[Haskell-cafe] Fast JSON validation - reducing allocations

[Bardur Arantsson]

On 2017-05-11 18:12, David Turner wrote:
> Dear Café,
> 
> We have a stream of ~900byte JSON documents (lazy ByteStrings) that we
> would like to validate (NB not parse) as quickly as possible. We just
> need to check that the bytes are a syntactically valid JSON document and
> then pass them elsewhere; we do not care what the content is. The
> documents are encoded in UTF-8 and have no extraneous whitespace.
> 

No particular recommendations, but you might want to look into
semi-indexing[1] as a strategy. It looks plausible that it would be
possible to do that without a lot of allocation; see the paper for
details. (I there's also a demo implementation in Python on GH.)

[1] http://www.di.unipi.it/~ottavian/files/semi_index_cikm.pdf