encoding/xml: Unmarshal panic when use generic

[nzlov]

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

$ go version
go version devel go1.18-c69f5c0d76 Fri Sep 10 10:45:59 2021 +0000 darwin/amd64

Does this issue reproduce with the latest release?

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

go env Output

$ go env
GO111MODULE="on"
GOARCH="amd64"
GOBIN=""
GOCACHE="/Users/nzlov/Library/Caches/go-build"
GOENV="/Users/nzlov/Library/Application Support/go/env"
GOEXE=""
GOEXPERIMENT=""
GOFLAGS=""
GOHOSTARCH="amd64"
GOHOSTOS="darwin"
GOINSECURE=""
GOMODCACHE="/Users/nzlov/workspace/go/pkg/mod"
GONOPROXY=""
GONOSUMDB=""
GOOS="darwin"
GOPATH="/Users/nzlov/workspace/go"
GOPRIVATE=""
GOPROXY="https://goproxy.cn,direct"
GOROOT="/Users/nzlov/program/go"
GOSUMDB="sum.golang.org"
GOTMPDIR=""
GOTOOLDIR="/Users/nzlov/program/go/pkg/tool/darwin_amd64"
GOVCS=""
GOVERSION="devel go1.18-c69f5c0d76 Fri Sep 10 10:45:59 2021 +0000"
GCCGO="gccgo"
AR="ar"
CC="clang"
CXX="clang++"
CGO_ENABLED="1"
GOMOD="/Users/nzlov/workspace/vgo/generices/go.mod"
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -arch x86_64 -m64 -pthread -fno-caret-diagnostics -Qunused-arguments -fmessage-length=0 -fdebug-prefix-map=/var/folders/71/4t6lk06x5y77f2wyln9vqnlw0000gn/T/go-build3474694710=/tmp/go-build -gno-record-gcc-switches -fno-common"

What did you do?

package main

import (
	"encoding/json"
	"encoding/xml"
	"fmt"
)

type A[T any] struct {
	Name          string  `json:"name" xml:"name"`
	Data          T       `json:"data" xml:"data"`
	ErrorResponse ErrResp `json:"errorResponse" xml:"errorResponse"`
}

type ErrResp struct {
	Age int `xml:"age"`
}

func main() {
	fmt.Printf("%+v\n", *b[int]())
}

func b[T any]() *T {
	// data :=`<A[int]><name>name</name><data>1</data><errorResponse><age>10</age></errorResponse></A[int]>`
	data, err := xml.Marshal(&A[int]{Name: "name", Data: 1, ErrorResponse: ErrResp{Age: 10}})
	if err != nil {
		panic(err)
	}
	fmt.Printf("%+v\n", string(data))
	a1 := A[T]{}
	err = xml.Unmarshal([]byte(data), &a1)
	if err != nil {
		panic(err)
	}
	fmt.Printf("%+v\n", a1)
	return &a1.Data
}

What did you expect to see?

1

What did you see instead?

<A[int]><name>name</name><data>1</data><errorResponse><age>10</age></errorResponse></A[int]>
panic: XML syntax error on line 1: expected attribute name in element

goroutine 1 [running]:
main.b[.shape.int](0x10f3f80)
	/Users/nzlov/workspace/vgo/generices/main.go:45 +0x1dd
main.main()
	/Users/nzlov/workspace/vgo/generices/main.go:21 +0x25
exit status 2

[thanm]

@randall77 @danscales

[randall77]

This patch "fixes" the problem, but makes the parser out-of-spec, so I'm not sure we can do it.

diff --git a/src/encoding/xml/xml.go b/src/encoding/xml/xml.go
index c14954df15..9a9acbab9b 100644
--- a/src/encoding/xml/xml.go
+++ b/src/encoding/xml/xml.go
@@ -1222,7 +1222,7 @@ func isNameByte(c byte) bool {
        return 'A' <= c && c <= 'Z' ||
                'a' <= c && c <= 'z' ||
                '0' <= c && c <= '9' ||
-               c == '_' || c == ':' || c == '.' || c == '-'
+               c == '_' || c == ':' || c == '.' || c == '-' || c == '[' || c == ']'
 }
 
 func isName(s []byte) bool {
@@ -1477,6 +1477,8 @@ var second = &unicode.RangeTable{
        R16: []unicode.Range16{
                {0x002D, 0x002E, 1},
                {0x0030, 0x0039, 1},
+               {0x005b, 0x005c, 1}, // [
+               {0x005d, 0x005e, 1}, // ]
                {0x00B7, 0x00B7, 1},
                {0x02D0, 0x02D1, 1},
                {0x0300, 0x0345, 1},

We may just need to accept that generic types cannot be encoded into XML. Or we need a different naming scheme for generics in XML that uses approved punctuation.

In any case, there's something to do here. The marshaler should return an error if the unmarshaler will return an error.

@rsc from owners

[rsc]

https://play.golang.org/p/OoUqANfbaSl shows what happens for an unnamed struct type: we get an unsupported type error.
This should only affect the top-level element: the inner elements are named for the fields they come from.
We could say that named generic structs are disallowed, but I think it is also fine and probably better to chop the name at the first [.

[rsc]

The docs say:

    The name for the XML elements is taken from, in order of preference:

        - the tag on the XMLName field, if the data is a struct
        - the value of the XMLName field of type Name
        - the tag of the struct field used to obtain the data
        - the name of the struct field used to obtain the data
        - the name of the marshaled type

So there are four different preferred choices before we get to this one - plenty of ways to override the decision to chop at the [.

[gopherbot]

Change https://golang.org/cl/349349 mentions this issue: encoding/xml: truncate generic type names