Use encName in rgetResolveSFI to allow tag aliasing of shadowed anonymous fields

[johnniedoe]

I think that this would resolve the first use case mentioned in #159

The second use case seems a bit more involved.

[ugorji]

It's not clear to me what this change does. It seems like it just changes the name of the field, which seems to be a change that does nothing.

[johnniedoe]

The change causes the list of struct fields to be keyed off of the encName rather than the fieldName. As a result, a shadowed struct field can still be encoded so long as the tag-specified alias is not shadowed. In my tests, this fixes the first variable shadowing issue mentioned in #159.

If you'd like, I can add a test to the unit tests demonstrating it.

[ugorji]

But if I look at https://github.com/ugorji/go/pull/160/commits/b53f7c41dc5efa3112caedaecc67e2b2f72d4624 , it just looks like you changed the word "fieldName" to "encName".

I can't see how that change actually does anything. It seems more like a refactoring.

Please take a look at the actual change again.

[johnniedoe]

The rename of "fieldName" to "encName" within the sfiIdx struct was made for consistency/clarity, but shouldn't have any other runtime effects.

The change that really makes a difference is on 1011. The function is forcing uniqueness of fields by comparing them against a list of already "found" ones. Before the change, it was using the "fieldName" field of "structFieldInfo" as the key for comparison. Post-change, it is using the "encName" field. These are two different fields in a pre-existing struct.

To see the difference that it makes, try running this sample program against both versions:

import (
    "bytes"
    "encoding/json"
    "fmt"

    "github.com/ugorji/go/codec"
)

type Composite struct {
    X string
    Embedded
}

type Embedded struct {
    X string `json:"x"`
}

func main() {
    d := Composite{X:"a",Embedded:Embedded{X:"b"}}
    msg, _ := json.Marshal(d)
    fmt.Println(" json: ", string(msg))

    buf := &bytes.Buffer{}
    codec.NewEncoder(buf, &codec.JsonHandle{}).Encode(d)
    fmt.Println("codec: ", buf.String())
}

The existing code should produce:

 json:  {"X":"a","x":"b"}
codec:  {"X":"a"}

while the updated code should produce:

 json:  {"X":"a","x":"b"}
codec:  {"X":"a","x":"b"}

I did verify that the updated version passes the existing unit tests. However, there may be case(s) that I'm not thinking of that aren't in the unit tests...

[ugorji]

Looks good. I have validated it.

I can't merge it, because there are a few other merges that don't apply.

If you can make this a standalone merge, I will merge it. Else, I will just include the fix in my next updates (but you lose credit for the change). Let me know your preference.

[ugorji]

@johnniedoe ^^ (please read and respond to prior message)

[ugorji]

Updated full fix for #159 in change c0d481cf0ab1f0d68f8257fd3ffa267506dfc04b . This pull request will no longer work.

[johnniedoe]

Sorry for the slow reply. I'd been away on vacation.

Yours looks like the more comprehensive solution. Mine only handled one of the issues mentioned in #159, while your appears to handle both. Nice.

Thanks again for the fix.