Better way to modify existing extractor.

[racinmat]

Currently we have e.g. in mutagenesis example https://github.com/CTUAvastLab/JsonGrinder.jl/blob/master/examples/mutagenesis.jl
following, cumbersome way of modifying current extractor after suggestion:

extractor[:atoms].item[:bonds].item.dict[:charge] = JsonGrinder.ExtractCategorical(sch[:atoms].items[:bonds].items[:charge])

which is lengthy and we need to specify the path twice.
Instead it would be nice to e.g. use lenses, but that would require us to have unique code2lens, which is not now, because the logic is: takt value under that code and match/return everything that is equal as element under that code.
Which is troublesome.

printtree(sch, trav=true)
printtree(extractor, trav=true)
sch["BK"]
code2lens(sch, "BK")
sch_lens = only(code2lens(sch, "BK"))
code2lens(extractor, "BK")
ext_lens = first(code2lens(extractor, "BK"))

using Setfield
get(sch, sch_lens)
get(extractor, ext_lens)

also the schema lens to extractor lens is not working now, I would expect this to work:

Setfield.set(extractor, ext_lens, JsonGrinder.ExtractCategorical(get(sch, sch_lens)))

But maybe it may suffice to have yet another parameter to suggestextractor, which woud be similar as fsm field in reflectinmodel, where one would pass dict of path in schema => suggestion of extractor baed on shema.
But the path used by string code is unstable and code changes based on another fields. Maybe putting lens there would be worth it?

[simonmandlik]

IMO, adding this to suggestextractor makes things too complicated, when this is a simple one liner:

sch = schema(...)
e = suggestextractor(sch)

using Accessors

set(e, code2lens(e, "BK"), ...)

But some examples may be added to the docs: #137