x/text/language: language.TraditionalChinese.Parent() != language.Chinese

[nicksnyder]

What version of Go are you using (go version)?

go version go1.10.2 darwin/amd64

Does this issue reproduce with the latest release?

Yes

What operating system and processor architecture are you using (go env)?

GOARCH="amd64"
GOBIN=""
GOCACHE="/Users/nick/Library/Caches/go-build"
GOEXE=""
GOHOSTARCH="amd64"
GOHOSTOS="darwin"
GOOS="darwin"
GOPATH="/Users/nick/dev/gopath"
GORACE=""
GOROOT="/usr/local/Cellar/go/1.10.2/libexec"
GOTMPDIR=""
GOTOOLDIR="/usr/local/Cellar/go/1.10.2/libexec/pkg/tool/darwin_amd64"
GCCGO="gccgo"
CC="clang"
CXX="clang++"
CGO_ENABLED="1"
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -m64 -pthread -fno-caret-diagnostics -Qunused-arguments -fmessage-length=0 -fdebug-prefix-map=/var/folders/l6/djg_zw1j5lgbwz4h5q92lnwr0000gn/T/go-build319403565=/tmp/go-build -gno-record-gcc-switches -fno-common"

What did you do?

package main

import "fmt"
import "golang.org/x/text/language"

func main() {
	hans := language.SimplifiedChinese.Parent() == language.Chinese
	hant := language.TraditionalChinese.Parent() == language.Chinese
	fmt.Println(hans, hant) // true, false
}

What did you expect to see?

I expected language.TraditionalChinese.Parent() == language.Chinese

What did you see instead?

language.TraditionalChinese.Parent() == language.Und

[meirf]

The code that's causing the divergent behavior has the following explanation:

// The parent for an base-script pair with a non-default script is
// "und" instead of the base language."

[nicksnyder]

I guess my real question is “why?”

[meirf]

cc @mpvl

[ALTree]

That comment seems to rule out the possibility that the behaviour you're observing is a bug. In general we don't use the issue tracker for questions, it's just for actual bugs.

I guess my real question is “why?”

Are you just curious, or the observed (and documented) behaviour is causing you troubles?

[nicksnyder]

Ultimately I am questioning the correctness/usefulness of this behavior.

A simple reading of the godoc doesn’t give any guidance to this behavior so I think the expected behavior described above is reasonable and the actual behavior is surprising.

As a user of this API, this caused an unexpected bug (which I was able to workaround): nicksnyder/go-i18n#107

If this is indeed working as intended then I think better documentation would satisfy me. I am also personally curious to understand the rational for this particular behavior.

[ALTree]

Okay, leaving this open then. Package owner @mpvl is likely the one that can answer your questions and make a decision about this.

[gopherbot]

Change https://golang.org/cl/145437 mentions this issue: language: improve docs for Parent

[mpvl]

This behavior is defined in http://unicode.org/reports/tr35/#Parent_Locales

The parent relationship is used to determine fallback data, or fields that can be inherited from "parent" languages. Generally this corresponds to stripping away the last part of a BCP 47 tag, but not always. In general, single-component tags typically apply some default script. For instance, Dutch is typically written in Latin script, so no use to write nl-Latn. However, some languages are commonly written in more than one script, and it would be bad for a user to fall back on a language for which the script is not understood. In the case of zh and zh_Hant it is even worse, as they do not only imply a different script, but also, with high likelihood, a different language (cmn_Hans and yue_Hant, respectively). I guess zh is still widely used as there is still a decent amount of mutual intelligibility between the written form of Chinese languages if written in the same script. But one should not assume mutual intelligibility between different scripts. Hence.

[nicksnyder]

Thanks for the explanation! The proposed documentation looks good to me.