to choose either efficient compilation, efficient execution, or ease of programming; all three were not available in the same mainstream language. Programmers who could were choosing ease over safety and efficiency by moving to dynamically typed languages such as Python and JavaScript rather than C++ or, to a lesser extent, Java.
Go is an attempt to combine the ease of programming of an interpreted, dynamically typed language with the efficiency and safety of a statically typed, compiled language. It also aims to be modern, with support for networked and multicore computing. Finally, working with Go is intended to be fast: it should take at most a few seconds to build a large executable on a single computer. To meet these goals required addressing a number of linguistic issues: an expressive but lightweight type system; concurrency and garbage collection; rigid dependency specification; and so on. These cannot be addressed well by libraries or tools; a new language was called for.
The article Go at Google discusses the background and motivation behind the design of the Go language, as well as providing more detail about many of the answers presented in this FAQ.
Go is mostly in the C family (basic syntax), with significant input from the Pascal/Modula/Oberon family (declarations, packages), plus some ideas from languages inspired by Tony Hoare's CSP, such as Newsqueak and Limbo (concurrency). However, it is a new language across the board. In every respect the language was designed by thinking about what programmers do and how to make programming, at least the kind of programming we do, more effective, which means more fun.
Programming today involves too much bookkeeping, repetition, and clerical work. As Dick Gabriel says, “Old programs read like quiet conversations between a well-spoken research worker and a well-studied mechanical colleague, not as a debate with a compiler. Who'd have guessed sophistication bought such noise?” The sophistication is worthwhile—no one wants to go back to the old languages—but can it be more quietly achieved?
Go attempts to reduce the amount of typing in both senses of the word.
Throughout its design, we have tried to reduce clutter and
complexity. There are no forward declarations and no header files;
everything is declared exactly once. Initialization is expressive,
automatic, and easy to use. Syntax is clean and light on keywords.
Stuttering (foo.Foo* myFoo = new(foo.Foo)) is reduced by
simple type derivation using the :=
declare-and-initialize construct. And perhaps most radically, there
is no type hierarchy: types just are, they don't have to
announce their relationships. These simplifications allow Go to be
expressive yet comprehensible without sacrificing, well, sophistication.
Another important principle is to keep the concepts orthogonal. Methods can be implemented for any type; structures represent data while interfaces represent abstraction; and so on. Orthogonality makes it easier to understand what happens when things combine.
Yes. There are now several Go programs deployed in
production inside Google. A public example is the server behind
golang.org.
It's just the godoc
document server running in a production configuration on
Google App Engine.
Other examples include the Vitess
system for large-scale SQL installations and Google's download server, dl.google.com,
which delivers Chrome binaries and other large installables such as apt-get
packages.
There are two Go compiler implementations, gc
and gccgo.
Gc uses a different calling convention and linker and can
therefore only be linked with C programs using the same convention.
There is such a C compiler but no C++ compiler.
Gccgo is a GCC front-end that can, with care, be linked with
GCC-compiled C or C++ programs.
The cgo program provides the mechanism for a “foreign function interface” to allow safe calling of C libraries from Go code. SWIG extends this capability to C++ libraries.
A separate open source project provides the necessary compiler plugin and library. It is available at github.com/golang/protobuf/
Absolutely. We encourage developers to make Go Language sites in their own languages. However, if you choose to add the Google logo or branding to your site (it does not appear on golang.org), you will need to abide by the guidelines at www.google.com/permissions/guidelines.html
Go does have an extensive library, called the runtime,
that is part of every Go program.
The runtime library implements garbage collection, concurrency,
stack management, and other critical features of the Go language.
Although it is more central to the language, Go's runtime is analogous
to libc, the C library.
It is important to understand, however, that Go's runtime does not include a virtual machine, such as is provided by the Java runtime. Go programs are compiled ahead of time to native machine code. Thus, although the term is often used to describe the virtual environment in which a program runs, in Go the word “runtime” is just the name given to the library providing critical language services.
It was important to us to extend the space of identifiers from the confines of ASCII. Go's rule—identifier characters must be letters or digits as defined by Unicode—is simple to understand and to implement but has restrictions. Combining characters are excluded by design, for instance. Until there is an agreed external definition of what an identifier might be, plus a definition of canonicalization of identifiers that guarantees no ambiguity, it seemed better to keep combining characters out of the mix. Thus we have a simple rule that can be expanded later without breaking programs, one that avoids bugs that would surely arise from a rule that admits ambiguous identifiers.
On a related note, since an exported identifier must begin with an
upper-case letter, identifiers created from “letters”
in some languages can, by definition, not be exported. For now the
only solution is to use something like X日本語, which
is clearly unsatisfactory; we are considering other options. The
case-for-visibility rule is unlikely to change however; it's one
of our favorite features of Go.
Every language contains novel features and omits someone's favorite feature. Go was designed with an eye on felicity of programming, speed of compilation, orthogonality of concepts, and the need to support features such as concurrency and garbage collection. Your favorite feature may be missing because it doesn't fit, because it affects compilation speed or clarity of design, or because it would make the fundamental system model too difficult.
If it bothers you that Go is missing feature X, please forgive us and investigate the features that Go does have. You might find that they compensate in interesting ways for the lack of X.
Generics may well be added at some point. We don't feel an urgency for them, although we understand some programmers do.
Generics are convenient but they come at a cost in complexity in the type system and run-time. We haven't yet found a design that gives value proportionate to the complexity, although we continue to think about it. Meanwhile, Go's built-in maps and slices, plus the ability to use the empty interface to construct containers (with explicit unboxing) mean in many cases it is possible to write code that does what generics would enable, if less smoothly.
The topic remains open. For a look at several previous unsuccessful attempts to design a good generics solution for Go, see this proposal.
We believe that coupling exceptions to a control
structure, as in the try-catch-finally idiom, results in
convoluted code. It also tends to encourage programmers to label
too many ordinary errors, such as failing to open a file, as
exceptional.
Go takes a different approach. For plain error handling, Go's multi-value returns make it easy to report an error without overloading the return value. A canonical error type, coupled with Go's other features, makes error handling pleasant but quite different from that in other languages.
Go also has a couple of built-in functions to signal and recover from truly exceptional conditions. The recovery mechanism is executed only as part of a function's state being torn down after an error, which is sufficient to handle catastrophe but requires no extra control structures and, when used well, can result in clean error-handling code.
See the Defer, Panic, and Recover article for details.
Go doesn't provide assertions. They are undeniably convenient, but our experience has been that programmers use them as a crutch to avoid thinking about proper error handling and reporting. Proper error handling means that servers continue operation after non-fatal errors instead of crashing. Proper error reporting means that errors are direct and to the point, saving the programmer from interpreting a large crash trace. Precise errors are particularly important when the programmer seeing the errors is not familiar with the code.
We understand that this is a point of contention. There are many things in the Go language and libraries that differ from modern practices, simply because we feel it's sometimes worth trying a different approach.
Concurrency and multi-threaded programming have a reputation for difficulty. We believe this is due partly to complex designs such as pthreads and partly to overemphasis on low-level details such as mutexes, condition variables, and memory barriers. Higher-level interfaces enable much simpler code, even if there are still mutexes and such under the covers.
One of the most successful models for providing high-level linguistic support for concurrency comes from Hoare's Communicating Sequential Processes, or CSP. Occam and Erlang are two well known languages that stem from CSP. Go's concurrency primitives derive from a different part of the family tree whose main contribution is the powerful notion of channels as first class objects. Experience with several earlier languages has shown that the CSP model fits well into a procedural language framework.
Goroutines are part of making concurrency easy to use. The idea, which has been around for a while, is to multiplex independently executing functions—coroutines—onto a set of threads. When a coroutine blocks, such as by calling a blocking system call, the run-time automatically moves other coroutines on the same operating system thread to a different, runnable thread so they won't be blocked. The programmer sees none of this, which is the point. The result, which we call goroutines, can be very cheap: they have little overhead beyond the memory for the stack, which is just a few kilobytes.
To make the stacks small, Go's run-time uses resizable, bounded stacks. A newly minted goroutine is given a few kilobytes, which is almost always enough. When it isn't, the run-time grows (and shrinks) the memory for storing the stack automatically, allowing many goroutines to live in a modest amount of memory. The CPU overhead averages about three cheap instructions per function call. It is practical to create hundreds of thousands of goroutines in the same address space. If goroutines were just threads, system resources would run out at a much smaller number.
After long discussion it was decided that the typical use of maps did not require safe access from multiple goroutines, and in those cases where it did, the map was probably part of some larger data structure or computation that was already synchronized. Therefore requiring that all map operations grab a mutex would slow down most programs and add safety to few. This was not an easy decision, however, since it means uncontrolled map access can crash the program.
The language does not preclude atomic map updates. When required, such as when hosting an untrusted program, the implementation could interlock map access.
People often suggest improvements to the language—the mailing list contains a rich history of such discussions—but very few of these changes have been accepted.
Although Go is an open source project, the language and libraries are protected by a compatibility promise that prevents changes that break existing programs. If your proposal violates the Go 1 specification we cannot even entertain the idea, regardless of its merit. A future major release of Go may be incompatible with Go 1, but we're not ready to start talking about what that might be.
Even if your proposal is compatible with the Go 1 spec, it might not be in the spirit of Go's design goals. The article Go at Google: Language Design in the Service of Software Engineering explains Go's origins and the motivation behind its design.
Yes and no. Although Go has types and methods and allows an object-oriented style of programming, there is no type hierarchy. The concept of “interface” in Go provides a different approach that we believe is easy to use and in some ways more general. There are also ways to embed types in other types to provide something analogous—but not identical—to subclassing. Moreover, methods in Go are more general than in C++ or Java: they can be defined for any sort of data, even built-in types such as plain, “unboxed” integers. They are not restricted to structs (classes).
Also, the lack of a type hierarchy makes “objects” in Go feel much more lightweight than in languages such as C++ or Java.
The only way to have dynamically dispatched methods is through an interface. Methods on a struct or any other concrete type are always resolved statically.
Object-oriented programming, at least in the best-known languages, involves too much discussion of the relationships between types, relationships that often could be derived automatically. Go takes a different approach.
Rather than requiring the programmer to declare ahead of time that two types are related, in Go a type automatically satisfies any interface that specifies a subset of its methods. Besides reducing the bookkeeping, this approach has real advantages. Types can satisfy many interfaces at once, without the complexities of traditional multiple inheritance. Interfaces can be very lightweight—an interface with one or even zero methods can express a useful concept. Interfaces can be added after the fact if a new idea comes along or for testing—without annotating the original types. Because there are no explicit relationships between types and interfaces, there is no type hierarchy to manage or discuss.
It's possible to use these ideas to construct something analogous to
type-safe Unix pipes. For instance, see how fmt.Fprintf
enables formatted printing to any output, not just a file, or how the
bufio package can be completely separate from file I/O,
or how the image packages generate compressed
image files. All these ideas stem from a single interface
(io.Writer) representing a single method
(Write). And that's only scratching the surface.
Go's interfaces have a profound influence on how programs are structured.
It takes some getting used to but this implicit style of type dependency is one of the most productive things about Go.
len a function and not a method?
We debated this issue but decided
implementing len and friends as functions was fine in practice and
didn't complicate questions about the interface (in the Go type sense)
of basic types.
Method dispatch is simplified if it doesn't need to do type matching as well. Experience with other languages told us that having a variety of methods with the same name but different signatures was occasionally useful but that it could also be confusing and fragile in practice. Matching only by name and requiring consistency in the types was a major simplifying decision in Go's type system.
Regarding operator overloading, it seems more a convenience than an absolute requirement. Again, things are simpler without it.
A Go type satisfies an interface by implementing the methods of that interface, nothing more. This property allows interfaces to be defined and used without having to modify existing code. It enables a kind of structural typing that promotes separation of concerns and improves code re-use, and makes it easier to build on patterns that emerge as the code develops. The semantics of interfaces is one of the main reasons for Go's nimble, lightweight feel.
See the question on type inheritance for more detail.
You can ask the compiler to check that the type T implements the
interface I by attempting an assignment using the zero value for
T or pointer to T, as appropriate:
type T struct{}
var _ I = T{} // Verify that T implements I.
var _ I = (*T)(nil) // Verify that *T implements I.
If T (or *T, accordingly) doesn't implement
I, the mistake will be caught at compile time.
If you wish the users of an interface to explicitly declare that they implement it, you can add a method with a descriptive name to the interface's method set. For example:
type Fooer interface {
Foo()
ImplementsFooer()
}
A type must then implement the ImplementsFooer method to be a
Fooer, clearly documenting the fact and announcing it in
godoc's output.
type Bar struct{}
func (b Bar) ImplementsFooer() {}
func (b Bar) Foo() {}
Most code doesn't make use of such constraints, since they limit the utility of the interface idea. Sometimes, though, they're necessary to resolve ambiguities among similar interfaces.
Consider this simple interface to represent an object that can compare itself with another value:
type Equaler interface {
Equal(Equaler) bool
}
and this type, T:
type T int
func (t T) Equal(u T) bool { return t == u } // does not satisfy Equaler
Unlike the analogous situation in some polymorphic type systems,
T does not implement Equaler.
The argument type of T.Equal is T,
not literally the required type Equaler.
In Go, the type system does not promote the argument of
Equal; that is the programmer's responsibility, as
illustrated by the type T2, which does implement
Equaler:
type T2 int
func (t T2) Equal(u Equaler) bool { return t == u.(T2) } // satisfies Equaler
Even this isn't like other type systems, though, because in Go any
type that satisfies Equaler could be passed as the
argument to T2.Equal, and at run time we must
check that the argument is of type T2.
Some languages arrange to make that guarantee at compile time.
A related example goes the other way:
type Opener interface {
Open() Reader
}
func (t T3) Open() *os.File
In Go, T3 does not satisfy Opener,
although it might in another language.
While it is true that Go's type system does less for the programmer in such cases, the lack of subtyping makes the rules about interface satisfaction very easy to state: are the function's names and signatures exactly those of the interface? Go's rule is also easy to implement efficiently. We feel these benefits offset the lack of automatic type promotion. Should Go one day adopt some form of polymorphic typing, we expect there would be a way to express the idea of these examples and also have them be statically checked.
Not directly, because they do not have the same representation in memory.
It is necessary to copy the elements individually to the destination
slice. This example converts a slice of int to a slice of
interface{}:
t := []int{1, 2, 3, 4}
s := make([]interface{}, len(t))
for i, v := range t {
s[i] = v
}
type T1 int type T2 int var t1 T1 var x = T2(t1) // OK var st1 []T1 var sx = ([]T2)(st1) // NOT OK
In Go, types are closely tied to methods, in that every named type has a (possibly empty) method set. The general rule is that you can change the name of the type being converted (and thus possibly change its method set) but you can't change the name (and method set) of elements of a composite type. Go requires you to be explicit about type conversions.
Under the covers, interfaces are implemented as two elements, a type and a value.
The value, called the interface's dynamic value,
is an arbitrary concrete value and the type is that of the value.
For the int value 3, an interface value contains,
schematically, (int, 3).
An interface value is nil only if the inner value and type are both unset,
(nil, nil).
In particular, a nil interface will always hold a nil type.
If we store a nil pointer of type *int inside
an interface value, the inner type will be *int regardless of the value of the pointer:
(*int, nil).
Such an interface value will therefore be non-nil
even when the pointer inside is nil.
This situation can be confusing, and arises when a nil value is
stored inside an interface value such as an error return:
func returnsError() error {
var p *MyError = nil
if bad() {
p = ErrBad
}
return p // Will always return a non-nil error.
}
If all goes well, the function returns a nil p,
so the return value is an error interface
value holding (*MyError, nil).
This means that if the caller compares the returned error to nil,
it will always look as if there was an error even if nothing bad happened.
To return a proper nil error to the caller,
the function must return an explicit nil:
func returnsError() error {
if bad() {
return ErrBad
}
return nil
}
It's a good idea for functions
that return errors always to use the error type in
their signature (as we did above) rather than a concrete type such
as *MyError, to help guarantee the error is
created correctly. As an example,
os.Open
returns an error even though, if not nil,
it's always of concrete type
*os.PathError.
Similar situations to those described here can arise whenever interfaces are used.
Just keep in mind that if any concrete value
has been stored in the interface, the interface will not be nil.
For more information, see
The Laws of Reflection.
Untagged unions would violate Go's memory safety guarantees.
Variant types, also known as algebraic types, provide a way to specify that a value might take one of a set of other types, but only those types. A common example in systems programming would specify that an error is, say, a network error, a security error or an application error and allow the caller to discriminate the source of the problem by examining the type of the error. Another example is a syntax tree in which each node can be a different type: declaration, statement, assignment and so on.
We considered adding variant types to Go, but after discussion decided to leave them out because they overlap in confusing ways with interfaces. What would happen if the elements of a variant type were themselves interfaces?
Also, some of what variant types address is already covered by the language. The error example is easy to express using an interface value to hold the error and a type switch to discriminate cases. The syntax tree example is also doable, although not as elegantly.
Covariant result types would mean that an interface like
type Copyable interface {
Copy() interface{}
}
would be satisfied by the method
func (v Value) Copy() Value
because Value implements the empty interface.
In Go method types must match exactly, so Value does not
implement Copyable.
Go separates the notion of what a
type does—its methods—from the type's implementation.
If two methods return different types, they are not doing the same thing.
Programmers who want covariant result types are often trying to
express a type hierarchy through interfaces.
In Go it's more natural to have a clean separation between interface
and implementation.
The convenience of automatic conversion between numeric types in C is outweighed by the confusion it causes. When is an expression unsigned? How big is the value? Does it overflow? Is the result portable, independent of the machine on which it executes? It also complicates the compiler; “the usual arithmetic conversions” are not easy to implement and inconsistent across architectures. For reasons of portability, we decided to make things clear and straightforward at the cost of some explicit conversions in the code. The definition of constants in Go—arbitrary precision values free of signedness and size annotations—ameliorates matters considerably, though.
A related detail is that, unlike in C, int and int64
are distinct types even if int is a 64-bit type. The int
type is generic; if you care about how many bits an integer holds, Go
encourages you to be explicit.
A blog post titled Constants explores this topic in more detail.
The same reason strings are: they are such a powerful and important data structure that providing one excellent implementation with syntactic support makes programming more pleasant. We believe that Go's implementation of maps is strong enough that it will serve for the vast majority of uses. If a specific application can benefit from a custom implementation, it's possible to write one but it will not be as convenient syntactically; this seems a reasonable tradeoff.
Map lookup requires an equality operator, which slices do not implement. They don't implement equality because equality is not well defined on such types; there are multiple considerations involving shallow vs. deep comparison, pointer vs. value comparison, how to deal with recursive types, and so on. We may revisit this issue—and implementing equality for slices will not invalidate any existing programs—but without a clear idea of what equality of slices should mean, it was simpler to leave it out for now.
In Go 1, unlike prior releases, equality is defined for structs and arrays, so such types can be used as map keys. Slices still do not have a definition of equality, though.
There's a lot of history on that topic. Early on, maps and channels were syntactically pointers and it was impossible to declare or use a non-pointer instance. Also, we struggled with how arrays should work. Eventually we decided that the strict separation of pointers and values made the language harder to use. Changing these types to act as references to the associated, shared data structures resolved these issues. This change added some regrettable complexity to the language but had a large effect on usability: Go became a more productive, comfortable language when it was introduced.
There is a program, godoc, written in Go, that extracts
package documentation from the source code. It can be used on the
command line or on the web. An instance is running at
golang.org/pkg/.
In fact, godoc implements the full site at
golang.org/.
A godoc instance may be configured to provide rich,
interactive static analyses of symbols in the programs it displays; details are
listed here.
For access to documentation from the command line, the go tool has a doc subcommand that provides a textual interface to the same information.
Eventually, there may be a small number of rules to guide things
like naming, layout, and file organization.
The document Effective Go
contains some style advice.
More directly, the program gofmt is a pretty-printer
whose purpose is to enforce layout rules; it replaces the usual
compendium of do's and don'ts that allows interpretation.
All the Go code in the repository has been run through gofmt.
The document titled Go Code Review Comments is a collection of very short essays about details of Go idiom that are often missed by programmers. It is a handy reference for people doing code reviews for Go projects.
The library sources are in the src directory of the repository.
If you want to make a significant change, please discuss on the mailing list before embarking.
See the document Contributing to the Go project for more information about how to proceed.
Companies often permit outgoing traffic only on the standard TCP ports 80 (HTTP)
and 443 (HTTPS), blocking outgoing traffic on other ports, including TCP port 9418
(git) and TCP port 22 (SSH).
When using HTTPS instead of HTTP, git enforces certificate validation by
default, providing protection against man-in-the-middle, eavesdropping and tampering attacks.
The go get command therefore uses HTTPS for safety.
If you use git and prefer to push changes through SSH using your existing key
it's easy to work around this. For GitHub, try one of these solutions:
$ cd src/github.com/username $ git clone git@github.com:username/package.git
git push to use the SSH protocol by appending
these two lines to ~/.gitconfig:
[url "git@github.com:"] pushInsteadOf = https://github.com/
"Go get" does not have any explicit concept of package versions.
Versioning is a source of significant complexity, especially in large code bases,
and we are unaware of any approach that works well at scale in a large enough
variety of situations to be appropriate to force on all Go users.
What "go get" and the larger Go toolchain do provide is isolation of
packages with different import paths.
For example, the standard library's html/template and text/template
coexist even though both are "package template".
This observation leads to some advice for package authors and package users.
Packages intended for public use should try to maintain backwards compatibility as they evolve. The Go 1 compatibility guidelines are a good reference here: don't remove exported names, encourage tagged composite literals, and so on. If different functionality is required, add a new name instead of changing an old one. If a complete break is required, create a new package with a new import path.
If you're using an externally supplied package and worry that it might change in unexpected ways, the simplest solution is to copy it to your local repository. (This is the approach Google takes internally.) Store the copy under a new import path that identifies it as a local copy. For example, you might copy "original.com/pkg" to "you.com/external/original.com/pkg". The gomvpkg program is one tool to help automate this process.
The Go 1.5 release includes an experimental facility to the go command that makes it easier to manage external dependencies by "vendoring" them into a special directory near the package that depends upon them. See the design document for details.
As in all languages in the C family, everything in Go is passed by value.
That is, a function always gets a copy of the
thing being passed, as if there were an assignment statement assigning the
value to the parameter. For instance, passing an int value
to a function makes a copy of the int, and passing a pointer
value makes a copy of the pointer, but not the data it points to.
(See a later
section for a discussion of how this affects method receivers.)
Map and slice values behave like pointers: they are descriptors that contain pointers to the underlying map or slice data. Copying a map or slice value doesn't copy the data it points to. Copying an interface value makes a copy of the thing stored in the interface value. If the interface value holds a struct, copying the interface value makes a copy of the struct. If the interface value holds a pointer, copying the interface value makes a copy of the pointer, but again not the data it points to.
Note that this discussion is about the semantics of the operations. Actual implementations may apply optimizations to avoid copying as long as the optimizations do not change the semantics.
Almost never. Pointers to interface values arise only in rare, tricky situations involving disguising an interface value's type for delayed evaluation.
It is however a common mistake to pass a pointer to an interface value to a function expecting an interface. The compiler will complain about this error but the situation can still be confusing, because sometimes a pointer is necessary to satisfy an interface. The insight is that although a pointer to a concrete type can satisfy an interface, with one exception a pointer to an interface can never satisfy an interface.
Consider the variable declaration,
var w io.Writer
The printing function fmt.Fprintf takes as its first argument
a value that satisfies io.Writer—something that implements
the canonical Write method. Thus we can write
fmt.Fprintf(w, "hello, world\n")
If however we pass the address of w, the program will not compile.
fmt.Fprintf(&w, "hello, world\n") // Compile-time error.
The one exception is that any value, even a pointer to an interface, can be assigned to
a variable of empty interface type (interface{}).
Even so, it's almost certainly a mistake if the value is a pointer to an interface;
the result can be confusing.
func (s *MyStruct) pointerMethod() { } // method on pointer
func (s MyStruct) valueMethod() { } // method on value
For programmers unaccustomed to pointers, the distinction between these
two examples can be confusing, but the situation is actually very simple.
When defining a method on a type, the receiver (s in the above
examples) behaves exactly as if it were an argument to the method.
Whether to define the receiver as a value or as a pointer is the same
question, then, as whether a function argument should be a value or
a pointer.
There are several considerations.
First, and most important, does the method need to modify the
receiver?
If it does, the receiver must be a pointer.
(Slices and maps act as references, so their story is a little
more subtle, but for instance to change the length of a slice
in a method the receiver must still be a pointer.)
In the examples above, if pointerMethod modifies
the fields of s,
the caller will see those changes, but valueMethod
is called with a copy of the caller's argument (that's the definition
of passing a value), so changes it makes will be invisible to the caller.
By the way, pointer receivers are identical to the situation in Java, although in Java the pointers are hidden under the covers; it's Go's value receivers that are unusual.
Second is the consideration of efficiency. If the receiver is large,
a big struct for instance, it will be much cheaper to
use a pointer receiver.
Next is consistency. If some of the methods of the type must have pointer receivers, the rest should too, so the method set is consistent regardless of how the type is used. See the section on method sets for details.
For types such as basic types, slices, and small structs,
a value receiver is very cheap so unless the semantics of the method
requires a pointer, a value receiver is efficient and clear.
In short: new allocates memory, make initializes
the slice, map, and channel types.
See the relevant section of Effective Go for more details.
int on a 64 bit machine?
The sizes of int and uint are implementation-specific
but the same as each other on a given platform.
For portability, code that relies on a particular
size of value should use an explicitly sized type, like int64.
Prior to Go 1.1, the 64-bit Go compilers (both gc and gccgo) used
a 32-bit representation for int. As of Go 1.1 they use
a 64-bit representation.
On the other hand, floating-point scalars and complex
types are always sized (there are no float or complex basic types),
because programmers should be aware of precision when using floating-point numbers.
The default type used for an (untyped) floating-point constant is float64.
Thus foo := 3.0 declares a variable foo
of type float64.
For a float32 variable initialized by an (untyped) constant, the variable type
must be specified explicitly in the variable declaration:
var foo float32 = 3.0
Alternatively, the constant must be given a type with a conversion as in
foo := float32(3.0).
From a correctness standpoint, you don't need to know. Each variable in Go exists as long as there are references to it. The storage location chosen by the implementation is irrelevant to the semantics of the language.
The storage location does have an effect on writing efficient programs. When possible, the Go compilers will allocate variables that are local to a function in that function's stack frame. However, if the compiler cannot prove that the variable is not referenced after the function returns, then the compiler must allocate the variable on the garbage-collected heap to avoid dangling pointer errors. Also, if a local variable is very large, it might make more sense to store it on the heap rather than the stack.
In the current compilers, if a variable has its address taken, that variable is a candidate for allocation on the heap. However, a basic escape analysis recognizes some cases when such variables will not live past the return from the function and can reside on the stack.
The Go memory allocator reserves a large region of virtual memory as an arena for allocations. This virtual memory is local to the specific Go process; the reservation does not deprive other processes of memory.
To find the amount of actual memory allocated to a Go process, use the Unix
top command and consult the RES (Linux) or
RSIZE (Mac OS X) columns.
We haven't fully defined it all yet, but some details about atomicity are available in the Go Memory Model specification.
Regarding mutexes, the sync package implements them, but we hope Go programming style will encourage people to try higher-level techniques. In particular, consider structuring your program so that only one goroutine at a time is ever responsible for a particular piece of data.
Do not communicate by sharing memory. Instead, share memory by communicating.
See the Share Memory By Communicating code walk and its associated article for a detailed discussion of this concept.
The number of CPUs available simultaneously to executing goroutines is
controlled by the GOMAXPROCS shell environment variable.
In earlier releases of Go, the default value was 1, but as of Go 1.5 the default
value is the number of cores available.
Therefore programs compiled after 1.5 should demonstrate parallel execution
of multiple goroutines.
To change the behavior, set the environment variable or use the similarly-named
function
of the runtime package to configure the
run-time support to utilize a different number of threads.
Programs that perform parallel computation might benefit from a further increase in
GOMAXPROCS.
However, be aware that
concurrency
is not parallelism.
GOMAXPROCS > 1 sometimes make my program
slower?It depends on the nature of your program. Problems that are intrinsically sequential cannot be sped up by adding more goroutines. Concurrency only becomes parallelism when the problem is intrinsically parallel.
In practical terms, programs that spend more time
communicating on channels than doing computation
may experience performance degradation when using
multiple OS threads.
This is because sending data between threads involves switching
contexts, which has significant cost.
For instance, the prime sieve example
from the Go specification has no significant parallelism although it launches many
goroutines; increasing GOMAXPROCS is more likely to slow it down than
to speed it up.
Go's goroutine scheduler is not as good as it needs to be, although it
has improved in recent releases.
In the future, it may better optimize its use of OS threads.
For now, if there are performance issues,
setting GOMAXPROCS on a per-application basis may help.
For more detail on this topic see the talk entitled, Concurrency is not Parallelism.
From the Go Spec:
The method set of any other named typeTconsists of all methods with receiver typeT. The method set of the corresponding pointer type*Tis the set of all methods with receiver*TorT(that is, it also contains the method set ofT).
If an interface value contains a pointer *T,
a method call can obtain a value by dereferencing the pointer,
but if an interface value contains a value T,
there is no useful way for a method call to obtain a pointer.
Even in cases where the compiler could take the address of a value
to pass to the method, if the method modifies the value the changes
will be lost in the caller.
As an example, if the Write method of
bytes.Buffer
used a value receiver rather than a pointer,
this code:
var buf bytes.Buffer io.Copy(buf, os.Stdin)
would copy standard input into a copy of buf,
not into buf itself.
This is almost never the desired behavior.
Some confusion may arise when using closures with concurrency. Consider the following program:
func main() {
done := make(chan bool)
values := []string{"a", "b", "c"}
for _, v := range values {
go func() {
fmt.Println(v)
done <- true
}()
}
// wait for all goroutines to complete before exiting
for _ = range values {
<-done
}
}
One might mistakenly expect to see a, b, c as the output.
What you'll probably see instead is c, c, c. This is because
each iteration of the loop uses the same instance of the variable v, so
each closure shares that single variable. When the closure runs, it prints the
value of v at the time fmt.Println is executed,
but v may have been modified since the goroutine was launched.
To help detect this and other problems before they happen, run
go vet.
To bind the current value of v to each closure as it is launched, one
must modify the inner loop to create a new variable each iteration.
One way is to pass the variable as an argument to the closure:
for _, v := range values {
go func(u string) {
fmt.Println(u)
done <- true
}(v)
}
In this example, the value of v is passed as an argument to the
anonymous function. That value is then accessible inside the function as
the variable u.
Even easier is just to create a new variable, using a declaration style that may seem odd but works fine in Go:
for _, v := range values {
v := v // create a new 'v'.
go func() {
fmt.Println(v)
done <- true
}()
}
?: operator?There is no ternary testing operation in Go. You may use the following to achieve the same result:
if expr {
n = trueVal
} else {
n = falseVal
}
Put all the source files for the package in a directory by themselves. Source files can refer to items from different files at will; there is no need for forward declarations or a header file.
Other than being split into multiple files, the package will compile and test just like a single-file package.
Create a new file ending in _test.go in the same directory
as your package sources. Inside that file, import "testing"
and write functions of the form
func TestFoo(t *testing.T) {
...
}
Run go test in that directory.
That script finds the Test functions,
builds a test binary, and runs it.
See the How to Write Go Code document,
the testing package
and the go test subcommand for more details.
Go's standard testing package makes it easy to write unit tests, but it lacks
features provided in other language's testing frameworks such as assertion functions.
An earlier section of this document explained why Go
doesn't have assertions, and
the same arguments apply to the use of assert in tests.
Proper error handling means letting other tests run after one has failed, so
that the person debugging the failure gets a complete picture of what is
wrong. It is more useful for a test to report that
isPrime gives the wrong answer for 2, 3, 5, and 7 (or for
2, 4, 8, and 16) than to report that isPrime gives the wrong
answer for 2 and therefore no more tests were run. The programmer who
triggers the test failure may not be familiar with the code that fails.
Time invested writing a good error message now pays off later when the
test breaks.
A related point is that testing frameworks tend to develop into mini-languages of their own, with conditionals and controls and printing mechanisms, but Go already has all those capabilities; why recreate them? We'd rather write tests in Go; it's one fewer language to learn and the approach keeps the tests straightforward and easy to understand.
If the amount of extra code required to write
good errors seems repetitive and overwhelming, the test might work better if
table-driven, iterating over a list of inputs and outputs defined
in a data structure (Go has excellent support for data structure literals).
The work to write a good test and good error messages will then be amortized over many
test cases. The standard Go library is full of illustrative examples, such as in
the formatting tests for the fmt package.
The standard library's purpose is to support the runtime, connect to the operating system, and provide key functionality that many Go programs require, such as formatted I/O and networking. It also contains elements important for web programming, including cryptography and support for standards like HTTP, JSON, and XML.
There is no clear criterion that defines what is included because for a long time, this was the only Go library. There are criteria that define what gets added today, however.
New additions to the standard library are rare and the bar for inclusion is high. Code included in the standard library bears a large ongoing maintenance cost (often borne by those other than the original author), is subject to the Go 1 compatibility promise (blocking fixes to any flaws in the API), and is subject to the Go release schedule, preventing bug fixes from being available to users quickly.
Most new code should live outside of the standard library and be accessible
via the go tool's
go get command.
Such code can have its own maintainers, release cycle,
and compatibility guarantees.
Users can find packages and read their documentation at
godoc.org.
Although there are pieces in the standard library that don't really belong,
such as log/syslog, we continue to maintain everything in the
library because of the Go 1 compatibility promise.
But we encourage most new code to live elsewhere.
Gccgo has a front end written in C++, with a recursive descent parser coupled to the
standard GCC back end. Gc is written in Go with a recursive descent parser
and uses a custom loader, also written in Go but
based on the Plan 9 loader, to generate ELF/Mach-O/PE binaries.
We considered using LLVM for gc but we felt it was too large and
slow to meet our performance goals.
The original gc, the Go compiler, was written in C
because of the difficulties of bootstrapping—you'd need a Go compiler to
set up a Go environment.
But things have advanced and as of Go 1.5 the compiler is written in Go.
It was converted from C to Go using automatic translation tools, as
described in this design document
and a recent talk.
Thus the compiler is now "self-hosting", which means we must face
the bootstrapping problem.
The solution, naturally, is to have a working Go installation already,
just as one normally has a working C installation in place.
The story of how to bring up a new Go installation from source
is described separately.
Go is a fine language in which to implement a Go compiler.
Although gc does not use them (yet?), a native lexer and
parser are available in the go package
and there is also a type checker.
Again due to bootstrapping issues, the run-time code was originally written mostly in C (with a
tiny bit of assembler) but it has since been translated to Go
(except for some assembler bits).
Gccgo's run-time support uses glibc.
The gccgo compiler implements goroutines using
a technique called segmented stacks,
supported by recent modifications to the gold linker.
The linker in the gc tool chain
creates statically-linked binaries by default. All Go binaries therefore include the Go
run-time, along with the run-time type information necessary to support dynamic
type checks, reflection, and even panic-time stack traces.
A simple C "hello, world" program compiled and linked statically using gcc
on Linux is around 750 kB,
including an implementation of printf.
An equivalent Go program using fmt.Printf
is around 1.5 MB, but
that includes more powerful run-time support and type information.
The presence of an unused variable may indicate a bug, while unused imports just slow down compilation, an effect that can become substantial as a program accumulates code and programmers over time. For these reasons, Go refuses to compile programs with unused variables or imports, trading short-term convenience for long-term build speed and program clarity.
Still, when developing code, it's common to create these situations temporarily and it can be annoying to have to edit them out before the program will compile.
Some have asked for a compiler option to turn those checks off or at least reduce them to warnings. Such an option has not been added, though, because compiler options should not affect the semantics of the language and because the Go compiler does not report warnings, only errors that prevent compilation.
There are two reasons for having no warnings. First, if it's worth complaining about, it's worth fixing in the code. (And if it's not worth fixing, it's not worth mentioning.) Second, having the compiler generate warnings encourages the implementation to warn about weak cases that can make compilation noisy, masking real errors that should be fixed.
It's easy to address the situation, though. Use the blank identifier to let unused things persist while you're developing.
import "unused"
// This declaration marks the import as used by referencing an
// item from the package.
var _ = unused.Item // TODO: Delete before committing!
func main() {
debugData := debug.Profile()
_ = debugData // Used only during debugging.
....
}
Nowadays, most Go programmers use a tool, goimports, which automatically rewrites a Go source file to have the correct imports, eliminating the unused imports issue in practice. This program is easily connected to most editors to run automatically when a Go source file is written.
One of Go's design goals is to approach the performance of C for comparable programs, yet on some benchmarks it does quite poorly, including several in golang.org/x/exp/shootout. The slowest depend on libraries for which versions of comparable performance are not available in Go. For instance, pidigits.go depends on a multi-precision math package, and the C versions, unlike Go's, use GMP (which is written in optimized assembler). Benchmarks that depend on regular expressions (regex-dna.go, for instance) are essentially comparing Go's native regexp package to mature, highly optimized regular expression libraries like PCRE.
Benchmark games are won by extensive tuning and the Go versions of most of the benchmarks need attention. If you measure comparable C and Go programs (reverse-complement.go is one example), you'll see the two languages are much closer in raw performance than this suite would indicate.
Still, there is room for improvement. The compilers are good but could be better, many libraries need major performance work, and the garbage collector isn't fast enough yet. (Even if it were, taking care not to generate unnecessary garbage can have a huge effect.)
In any case, Go can often be very competitive. There has been significant improvement in the performance of many programs as the language and tools have developed. See the blog post about profiling Go programs for an informative example.
Other than declaration syntax, the differences are not major and stem from two desires. First, the syntax should feel light, without too many mandatory keywords, repetition, or arcana. Second, the language has been designed to be easy to analyze and can be parsed without a symbol table. This makes it much easier to build tools such as debuggers, dependency analyzers, automated documentation extractors, IDE plug-ins, and so on. C and its descendants are notoriously difficult in this regard.
They're only backwards if you're used to C. In C, the notion is that a
variable is declared like an expression denoting its type, which is a
nice idea, but the type and expression grammars don't mix very well and
the results can be confusing; consider function pointers. Go mostly
separates expression and type syntax and that simplifies things (using
prefix * for pointers is an exception that proves the rule). In C,
the declaration
int* a, b;
declares a to be a pointer but not b; in Go
var a, b *int
declares both to be pointers. This is clearer and more regular.
Also, the := short declaration form argues that a full variable
declaration should present the same order as := so
var a uint64 = 1
has the same effect as
a := uint64(1)
Parsing is also simplified by having a distinct grammar for types that
is not just the expression grammar; keywords such as func
and chan keep things clear.
See the article about Go's Declaration Syntax for more details.
Safety. Without pointer arithmetic it's possible to create a language that can never derive an illegal address that succeeds incorrectly. Compiler and hardware technology have advanced to the point where a loop using array indices can be as efficient as a loop using pointer arithmetic. Also, the lack of pointer arithmetic can simplify the implementation of the garbage collector.
++ and -- statements and not expressions? And why postfix, not prefix?
Without pointer arithmetic, the convenience value of pre- and postfix
increment operators drops. By removing them from the expression
hierarchy altogether, expression syntax is simplified and the messy
issues around order of evaluation of ++ and --
(consider f(i++) and p[i] = q[++i])
are eliminated as well. The simplification is
significant. As for postfix vs. prefix, either would work fine but
the postfix version is more traditional; insistence on prefix arose
with the STL, a library for a language whose name contains, ironically, a
postfix increment.
Go uses brace brackets for statement grouping, a syntax familiar to programmers who have worked with any language in the C family. Semicolons, however, are for parsers, not for people, and we wanted to eliminate them as much as possible. To achieve this goal, Go borrows a trick from BCPL: the semicolons that separate statements are in the formal grammar but are injected automatically, without lookahead, by the lexer at the end of any line that could be the end of a statement. This works very well in practice but has the effect that it forces a brace style. For instance, the opening brace of a function cannot appear on a line by itself.
Some have argued that the lexer should do lookahead to permit the
brace to live on the next line. We disagree. Since Go code is meant
to be formatted automatically by
gofmt,
some style must be chosen. That style may differ from what
you've used in C or Java, but Go is a new language and
gofmt's style is as good as any other. More
important—much more important—the advantages of a single,
programmatically mandated format for all Go programs greatly outweigh
any perceived disadvantages of the particular style.
Note too that Go's style means that an interactive implementation of
Go can use the standard syntax one line at a time without special rules.
One of the biggest sources of bookkeeping in systems programs is memory management. We feel it's critical to eliminate that programmer overhead, and advances in garbage collection technology in the last few years give us confidence that we can implement it with low enough overhead and no significant latency.
Another point is that a large part of the difficulty of concurrent and multi-threaded programming is memory management; as objects get passed among threads it becomes cumbersome to guarantee they become freed safely. Automatic garbage collection makes concurrent code far easier to write. Of course, implementing garbage collection in a concurrent environment is itself a challenge, but meeting it once rather than in every program helps everyone.
Finally, concurrency aside, garbage collection makes interfaces simpler because they don't need to specify how memory is managed across them.
The current implementation is a parallel mark-and-sweep collector. Recent improvements, documented in this design document, have introduced bounded pause times and improved the parallelism. Future versions might attempt new approaches.
On the topic of performance, keep in mind that Go gives the programmer considerable control over memory layout and allocation, much more than is typical in garbage-collected languages. A careful programmer can reduce the garbage collection overhead dramatically by using the language well; see the article about profiling Go programs for a worked example, including a demonstration of Go's profiling tools.