add support for strict errors #796
Conversation
There was a problem hiding this comment.
I'm still not convinced (more context in thread kubernetes-sigs/yaml#61) that we should allow yaml.v3 to be non-spec compliant and allow duplicate fields at any time, although yaml.v2 allows duplicate fields by default.
JSON just "recommends" unique keys, for which it then makes sense that unique keys is a sense of strictness, but YAML mandates their uniqueness, hence I do not think of that as a strictness measure.
I think the change to separate strict and non-strict errors is good though.
decode.go
Outdated
| doc *Node | ||
| aliases map[*Node]bool | ||
| terrors []string | ||
| strictTerrors []string |
There was a problem hiding this comment.
this should probably be strictErrors? I guess t stands for type, but it is still a bit weird when reading
Signed-off-by: Inteon <42113979+inteon@users.noreply.github.com>
|
@luxas as requested, I now only consider "unknown field" errors as strict errors. |
| @@ -129,8 +129,8 @@ func (dec *Decoder) Decode(v interface{}) (err error) { | |||
| out = out.Elem() | |||
| } | |||
| d.unmarshal(node, out) | |||
| if len(d.terrors) > 0 { | |||
| return &TypeError{d.terrors} | |||
| if len(d.terrors) > 0 || len(d.strictErrors) > 0 { | |||
There was a problem hiding this comment.
not sure if this is now common enough to have its own helper func attached to the decoder, something like if err := d.checkErr(); err != nil { return err } ... it would add some overhead due to one extra fn call so not sure if it's worth it.
|
I'm closing this PR, since I'm currently working on other things. |
based on #752
Now, "unknown field" errors are considered strict errors.
This means that unmarshalling will still return a valid response, but a strict error is returned too.
The library user can now decide eg. to use the result and also display a warning.
ref: kubernetes-sigs/yaml#61