Add ErrorContains

[alxn]

Summary

It's common practice to wrap errors, so that you do not have a transitive dependency on the error format of a third party library, but also so that you are able to exactly match you are testing a specific error case correctly.

This change adds an ErrorContains() method to check that both an error is reurned, and it contains a provided string.

Changes

Added ErrorContains() and ErrorContainsf().

Motivation

When I wrap errors returned from third party modules:

func method(f foo, b bar) error {
        if err := f(); err != nil {
                return fmt.Errorf("foo failed: %w", err)
        }
        if err := b(); err != nil {
                return fmt.Errorf("bar failed: %w", err)
        }
        return nil
}

This then forces me to write:

f := func() error { return assert.AnError }
b := func() error { return nil }
err := method(f, b)
require.Error(t, err)
assert.Contains(t, err.Error(), "foo failed")

f = func() error { return nil }
b = func() error { return assert.AnError }
err = method(f, b)
require.Error(t, err)
assert.Contains(t, err.Error(), "bar failed")

This change removes the require and Error() unpacking:

f := func() error { return assert.AnError }
b := func() error { return nil }
err := method(f, b)
assert.ErrorContains(t, err, "foo failed")

f = func() error { return nil }
b = func() error { return assert.AnError }
err = method(f, b)
assert.ErrorContains(t, err, "bar failed")

Related issues

[boyan-soubachov]

I'd love to get community feedback on this. I'm not sure the value this extra function adds is greater than the confusion/complexity it creates by having yet another function in the assert/require packages.

In fact, how is this different from the EqualError assertions?

[alxn]

EqualError is different because it requires an exact error match. Here, I have the problem where I have a decorated error. I don't know the underlying error, only the wrapping.

[boyan-soubachov]

I see. Would it not make sense to rename this function to ContainsError to be consistent with EqualError then?

[alxn]

assert.ErrorContains(t, err, "bar failed")

Semantically I'm asserting that the Error err Contains "bar failed".

ContainsError scans more like:

assert.ContainsError(t, something, err)

I'd written ErrorContains to be a variant of ErrorIs:

func EqualError(t TestingT, theError error, errString string, msgAndArgs ...interface{}) bool
func Error(t TestingT, err error, msgAndArgs ...interface{}) bool
func ErrorAs(t TestingT, err error, target interface{}, msgAndArgs ...interface{}) bool
func ErrorIs(t TestingT, err, target error, msgAndArgs ...interface{}) bool
func NoError(t TestingT, err error, msgAndArgs ...interface{}) bool
func NotErrorIs(t TestingT, err, target error, msgAndArgs ...interface{}) bool
func PanicsWithError(t TestingT, errString string, f PanicTestFunc, msgAndArgs ...interface{}) bool

[boyan-soubachov]

Fair point, that makes sense :)

[yarikk]

This was merged 6 months ago but still unavailable in @latest which currently resolves to v1.7.0. When to expect it version-tagged?

[webmaster128]

1.7.1 is realeased and seems to include this feature. Thank you everyone.