Integer casts in posflag.FlagVal result in unexpected error #362
Description
Describe the bug
The function posflag.FlagVal uses a type switch to get a flag's value in the correct type, but then casts each signed integer to int64 and each unsigned integer to uint64.
With the following configuration, this behavior will result in an unexpected error:
- Use
koanf.Conf.StrictMerge: trueto prevent the case where e.g. abooloverwrites anint. - Load the initial config via another provider, e.g.
confmapthat sets anintvalue. - Use the posflag provider to map an
intflag to theintconfig value viaposflag.ProviderWithFlag. Useposflag.FlagValto get the value.
Even though the initial config uses int and the flag is defined as an int, loading the values via the posflag provider results in the following error when koanf tries to merge both values:
incorrect types at key , type int != int64
To Reproduce
package main
import (
"fmt"
"github.com/knadh/koanf/providers/confmap"
"github.com/knadh/koanf/providers/posflag"
"github.com/knadh/koanf/v2"
"github.com/spf13/pflag"
)
func main() {
k := koanf.NewWithConf(koanf.Conf{
Delim: ".",
// Setting it to false won't trigger the bug, but it would also allow
// overwriting a bool with an int, which I want to prevent.
StrictMerge: true,
})
// Making "1" an int64 also "fixes" the issue.
confmapProvider := confmap.Provider(map[string]any{"a": int(1)}, "")
if err := k.Load(confmapProvider, nil); err != nil {
panic(err)
}
// Output: 1 (int)
fmt.Printf("%v (%T)\n", k.Get("a"), k.Get("a"))
fs := pflag.NewFlagSet("", pflag.PanicOnError)
fs.Int("a-flag", 0, "")
if err := fs.Set("a-flag", "2"); err != nil {
panic(err)
}
flagNameToConfig := map[string]string{"a-flag": "a"}
posflagProvider := posflag.ProviderWithFlag(fs, ".", nil, func(f *pflag.Flag) (string, any) {
posflagVal := posflag.FlagVal(fs, f)
// Output: 2 (int64)
fmt.Printf("%v (%T)\n", posflagVal, posflagVal)
return flagNameToConfig[f.Name], posflagVal
})
if err := k.Load(posflagProvider, nil); err != nil {
// Results in error: incorrect types at key , type int != int64
panic(err)
}
}Expected behavior
I expected that posflag.FlagVal returns the value with the same type that the flag is defined with and therefore for no error to occur.
Making posflag.FlagVal assign i directly to val instead of casting it to int64/uint64 and then assigning it to val fixes the issue; but I don't know if I'm missing the reason why the casts are there.
Please provide the following information):
- OS: linux
- Koanf Version v2.2.0