Documentation
¶
Overview ¶
Package generator defines an interface for code generators to implement.
To use this package, you'll implement the "Package" and "Generator" interfaces; you'll call NewContext to load up the types you want to work with, and then you'll call one or more of the Execute methods. See the interface definitions for explanations. All output will have gofmt called on it automatically, so you do not need to worry about generating correct indentation.
This package also exposes SnippetWriter. SnippetWriter reduces to a minimum the boilerplate involved in setting up a template from go's text/template package. Additionally, all naming systems in the Context will be added as functions to the parsed template, so that they can be called directly from your templates!
Index ¶
- type Context
- type DefaultGen
- func (d DefaultGen) Filename() string
- func (d DefaultGen) Filter(*Context, *types.Type) bool
- func (d DefaultGen) GenerateType(*Context, *types.Type, io.Writer) error
- func (d DefaultGen) Imports(*Context) []string
- func (d DefaultGen) Init(c *Context, w io.Writer) error
- func (d DefaultGen) Name() string
- func (d DefaultGen) Namers(*Context) namer.NameSystems
- func (d DefaultGen) PackageConsts(*Context) []string
- func (d DefaultGen) PackageVars(*Context) []string
- type DefaultPackage
- type ErrorTracker
- type Generator
- type ImportTracker
- type Package
- type Packages
- type SnippetWriter
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Context ¶
type Context struct {
// A map from the naming system to the names for that system. E.g., you
// might have public names and several private naming systems.
Namers namer.NameSystems
// All the types, in case you want to look up something.
Universe types.Universe
// The canonical ordering of the types (will be filtered by both the
// Package's and Generator's Filter methods).
Order []*types.Type
}
Context is global context for individual generators to consume.
func NewContext ¶
func NewContext(b *parser.Builder, nameSystems namer.NameSystems, canonicalOrderName string) (*Context, error)
NewContext generates a context from the given builder, naming systems, and the naming system you wish to construct the canonical ordering from.
func (*Context) ExecutePackage ¶
ExecutePackage executes a single package. 'outDir' is the base directory in which to place the package; it should be a physical path on disk, not an import path. e.g.: '/path/to/home/path/to/gopath/src/' The package knows its import path already, this will be appended to 'outDir'.
func (*Context) ExecutePackages ¶
ExecutePackages runs the generators for every package in 'packages'. 'outDir' is the base directory in which to place all the generated packages; it should be a physical path on disk, not an import path. e.g.: /path/to/home/path/to/gopath/src/ Each package has its import path already, this will be appended to 'outDir'.
type DefaultGen ¶
type DefaultGen struct {
// OptionalName, if present, will be used for the generator's name, and
// the filename (with ".go" appended).
OptionalName string
// OptionalBody, if present, will be used as the return from the "Init"
// method. This causes it to be static content for the entire file if
// no other generator touches the file.
OptionalBody []byte
}
DefaultGen implements a do-nothing Generator.
It can be used to implement static content files.
func (DefaultGen) Filename ¶
func (d DefaultGen) Filename() string
func (DefaultGen) GenerateType ¶
func (DefaultGen) Imports ¶
func (d DefaultGen) Imports(*Context) []string
func (DefaultGen) Name ¶
func (d DefaultGen) Name() string
func (DefaultGen) Namers ¶
func (d DefaultGen) Namers(*Context) namer.NameSystems
func (DefaultGen) PackageConsts ¶
func (d DefaultGen) PackageConsts(*Context) []string
func (DefaultGen) PackageVars ¶
func (d DefaultGen) PackageVars(*Context) []string
type DefaultPackage ¶
type DefaultPackage struct {
// Short name of package, used in the "package xxxx" line.
PackageName string
// Import path of the package, and the location on disk of the package.
PackagePath string
// Emitted at the top of every file.
HeaderText []byte
// Emitted only for a "doc.go" file; appended to the HeaderText for
// that file.
PackageDocumentation []byte
// If non-nil, will be called on "Generators"; otherwise, the static
// list will be used. So you should set only one of these two fields.
GeneratorFunc func(*Context) []Generator
GeneratorList []Generator
// Optional; filters the types exposed to the generators.
FilterFunc func(*Context, *types.Type) bool
}
DefaultPackage contains a default implentation of Package.
func (*DefaultPackage) Generators ¶
func (d *DefaultPackage) Generators(c *Context) []Generator
func (*DefaultPackage) Header ¶
func (d *DefaultPackage) Header(filename string) []byte
func (*DefaultPackage) Name ¶
func (d *DefaultPackage) Name() string
func (*DefaultPackage) Path ¶
func (d *DefaultPackage) Path() string
type ErrorTracker ¶
ErrorTracker tracks errors to the underlying writer, so that you can ignore them until you're ready to return.
func NewErrorTracker ¶
func NewErrorTracker(w io.Writer) *ErrorTracker
NewErrorTracker makes a new error tracker; note that it implements io.Writer.
func (*ErrorTracker) Error ¶
func (et *ErrorTracker) Error() error
Error returns nil if no error has occurred, otherwise it returns the error.
type Generator ¶
type Generator interface {
// The name of this generator. Will be included in generated comments.
Name() string
// Filter should return true if this generator cares about this type.
// (otherwise, GenerateType will not be called.)
//
// Filter is called before any of the generator's other functions;
// subsequent calls will get a context with only the types that passed
// this filter.
Filter(*Context, *types.Type) bool
// If this generator needs special namers, return them here. These will
// override the original namers in the context if there is a collision.
// You may return nil if you don't need special names. These names will
// be available in the context passed to the rest of the generator's
// functions.
//
// A use case for this is to return a namer that tracks imports.
Namers(*Context) namer.NameSystems
// Init should write an init function, and any other content that's not
// generated per-type. (It's not intended for generator specific
// initialization! Do that when your Package constructs the
// Generators.)
Init(*Context, io.Writer) error
// PackageVars should emit an array of variable lines. They will be
// placed in a var ( ... ) block. There's no need to include a leading
// \t or trailing \n.
PackageVars(*Context) []string
// PackageConsts should emit an array of constant lines. They will be
// placed in a const ( ... ) block. There's no need to include a leading
// \t or trailing \n.
PackageConsts(*Context) []string
// GenerateType should emit the code for a particular type.
GenerateType(*Context, *types.Type, io.Writer) error
// Imports should return a list of necessary imports. They will be
// formatted correctly. You do not need to include quotation marks,
// return only the package name; alternatively, you can also return
// imports in the format `name "path/to/pkg"`. Imports will be called
// after Init, PackageVars, PackageConsts, and GenerateType, to allow
// you to keep track of what imports you actually need.
Imports(*Context) []string
// Preferred file name of this generator, not including a path. It is
// allowed for multiple generators to use the same filename, but it's
// up to you to make sure they don't have colliding import names.
// TODO: provide per-file import tracking, removing the requirement
// that generators coordinate..
Filename() string
}
Generator is the contract for anything that wants to do auto-generation. It's expected that the io.Writers passed to the below functions will be ErrorTrackers; this allows implementations to not check for io errors, making more readable code.
The call order for the functions that take a Context is: 1. Filter() // Subsequent calls see only types that pass this. 2. Namers() // Subsequent calls see the namers provided by this. 3. PackageVars() 4. PackageConsts() 5. Init() 6. GenerateType() // Called N times, once per type in the context's Order. 7. Imports()
You may have multiple generators for the same file.
type ImportTracker ¶
type ImportTracker struct {
// contains filtered or unexported fields
}
ImportTracker may be passed to a namer.RawNamer, to track the imports needed for the types it names.
TODO: pay attention to the package name (instead of renaming every package). TODO: Figure out the best way to make names for packages that collide.
func NewImportTracker ¶
func NewImportTracker(types ...*types.Type) *ImportTracker
func (*ImportTracker) AddType ¶
func (tracker *ImportTracker) AddType(t *types.Type)
func (*ImportTracker) AddTypes ¶
func (tracker *ImportTracker) AddTypes(types ...*types.Type)
func (*ImportTracker) ImportLines ¶
func (tracker *ImportTracker) ImportLines() []string
func (*ImportTracker) LocalNameOf ¶
func (tracker *ImportTracker) LocalNameOf(path string) string
LocalNameOf returns the name you would use to refer to the package at the specified path within the body of a file.
type Package ¶
type Package interface {
// Name returns the package short name.
Name() string
// Path returns the package import path.
Path() string
// Filter should return true if this package cares about this type.
// Otherwise, this type will be ommitted from the type ordering for
// this package.
Filter(*Context, *types.Type) bool
// Header should return a header for the file, including comment markers.
// Useful for copyright notices and doc strings. Include an
// autogeneration notice! Do not include the "package x" line.
Header(filename string) []byte
// Generators returns the list of generators for this package. It is
// allowed for more than one generator to write to the same file.
// A Context is passed in case the list of generators depends on the
// input types.
Generators(*Context) []Generator
}
Package contains the contract for generating a package.
type SnippetWriter ¶
type SnippetWriter struct {
// contains filtered or unexported fields
}
SnippetWriter is an attempt to make the template library usable. Methods are chainable, and you don't have to check Error() until you're all done.
func NewSnippetWriter ¶
func NewSnippetWriter(w io.Writer, c *Context, left, right string) *SnippetWriter
w is the destination; left and right are the delimiters; @ and $ are both reasonable choices.
c is used to make a function for every naming system, to which you can pass a type and get the corresponding name.
func (*SnippetWriter) Do ¶
func (s *SnippetWriter) Do(format string, args interface{}) *SnippetWriter
Do parses format and runs args through it. You can have arbitrary logic in the format (see the text/template documentation), but consider running many short templaces, with ordinary go logic in between--this may be more readable. Do is chainable. Any error causes every other call to do to be ignored, and the error will be returned by Error(). So you can check it just once, at the end of your function.
'args' can be quite literally anything; read the text/template documentation for details. Maps and structs work particularly nicely. Conveniently, the types package is designed to have structs that are easily referencable from the template language.
Example:
sw := generator.NewSnippetWriter(outBuffer, context, "$", "$") sw.Do(`The public type name is: $.type|public$`, map[string]interface{}{"type": t}) return sw.Error()
Where:
- "$" starts a template directive
- "." references the entire thing passed as args
- "type" therefore sees a map and looks up the key "type"
- "|" means "pass the thing on the left to the thing on the right"
- "public" is the name of a naming system, so the SnippetWriter has given the template a function called "public" that takes a *types.Type and returns the naming system's name. E.g., if the type is "string" this might return "String".
- the second "$" ends the template directive.
The map is actually not necessary. The below does the same thing:
sw.Do(`The public type name is: $.|public$`, t)
You may or may not find it more readable to use the map with a descriptive key, but if you want to pass more than one arg, the map or a custom struct becomes a requirement. You can do arbitrary logic inside these templates, but you should consider doing the logic in go and stitching them together for the sake of your readers.
func (*SnippetWriter) Error ¶
func (s *SnippetWriter) Error() error
Error returns any encountered error.
Source Files