Go Language Modularization
Internal
Overview
Go modularization builds upon the concepts of package and module. Packages provide a namespace for their members, and they are a way to encapsulate code, hide implementation details and only expose features, such as variables, functions or type definitions that are meant to be publicly consumed. Packages can be published as part of modules. Modules have been introduced in Go 1.11.
Packages
Standard Library
Go comes with a set of over 100 "built-in" packages, which are available as part of the locally installed Go development environment.
Standard library package documentation is available online here:
The standard library is a good source of code examples, comments and style guidance.
Standard library packages:
pkg.go.dev
The place to look for published third-party packages is
Modules
Module-Aware or GOPATH Mode
The compiler must locate packages on the local file system every time it handles an import statement.
The go tool has two modes of resolving package dependencies: module-aware mode or GOPATH mode.
In module-aware mode, the go commands use go.mod files to find versioned dependencies and typically load packages out of the module cache, downloading modules if they are missing. As of Go 1.16, the module-aware mode is enabled by default, regardless of whether go.mod is present or not. The behavior can be controlled with the GO111MODULE environment variable.
In GOPATH mode, go commands use the value of the GOPATH environment variable and vendor directories to resolve packages.
Also see:
Workspace
The workspace is a concept introduced in Go 1.18. A workspace allows organizing the code for a project that has several modules which share a common list of dependencies. The workspace maintains metadata, especially dependency metadata, in a file called go.work. The dependencies declared in this file can span modules and anything declared in go.work will override dependencies in the modules's go.mod. The packages and modules maintained in a workspace are managed with the go tool.
A workspace may contain multiple projects.
The standard workspace layout is:
. ← GOPATH should point to this directory, it contains src, pkg and bin │ ├─ src │ ├─ a │ │ └─ b │ │ └─color # "color" package directory, with the "a/b/color" import path │ │ ├─ colors.go │ │ ├─ aux.go │ │ └─ ... │ │ │ ├─ weight # "weight" package directory, with the "weight" import path │ │ ├─ weights.go │ │ ├─ aux.go │ │ └─ ... │ │ │ ├─ novaordis.com │ │ └─ tools │ │ └─ hammer # "hammer" package directory, with the "novaordis.com/tools/hammer" import path │ │ └─ ... │ │ │ └─ github.com │ └─ blue-org │ └─ tools │ └─ wrench # "wrench" package directory, with the "github.com/blue-org/tools/wrench" import path │ ├─ .git │ └─ ... │ ├─ pkg │ └─ darwin_amd64 │ ├─ weights.go │ ├─ a/b/color.a │ ├─ novaordis.com/tools/hammer.a │ └─ github.com/bue-org/tools/wrench.a └─ bin
Content
src
The src subdirectory holds source code. Each package resides in a directory whose name relative to ${GOPATH}/src is the package's import path.
The src subdirectory may contain multiple version-control repository workareas.
pkg
The build tool stores compiled packages in the pkg directory, under ${GOOS}_${GOARCH} subdirectories.
bin
The bin directory is where the executables are stored.
Relationship between Workspace and GOPATH
GOPATH should point to the root of the workspace, the directory that contains src, pkg and bin.
Program
Go programs are constructed by linking together packages. There must be a main package, which contains the main(), to trigger the linker.
Project
Repository
A Go repository typically contains only one module, located in the root of the repository.
TO DEPLETE
Overview
A package is a unit of compiled code and a namespace: it consists in one or more source files that declare constants, variables, types and functions belonging to the package and which are accessible in all files of the same package. Packages are Go's way of organizing and reusing code, according to the principle Don't Repeat Yourself.
Every Go source code must belong to a package, and it must declare the package it belongs to with the package keyword.
All source file declared as being part of the same package must be in the same file system directory. You cannot split a package across multiple directories.
A directory cannot contain source files for more than one package (or the compiler will complain).
Identifiers declared by a package can be exported, and in order to refer those identifiers from other packages, the enclosing package must be imported and the identifiers must be prefixed with the name of the package:
package B
import "A"
...
func something() {
A.SomeFunctionExportedByA()
}
Each package can be imported and used individually, so developers can import only the functionality they need.
The benefits offered by the package mechanism are:
- Different namespaces allow for identifiers with the same name - as long as they belong to different namespaces. This allows keeping the identifier names short and succinct.
- It speeds up the compiler by only requiring compilation of smaller chunks of the program. For example, when we use "fmt" we don't need to recompile it every time we change our program.
It is common in Go to see packages that are relatively small: they have a small API and perform a single task. This is normal and expected.
The package keyword
The package keyword is used in package declarations. All Go files must start with a package declaration, which consists start with the package keyword and it is followed by the package name:
package blue ...
Package Names
A package name is an identifier, which means it consists in a sequence of one more letter, digits and possibly underscores, where the first character is a letter. Even if we might be tempted to think about packages hierarchically, "project1/blue" is NOT a package name, even if, as we will see below, we will use "project1/blue" string literal to import the package.
Conventionally, the name of the directory that contains the package files should coincide with the name of the package. It is usually a good idea to follow the convention: while compiling the file, the compiler won't complain if the convention is not followed, but it will run into troubles if other packages that are referred do not follow the convention. For more details, see GOPATH.
Note that in order to import the package, the package name may not be sufficient and we may need to use a path fragment in the import statement: the string literal that follows the import keyword and the package name are two different things. For more details on importing packages, see import.
Package Name Conventions
- A package name should coincide with the name of the directory that holds the files.
- The package names should be short, concise, lowercase names.
Writing a Package
Place the package files under a directory hierarchy that makes it obvious where the package comes from, if imported. For example, if planning to write a reusable "blue" package as part of the "airplane" project, place the package source under:
$LOCAL_PROJECT_HOME/src/io/novaordis/airplane/blue
Thus the package will be imported as:
import io/novaordis/airplane/blue
Exported Identifiers
In order to be visible outside the declaring package, an identifier must be exported. This is achieved by any of the following is true:
- The identifier's first character is an upper case letter.
- The identifier is declared in the package block or it is a field name or a method name of an exported type.
An identifier that is not explicitly exported is unexported.
Note that identifiers, and not values, can be exported or unexported.
It is a good practice to only expose the package elements that we explicitly want other package to use, and hide everything else. This allows to change the hidden parts later, without affecting the package's consumers.
Note that just because an identifier is unexported, it does not mean other packages can't indirectly access it: for example a function can return the value of an unexported type and this value is accessible by an calling function, even if the calling function has been declared in a different package. This can be coupled with the short declaration operator (:=), which infers the type of the variable. A variable of that type cannot be explicitly declared with var.
struct Fields Export
The fields of an exported struct type can be exported or unexported on a field basis, by naming them with an upper case and respectively lower case letter.
Embedded Type Export
If the name of an inner type starts with a lower case, it is unexported even if its outer type is exported. Even if the name of the inner type is not exported, its fields may be individually exported, and thus accessible from outside the package.
For more details on embedded types see:
The "main" Package
A package named "main" that exposes a main() function is used by the Go linker as a start point for walking the package dependency graph when building the binary executable. All of the executable programs must have a package called "main", and their "main" package must have a main() function, otherwise the executable won't be built. More about the main() function, the main package and Go executables can be found here: compiling an executable.
Package Block
Each package has a package block containing all source text for that package. More about blocks is available here:
Vendoring
"Vendoring" is the act of making a local copy of a third party package your project depends on. This copy is traditionally placed inside each project and then saved in the project repository.
Vendoring is using local copies of external dependencies to satisfy imports of those dependencies.
For a directory structure referred by GOPATH, code below a directory named "vendor" is importable only by code in the directory tree rooted at the parent of "vendor", and only using an import path that omits the prefix up to and including the vendor element.
Code in vendor directories deeper in the source tree shadows code in higher directories. Within the subtree rooted at foo, an import of "crash/bang" resolves to "foo/vendor/crash/bang", not the top-level "crash/bang". Code in vendor directories is not subject to import path checking.
When 'go get' checks out or updates a git repository, it now also updates submodules.
Vendor directories do not affect the placement of new repositories being checked out for the first time by 'go get': those are always placed in the main GOPATH, never in a vendor subtree.