What's the deal with trying the Equal() method about go-cmp

[dsnet]

The current paragraph that describes usage of the Equal method currently says:

• If the values have an Equal method of the form "(T) Equal(T) bool" or "(T) Equal(I) bool" where T is assignable to I, then use the result of x.Equal(y). Otherwise, no such method exists and evaluation proceeds to the next rule.

If it helps, I can add a phrase to suggest this edge case:

• If the values have an Equal method of the form "(T) Equal(T) bool" or "(T) Equal(I) bool" where T is assignable to I, then use the result of x.Equal(y) even if x or y are nil pointers. Otherwise, no such method exists and evaluation proceeds to the next rule.

Would that help?

[dsnet]

Go has no operator overloading for ==, so the Equal method has become the typical way that many objects declare how to compare themselves. This pattern occurs in the standard library, and an analysis of other Go code indicated that this was a common enough pattern to special case:

crypto/x509/x509.go:752:func (c *Certificate) Equal(other *Certificate) bool {
regexp/syntax/regexp.go:62:func (x *Regexp) Equal(y *Regexp) bool {
encoding/asn1/asn1.go:224:func (oi ObjectIdentifier) Equal(other ObjectIdentifier) bool {
time/time.go:252:func (t Time) Equal(u Time) bool {
net/ip.go:375:func (ip IP) Equal(x IP) bool {

[filmil]

I understand.

The reason I reported this was because of kubernetes/apimachinery@44878ce#diff-221c91463b9db024401e40d997ec741c

Turns out, I was looking at an old version of K8S source that had a problem here.

It would be nice, however, if go-cmp had some place that documented what requirements there are on Equal().

[filmil]

Yes, please.

…

On Thu, Dec 21, 2017 at 12:50 PM Joe Tsai ***@***.***> wrote: The current paragraph that describes usage of the Equal method currently says: - If the values have an Equal method of the form "(T) Equal(T) bool" or "(T) Equal(I) bool" where T is assignable to I, then use the result of x.Equal(y). Otherwise, no such method exists and evaluation proceeds to the next rule. If it helps, I can add a phrase to suggest this edge case: - If the values have an Equal method of the form "(T) Equal(T) bool" or "(T) Equal(I) bool" where T is assignable to I, then use the result of x.Equal(y) *even if x or y are nil pointers*. Otherwise, no such method exists and evaluation proceeds to the next rule. Would that help? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#61 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAPDMA2zKzIWlxxtCgVp-kSMWn3oSfpAks5tCsStgaJpZM4RKSkW> .

[filmil]

... especially because it is all too easy to depend on code that makes this mistake, and then you are fishing for reasons why it crashes.

…

On Thu, Dec 21, 2017 at 12:56 PM Filip Filmar ***@***.***> wrote: Yes, please. On Thu, Dec 21, 2017 at 12:50 PM Joe Tsai ***@***.***> wrote: > The current paragraph that describes usage of the Equal method currently > says: > > > - If the values have an Equal method of the form "(T) Equal(T) bool" > or "(T) Equal(I) bool" where T is assignable to I, then use the result of > x.Equal(y). Otherwise, no such method exists and evaluation proceeds to the > next rule. > > If it helps, I can add a phrase to suggest this edge case: > > > - If the values have an Equal method of the form "(T) Equal(T) bool" > or "(T) Equal(I) bool" where T is assignable to I, then use the result of > x.Equal(y) *even if x or y are nil pointers*. Otherwise, no such > method exists and evaluation proceeds to the next rule. > > Would that help? > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#61 (comment)>, or mute > the thread > <https://github.com/notifications/unsubscribe-auth/AAPDMA2zKzIWlxxtCgVp-kSMWn3oSfpAks5tCsStgaJpZM4RKSkW> > . >

[dsnet]

#63

[dsnet]

For future reference: I don't think we would ever change the semantics of when the Equal method is called since it useful for certain types to declare that the nil pointer to equal to the zero value of an allocated struct.

However, if this continues to bite people, we could consider recovering from a nil panic in Equal and performing the equivalent of if (x == nil || x == nil) { return x == nil && y == nil } on their behalf. However, this will slow down cmp.Equal further and is a hack I do not want to perform unless there is evidence of this being a pervasive problem.

[empijei]

Hi, sorry to necrobump, but I'd rather not open a duplicate issue.

Just wanted to point out that this problem bit someone today: golang/go#28743

This is probably due to the fact that the (T).Equal(T) method is not really defined, declared or guaranteed to behave like == but people expect it to, and this package does too.

[m1kola]

I had this issue with x509.Certificate recently. I guess a workaround while we are waiting for golang/go#28743 is to use a custom cmp.Comparer. Something like this:

certCmp := cmp.Comparer(func(x, y *x509.Certificate) bool {
	if x == nil || y == nil {
		return x == y
	}

	return x.Equal(y)
})

cmp.Diff(something, other, certCmp) // or cmp.Equal