feat: plain map and slice #464
Conversation
|
@nrwiersma As promised, quick lunch break fix. 😉 This also handles map in a similar way (with a separate option for it of course). |
| @@ -25,6 +25,8 @@ type config struct { | |||
| FullName bool | |||
| Encoders bool | |||
| FullSchema bool | |||
| PlainMap bool | |||
There was a problem hiding this comment.
The decoder will not work for map like it does for slice currently.
There was a problem hiding this comment.
Thanks for your feedback, however I'm not sure what you are referring to here. What would need to change in the decoder for this to work?
There was a problem hiding this comment.
@nrwiersma I'm not familiar with the whole codec logic yet but I had a look and tried to add support for nullable maps for encoding and decoding. Have a look at my latest commits and let me know what you think.
0xjac
force-pushed
the
feat/461-plain-map-slice
branch
from
ef47dfc to
4d8a8e2
Compare
Go maps and slices can be nil. An avro nullable union can thus be represented directly by a map or a slice without using an explicit pointer (`*`). This feature is hidden being the `PlainMap` and `PlainSlice` options to preserve the existing behavior. Closes hamba#461
Adds `-plain-map` and `-plain-slice` flags to expose the corresponding options.
Similar to how nullable slices are handles, this adds support for using plain `map[string]any` for nullable unions of maps instead of using an explicit pointer to a map.
0xjac
force-pushed
the
feat/461-plain-map-slice
branch
from
4d8a8e2 to
a733118
Compare
| if typ.(reflect2.MapType).Elem().Kind() != reflect.Interface { | ||
| if !schema.Nullable() { | ||
| break | ||
| } |
There was a problem hiding this comment.
There is an edge case here being:
["null", {"type": "map", "values": ["null", "string"]}]
or something like this. In the traditional struct layout this could be *map[string]any which is specifically excluded here. In the new layout being proposed, this would be map[string]any, and the en/decoder would not be able to tell the cases apart.
To deal with this case, IMO the en/decoder should always error towards the map union (existing functionality) and failing that, check the new non-pointer version. This would also need to be documented.
There was a problem hiding this comment.
@nrwiersma sorry my use case has changed and I no longer need this feature. Sadly I don't have time to fix it at the moment. Feel free to fix it, hand it over to another volunteer or close it.
Go map and slice types are already nullable and do not need an extra pointer to represent a nullable union.
This new behavior is hidden behind the
PainMap,PlainSliceoptions and corresponding-plain-map,-plain-sliceflags for the CLI to preserve compatibility.