Source code analyzer that helps you to make your Go programs more consistent.
This install the go-consistent
binary:
go install github.com/quasilyte/go-consistent
If go install location is under your system $PATH
, go-consistent
command should be available after that.
This should print the help message:
go-consistent --help
You can pass package names and separate Go filenames to the go-consistent
tool:
go-consistent foo.go bar.go mypkg
You can also use std
, ./...
and other conventional targets that are normally
understood by Go tools.
- If you want to check consistency of a single file or package, just provide their name
- If you want to check the whole project, you should pass all its packages as an arguments
To check the entire project, run go-consistent
like this:
go-consistent -v ./...
To understand what go-consistent
does, take a look at these 3 lines of code:
lit1 := map[K]V{}
lit2 := map[K]V{}
m := make(map[K]V)
lit1
, lit2
and m
are initialized to an empty, non-nil map.
The problem is that you have at least 2 ways to do it:
lit1
andlit2
use the first option, the map literalm
uses the second option, themake
function
Neither of these are the "best", but on the package or project level, you might want to prefer only one of them, for consistency reasons.
go-consistent
tool detects that map literal used more frequently (2 vs 1) in the example above,
so it suggest you to replace m
initialization expression to use map literal instead of make
function.
There are many similar cases where you have 2 or more options of expressing the same thing in Go,
go-consistent
tries to handle as much patterns as possible.
- Zero-configuration. Defaults should be good enough for most users. Other configuration is performed using command line arguments.
- Can handle projects of any size. This means that there should be no significant memory consumption growth with the increased number of packages being checked. There can be "fast, but memory-hungry" option that can work best for small-average projects, but it should be always possible to check huge projects on the developer machine.
Checkers that require types info:
Checkers that do not require types info:
// A: no parenthesis
import "fmt"
// B: with parenthesis
import (
"fmt"
)
// A: new call
new(T)
new([]T)
// B: address of literal
&T{}
&[]T{}
// A: make call
make([]T, 0)
// B: slice literal
[]T{}
// A: make call
make(map[K]V)
// B: map literal
map[K]V{}
// A: lower case a-f digits
0xff
// B: upper case A-F digits
0xFF
// A: left-aligned
x > low && x < high
// B: center-aligned
low < x && x < high
// A: using &^ operator (no space)
x &^ y
// B: using & and ^ (with space)
x & ^y
// A: explicit int/frac parts
0.0
1.0
// B: implicit int/frac parts
.0
1.
// A: all upper case
LABEL_NAME:
// B: upper camel case
LabelName:
// C: lower camel case
labelName:
// A: LHS type
var x int32 = 10
const y float32 = 1.6
// B: RHS type
var x = int32(10)
const y = float32(1.6)
// A: closing parenthesis on the same line
multiLineCall(
a,
b,
c)
// B: closing parenthesis on the next line
multiLineCall(
a,
b,
c,
)
// A: compare as "number of elems not equal to zero"
len(xs) != 0
// B: compare as "more than 0 elements"
len(xs) > 0
// C: compare as "at least 1 element"
len(xs) >= 1
// A: default case is the first one
switch {
default:
return "?"
case x > 10:
return "more than 10"
}
// B: default case is the last one
switch {
case x > 10:
return "more than 10"
default:
return "?"
}