Add handling of duplicate keys #714
Conversation
- This allows explicit handling of duplicate keys. - Also remove an outdated comment about the difference between the parsers value and json (they have been the same for a while).
|
Woohoo, a compiler bug on 7.6.3 - 7.10.3 https://travis-ci.org/bos/aeson/jobs/549001593 |
…f simplifier ticks
|
Cool, thanks! |
| objectValues str val = do | ||
| -- The object parser 'objectValues' is parameterized by the object constructor. | ||
| -- See also 'jsonWith'. | ||
| type HFromList = [(Text, Value)] -> Parser Object |
There was a problem hiding this comment.
I'd have Either String Object as a return type. Parser is "too big".
And actually wouldn't use a type synonym, hFromList is confusing naming, mkObject is what it is
EDIT: I see HFromList is because of HashMap.fromList, too far. mkObject is a domain specific, thus better.
There was a problem hiding this comment.
I see. I forgot we're in attoparsec's Parser, not aeson's Parser (which is closer to Either). That's a fair point about the synonym, I'll remove it. Thanks for the review!
| fmap (Array . Vector.fromList . ($ [])) . H.fromListWith (.) . (fmap . fmap) (:) | ||
|
|
||
| -- | @'fromListNoDup' kvs@ fails if @kvs@ contains duplicate keys. | ||
| parseListNoDup :: [(Text, Value)] -> Parser Object |
There was a problem hiding this comment.
point in case: here the HFromList is not used, so connecting the dots is difficult.
I'd remove the type alias.
Fixes #531.
This adds a parameterized parser
jsonWithto let users decide what to do with objects with duplicate keys. A few common variants are also included.Benchmarks to guard against regression (using the aeson-parse benchmark) (no difference on GHC 8.6, 8.4, there seems to be a 2% slowdown on 8.2.2) https://gist.github.com/Lysxia/beb1aa201c79f576b94c509af7f0abfa