add support for Func() and BoolFunc() #426

Add support for two features which landed in the 'flag' package
from the standard library: Func() and BoolFunc()
and their two pflag specific versions: FuncP() and BoolFuncP()

fixes #426

1 files changed, 147 insertions, 0 deletions

diff --git a/bool_func_test.go b/bool_func_test.go
new file mode 100644
index 0000000..c81970a
--- /dev/null
+++ b/bool_func_test.go
@@ -0,0 +1,147 @@
+package pflag
+
+import (
+	"errors"
+	"flag"
+	"io"
+	"strings"
+	"testing"
+)
+
+func TestBoolFunc(t *testing.T) {
+	var count int
+	fn := func(_ string) error {
+		count++
+		return nil
+	}
+
+	fset := NewFlagSet("test", ContinueOnError)
+	fset.BoolFunc("func", "Callback function", fn)
+
+	err := fset.Parse([]string{"--func", "--func=1", "--func=false"})
+	if err != nil {
+		t.Fatal("expected no error; got", err)
+	}
+
+	if count != 3 {
+		t.Fatalf("expected 3 calls to the callback, got %d calls", count)
+	}
+}
+
+func TestBoolFuncP(t *testing.T) {
+	var count int
+	fn := func(_ string) error {
+		count++
+		return nil
+	}
+
+	fset := NewFlagSet("test", ContinueOnError)
+	fset.BoolFuncP("bfunc", "b", "Callback function", fn)
+
+	err := fset.Parse([]string{"--bfunc", "--bfunc=0", "--bfunc=false", "-b", "-b=0"})
+	if err != nil {
+		t.Fatal("expected no error; got", err)
+	}
+
+	if count != 5 {
+		t.Fatalf("expected 5 calls to the callback, got %d calls", count)
+	}
+}
+
+func TestBoolFuncCompat(t *testing.T) {
+	// compare behavior with the stdlib 'flag' package
+	type BoolFuncFlagSet interface {
+		BoolFunc(name string, usage string, fn func(string) error)
+		Parse([]string) error
+	}
+
+	unitTestErr := errors.New("unit test error")
+	runCase := func(f BoolFuncFlagSet, name string, args []string) (values []string, err error) {
+		fn := func(s string) error {
+			values = append(values, s)
+			if s == "err" {
+				return unitTestErr
+			}
+			return nil
+		}
+		f.BoolFunc(name, "Callback function", fn)
+
+		err = f.Parse(args)
+		return values, err
+	}
+
+	t.Run("regular parsing", func(t *testing.T) {
+		flagName := "bflag"
+		args := []string{"--bflag", "--bflag=false", "--bflag=1", "--bflag=bar", "--bflag="}
+
+		// It turns out that, even though the function is called "BoolFunc",
+		// the stanard flag package does not try to parse the value assigned to
+		// that cli flag as a boolean. The string provided on the command line is
+		// passed as is to the callback.
+		//   e.g: with "--bflag=not_a_bool" on the command line, the FlagSet does not
+		// generate an error stating "invalid boolean value", and `fn` will be called
+		// with "not_a_bool" as an argument.
+
+		stdFSet := flag.NewFlagSet("std test", flag.ContinueOnError)
+		stdValues, err := runCase(stdFSet, flagName, args)
+		if err != nil {
+			t.Fatalf("std flag: expected no error, got %v", err)
+		}
+		expected := []string{"true", "false", "1", "bar", ""}
+		if !cmpLists(expected, stdValues) {
+			t.Fatalf("std flag: expected %v, got %v", expected, stdValues)
+		}
+
+		fset := NewFlagSet("pflag test", ContinueOnError)
+		pflagValues, err := runCase(fset, flagName, args)
+		if err != nil {
+			t.Fatalf("pflag: expected no error, got %v", err)
+		}
+		if !cmpLists(stdValues, pflagValues) {
+			t.Fatalf("pflag: expected %v, got %v", stdValues, pflagValues)
+		}
+	})
+
+	t.Run("error triggered by callback", func(t *testing.T) {
+		flagName := "bflag"
+		args := []string{"--bflag", "--bflag=err", "--bflag=after"}
+
+		// test behavior of standard flag.Fset with an error triggere by the callback:
+		// (note: as can be seen in 'runCase()', if the callback sees "err" as a value
+		//  for the bool flag, it will return an error)
+		stdFSet := flag.NewFlagSet("std test", flag.ContinueOnError)
+		stdFSet.SetOutput(io.Discard) // suppress output
+
+		// run test case with standard flag.Fset
+		stdValues, err := runCase(stdFSet, flagName, args)
+
+		// double check the standard behavior:
+		// - .Parse() should return an error, which contains the error message
+		if err == nil {
+			t.Fatalf("std flag: expected an error triggered by callback, got no error instead")
+		}
+		if !strings.HasSuffix(err.Error(), unitTestErr.Error()) {
+			t.Fatalf("std flag: expected unittest error, got unexpected error value: %T %v", err, err)
+		}
+		// - the function should have been called twice, with the first two values,
+		//   the final "=after" should not be recorded
+		expected := []string{"true", "err"}
+		if !cmpLists(expected, stdValues) {
+			t.Fatalf("std flag: expected %v, got %v", expected, stdValues)
+		}
+
+		// now run the test case on a pflag FlagSet:
+		fset := NewFlagSet("pflag test", ContinueOnError)
+		pflagValues, err := runCase(fset, flagName, args)
+
+		// check that there is a similar error (note: pflag will _wrap_ the error, while the stdlib
+		// currently keeps the original message but creates a flat errors.Error)
+		if !errors.Is(err, unitTestErr) {
+			t.Fatalf("pflag: got unexpected error value: %T %v", err, err)
+		}
+		// the callback should be called the same number of times, with the same values:
+		if !cmpLists(stdValues, pflagValues) {
+			t.Fatalf("pflag: expected %v, got %v", stdValues, pflagValues)
+		}
+	})
+}