Subset of table driven test - unit-testing

Question

For testing functions I could select which will run by option -run.
go test -run regex
Very common if we have dozens test cases is put it into array in order not to write function for each of that:
cases := []struct {
arg, expected string
} {
{"%a", "[%a]"},
{"%-a", "[%-a]"},
// and many others
}
for _, c := range cases {
res := myfn(c.arg)
if res != c.expected {
t.Errorf("myfn(%q) should return %q, but it returns %q", c.arg, c.expected, res)
}
}
This work good, but problem is with maintanance. When I add a new testcase, while debugging I want to start just a new test case, but I cannot say something like:
go test -run TestMyFn.onlyThirdCase
Is there any elegant way, how to have many testcases in array together with ability to choose which testcase will run?

Answer 1

With Go 1.6 (and below)
This is not supported directly by the testing package in Go 1.6 and below. You have to implement it yourself.
But it's not that hard. You can use flag package to easily access command line arguments.
Let's see an example. We define an "idx" command line parameter, which if present, only the case at that index will be executed, else all test cases.
Define flag:
var idx = flag.Int("idx", -1, "specify case index to run only")
Parse command line flags (actually, this is not required as go test already calls this, but just to be sure / complete):
func init() {
flag.Parse()
}
Using this parameter:
for i, c := range cases {
if *idx != -1 && *idx != i {
println("Skipping idx", i)
continue
}
if res := myfn(c.arg); res != c.expected {
t.Errorf("myfn(%q) should return %q, but it returns %q", c.arg, c.expected, res)
}
}
Testing it with 3 test cases:
cases := []struct {
arg, expected string
}{
{"%a", "[%a]"},
{"%-a", "[%-a]"},
{"%+a", "[%+a]"},
}
Without idx parameter:
go test
Output:
PASS
ok play 0.172s
Specifying an index:
go test -idx=1
Output:
Skipping idx 0
Skipping idx 2
PASS
ok play 0.203s
Of course you can implement more sophisticated filtering logic, e.g. you can have minidx and maxidx flags to run cases in a range:
var (
minidx = flag.Int("minidx", 0, "min case idx to run")
maxidx = flag.Int("maxidx", -1, "max case idx to run")
)
And the filtering:
if i < *minidx || *maxidx != -1 && i > *maxidx {
println("Skipping idx", i)
continue
}
Using it:
go test -maxidx=1
Output:
Skipping idx 2
PASS
ok play 0.188s
Starting with Go 1.7
Go 1.7 (to be released on August 18, 2016) adds the definition of subtests and sub-benchmarks:
The testing package now supports the definition of tests with subtests and benchmarks with sub-benchmarks. This support makes it easy to write table-driven benchmarks and to create hierarchical tests. It also provides a way to share common setup and tear-down code. See the package documentation for details.
With that, you can do things like:
func TestFoo(t *testing.T) {
// <setup code>
t.Run("A=1", func(t *testing.T) { ... })
t.Run("A=2", func(t *testing.T) { ... })
t.Run("B=1", func(t *testing.T) { ... })
// <tear-down code>
}
Where the subtests are named "A=1", "A=2", "B=1".
The argument to the -run and -bench command-line flags is a slash-separated list of regular expressions that match each name element in turn. For example:
go test -run Foo # Run top-level tests matching "Foo".
go test -run Foo/A= # Run subtests of Foo matching "A=".
go test -run /A=1 # Run all subtests of a top-level test matching "A=1".
How does this help your case? The names of subtests are string values, which can be generated on-the-fly, e.g.:
for i, c := range cases {
name := fmt.Sprintf("C=%d", i)
t.Run(name, func(t *testing.T) {
if res := myfn(c.arg); res != c.expected {
t.Errorf("myfn(%q) should return %q, but it returns %q",
c.arg, c.expected, res)
}
})
}
To run the case at index 2, you could start it like
go test -run /C=2
or
go test -run TestName/C=2

Answer 2

I wrote a simple code, that work fine with both, although with a bit different command line options. Version for 1.7 is:
// +build go1.7
package plist
import "testing"
func runTest(name string, fn func(t *testing.T), t *testing.T) {
t.Run(name, fn)
}
and 1.6 and older:
// +build !go1.7
package plist
import (
"flag"
"testing"
"runtime"
"strings"
"fmt"
)
func init() {
flag.Parse()
}
var pattern = flag.String("pattern", "", "specify which test(s) should be executed")
var verbose = flag.Bool("verbose", false, "write whether test was done")
// This is a hack, that a bit simulate t.Run available from go1.7
func runTest(name string, fn func(t *testing.T), t *testing.T) {
// obtain name of caller
var pc[10]uintptr
runtime.Callers(2, pc[:])
var fnName = ""
f := runtime.FuncForPC(pc[0])
if f != nil {
fnName = f.Name()
}
names := strings.Split(fnName, ".")
fnName = names[len(names)-1] + "/" + name
if strings.Contains(fnName, *pattern) {
if *verbose {
fmt.Printf("%s is executed\n", fnName)
}
fn(t)
} else {
if *verbose {
fmt.Printf("%s is skipped\n", fnName)
}
}
}