Moving down into an empty object or list fails. about waargonaut

[luke-clifton]

Ha, we posted at the same time. 👍

[mankyKitty]

There isn't anything in the index for that position, so it is considered a failure. I don't think I have explicit documentation for this behaviour, I'll add some in.

[luke-clifton]

So, what I want to do is, go down into an object, then do a foldCursor (or a rightwardSnoc). I'm perfectly happy for that to just give me the initial value of the accumulator if the object is empty. What is the alternative? If I check for failure on the down I still don't know why that is, it could be because it's a text or numeric field.

Or, rephrased, how can I check that my cursor is currently pointed at an object without parsing it all into a Json?

[luke-clifton]

Which, looking at the code just now, is exactly what your listAt is trying to do. So, if you listAt and the cursor is sitting in front of a text or numeric field, it will result in an empty list, which seems slight counter-intuitive.

[mankyKitty]

That listAt behaviour seems a bit more 'bug' than 'counter-intuitive', so thanks for that !

Currently there isn't a direct way to check if you're currently looking at an object or other specific type, without going to the Json structure then using the lenses and prisms to nose around. That's a useful feature though and wouldn't be difficult to include.

To keep you moving, you could take advantage of the Alt instance for Decoder to combine alternatives for what may exist at a given cursor position. Until I fix the listAt this will mean you need to be aware of ordering considerations, which isn't awesome.

Alternatively, you are able to use the functions from the library that powers the cursor itself hw-json. Specifically from the module HaskellWorks.Data.Json.Type.

jsonTypeAt (unJCurs cursor) :: Maybe JsonType

Using Alt would be my first preference though. Does any of that help?

[luke-clifton]

It seems that exposing jsonTypeAt is what I want.

[mankyKitty]

I'll leave this open until I've patched listAt others if I find more

[luke-clifton]

@mankyKitty I've posted a patch.

[luke-clifton]

Fixed by #39