Ruleguard takes special Go files as its configuration. These files define custom rules and have strict structure that is described in this document.
The advantage of a Go-compatible syntax is having convenient tooling working for ruleguard files.
The Go code from these files is never compiled by go build and/or executed by go run. Ruleguard parses these files and creates a special internal representation that can be used to efficiently execute them on the fly.
You write the rules, ruleguard tries to execute them precisely and efficiently.
We can describe a file structure like this:
- It has a package clause (package name should be
gorules). - An import clause (at the bare minimum, you'll need
dslpackage). - Function declarations.
There are 3 kinds of functions you can declare:
- Matcher functions that define a rule group
- Custom filter functions
init()function
Every matcher function accepts exactly 1 argument, a dsl.Matcher, and defines some rules.
Every rule definition starts with a Match() or MatchComment() method call.
- For
Match(), you specify one or more AST patterns that should represent what kind of Go code a rule is supposed to match. - For
MatchComment(), you provide one or more regular expressions that should match a comment of interest.
Another mandatory part is Report() or Suggest() that describe a rule match action. Report() will print a warning message while Suggest() can be used to provide a quickfix action (a syntax rewrite pattern).
Here is a small yet useful, example of ruleguard file:
package gorules
import "github.com/quasilyte/go-ruleguard/dsl"
func regexpMust(m dsl.Matcher) { // - regexpMust matcher func
m.Match(`regexp.Compile($pat)`, // - rule | (or "regexpMust rules group")
`regexp.CompilePOSIX($pat)`). // | |
Where(m["pat"].Const). // | |
Report(`can use MustCompile for const patterns`). // | |
Suggest(`regexp.MustCompile($pat)`) // - |
} // -A Report() argument string can use $<varname> notation to interpolate the named pattern submatches into the report message.
There is a special variable $$ which can be used to inject the entire pattern match into the message (like $0 in regular expressions).
It's a good practice to add structured documentation for your rule groups.
To add such documentation, use special pragmas when commenting a matcher function.
//doc:summary reports always false/true conditions
//doc:before strings.Count(s, "/") >= 0
//doc:after strings.Count(s, "/") > 0
//doc:tags diagnostic experimental
func badCond(m dsl.Matcher) {
m.Match(`strings.Count($_, $_) >= 0`).Report(`statement always true`)
m.Match(`bytes.Count($_, $_) >= 0`).Report(`statement always true`)
}summary- short one sentence descriptionbefore- code snippet of code that will violate ruleafter- code after a fix (one that complies to the rule)tags- space separated list of custom tagsnote- extra information, like issue links
The rule is matched if:
- At least 1 AST pattern from
Match()is matched - Filters from
Where()accept the given match
There are 2 types of filters that can be used in Where() call:
- Submatch (named variable-based) filters
- Context filters (current file, etc)
A match variable describes a named submatch of a pattern.
Here are some examples of supported filters:
- Submatch expression type is identical to
T - Submatch expression type is assignable to
T - Submatch expression type implements interface
I - Submatch expression is side-effect free
- Submatch expression is a const expression
- Submatch expression const value check
- Submatch text matches provided regexp
- Current files imports package
P
A match variable can be accessed with dsl.Matcher function argument indexing:
// m["a"] -- $a
// m["b"] -- $b
Where(m["a"].Type.Is(`int`) && !m["b"].Type.AssignableTo(`[]string`))If we had a pattern with $a and $b match variables, a filter above would only accept it
if $a expression had a type of int while $b is anything that is not assignable to []string.
Context-related filters can be applied through m members:
// Using m.File() to apply a file-related filter.
Where(m.File().Imports("io/ioutil"))When using MatchComment, submatches will have a type of *ast.Comment. Text-related filters can be used as usual.
The filter concept is crucial to avoid false-positives in rules.
Please refer to the godoc page of a dsl package to get an up-to-date list of supported filters.
When none of the DSL filters seem to do what you want, you can write a custom filter function.
package gorules
import (
"github.com/quasilyte/go-ruleguard/dsl"
"github.com/quasilyte/go-ruleguard/dsl/types"
)
func implementsStringer(ctx *dsl.VarFilterContext) bool {
stringer := ctx.GetInterface(`fmt.Stringer`)
return types.Implements(ctx.Type, stringer) ||
types.Implements(types.NewPointer(ctx.Type), stringer)
}
func stringerLiteral(m dsl.Matcher) {
m.Match(`$x{$*_}`).
Where(m["x"].Filter(implementsStringer)).
Report("$x implements stringer")
}Suppose that we have this Go file we want to check:
package target
type byValue struct{}
type byPtr struct{}
func (*byPtr) String() string { return "" }
func (byValue) String() string { return "" }
func f() {
_ = &byValue{}
_ = byValue{}
_ = &byPtr{}
_ = byPtr{} // does not implement fooer, but we still want it
}We'll get this output if we apply a stringerLiteral rule:
example.go:10:7: stringerLiterals: byValue implements stringer
example.go:11:6: stringerLiterals: byValue implements stringer
example.go:12:7: stringerLiterals: byPtr implements stringer
example.go:13:6: stringerLiterals: byPtr implements stringer
Custom filter functions are byte-compiled and interpreted like a scripting language. There are some limitations in the implementations; if you would like to see some feature to be implemented, please tell about it.
When you use a type filter, the parser must know how to match a given type against the type that is going to be found during execution.
In normal Go programs, the unqualified type name like Foo makes sense, it refers to a current package symbol table.
In ruleguard files, you either have to Import() the package and use the qualified name or use a fully-qualified name without import.
Here are two ways to use a Bar type from the foo/bar package in filters:
func qualifiedName(m dsl.Matcher) {
m.Import(`foo/bar`)
m.Match(`f($x)`).Where(m["x"].Type.Is(`bar.Baz`)).Report("$x is bar.Baz")
}
func fullyQualifiedName(m dsl.Matcher) {
m.Match(`f($x)`).
Where(m["x"].Type.Is(`foo/bar.Baz`)).
Report("$x is bar.Baz")
}For convenience, stdlib packages are pre-loaded into the imports table.
There are some collisions in the standard library, like text/template and html/template. By default, template is imported as text/template, but if you want template.Template to refer to the HTML package template, you can override the default imports by doing an explicit import:
m.Import(`html/template`)
// Now template.Template refers to a type from html/template package.Methods like ExprType.Is() accept a string argument that describes a Go type. It can be as simple as "[]string" that matches only a string slice, but it can also include a pattern-like variables:
[]$Tmatches any slice.[$len]$Tmatches any array.map[$K]$Vmatches any map.map[$T]$Tmatches a map where a key and value types are the same.struct{$*_}any struct type.struct{$x; $*_}struct that has $x-typed first field.struct{$*_; $x; $*_}struct that contains $x-typed field.struct{$*_; $x}struct that has $x-typed last field.
Note: when matching types, make sure to think whether you need to match a type or the underlying type.
To match the underlying type, use ExprType.Underlying() method.
You may recognize that it's the same pattern behavior as in AST patterns.
Some rules have a clear suggestion that can be used as a direct replacement of a matched code.
Consider this rule:
m.Match(`!!$x`).Report(`can simplify !!$x to $x`)It contains a fix inside its message. The user will have to fix it manually, by copy/pasting the suggestion from the report message. We can do better.
m.Match(`!!$x`).Suggest(`$x`)Now we're enabled the -fix support for that rule. If ruleguard is invoked with that argument,
code from the Suggest argument will replace the matched code chunk.
You can have both Report() and Suggest(), so the user can also have a more detailed
warning message when not using -fix.
When you use Suggest() and omit Report(), suggested string is used as a foundation of a report message.
The following 2 lines are identical:
m.Match(`!!$x`).Suggest(`$x`)
m.Match(`!!$x`).Suggest(`$x`).Report(`suggested: $x`)Be careful when using Suggest() with MatchComment(). As regexp may match a subset of the comment, you'll replace that exact comment portion with Suggest() pattern. If you want to replace an entire comment, be sure that your pattern contains ^ and $ anchors.
If you want to use a ruleguard file that is written by someone else, you have 2 main options:
- Copy the file and keep it somewhere inside your repository
- Import that ruleguard file as a bundle
The latter option gives you extra benefits:
- Versioning is easier. You can pin a bundle version in your
go.modfile - You can add your own rules without a risk of running into collisions
Bundles are installed with go get.
Here is how we can install the github.com/quasilyte/ruleguard-rules-test bundle:
# Make sure that Go modules are turned on.
export GO111MODULE=on
go get -v -u github.com/quasilyte/ruleguard-rules-test@masterIf your ruleguard file has a // +build ignore build tag, go get would install a bundle as indirect dependency. You'll probably want to use a tools.go idiom to make it a direct dependency.
// +build tools
package tools
import (
_ "github.com/quasilyte/go-ruleguard/dsl"
)If rules.go file has no ignore tag, go get should work properly without extra efforts.
Ruleguard bundle becomes your project explicit dependency that you can use in your ruleguard files. Note that it does not make your project bloated: bundle packages are never used in your builds.
In case if you don't want to have a direct bundle dependency, run a go get before running a ruleguard and then remove the installed package (go mod tidy will be enough if bundle package is an indirect dependency).
Installed bundle packages can be imported as normal Go packages.
Importing the bundle package is not enough to add its rules, you need to call ImportRules() function for that.
package gorules
import (
"github.com/quasilyte/go-ruleguard/dsl"
quasilyterules "github.com/quasilyte/ruleguard-rules-test"
)
func init() {
// Imported rules will have a "qrules" prefix.
dsl.ImportRules("qrules", quasilyterules.Bundle)
}
// Then you can define your own rules.
func emptyStringTest(m dsl.Matcher) {
m.Match(`len($s) == 0`).
Where(m["s"].Type.Is("string")).
Report(`maybe use $s == "" instead?`)
m.Match(`len($s) != 0`).
Where(m["s"].Type.Is("string")).
Report(`maybe use $s != "" instead?`)
}Let's try running that file:
$ ruleguard -rules rules.go test.go
test.go:4:6: emptyStringTest: maybe use s == "" instead? (rules.go:13)
test.go:5:6: qrules/boolComparison: omit bool literal in expression (rules1.go:8)It’s possible to use an empty ("") prefix, but you’ll risk getting a name collision. If you don’t define your own rules, then it’s perfectly fine to use an empty prefix.
A package that exports rules must define a Bundle object:
// Bundle holds the rules package metadata.
//
// In order to be importable from other gorules package,
// a package must define a Bundle variable.
var Bundle = dsl.Bundle{}That package should be a separate Go module. A rules bundle is versioned by its Go module.
It's possible to have several ruleguard files inside one Go module. Only one file should define a Bundle object. During a bundle import, all files will be exported.