Introduction of JSON tools for Scheme

I'm writing a library which can query JSON. It's still under development state but a bit of sample code wouldn't hurt so let me introduce it.

The repository is here, json-tools and handling JSON structure must be one of the Chicken Scheme egg's library json format. (I've ported it for R6RS Scheme implementations, it's in the repository as well.)

The library consists 2 parts one is JSON tools and the other one is JSON query which is based on the JSONSelect.

JSON Tools

This part of the library is highly inspired by SXPath. There are bunch of basic selectors which can be used by querying libraries. Following piece of code describe the flavour of this part.
(import (rnrs) (text json tools))

(define json '#(("name" . #(("first" . "Lloyd") ("last" . "Hilaiel")))
                ("favoriteColor" . "yellow")
                 #(("lang" . "Bulgarian") ("level" . "advanced"))
                 #(("lang" . "English")
                   ("level" . "native")
                   ("preferred" . #t))
                 #(("lang" . "Spanish") ("level" . "beginner")))
                ("seatingPreference" "window" "aisle")
                ("drinkPreference" "whiskey" "beer" "wine")
                ("weight" . 156)))

((json:filter (lambda (node)
  ;; The given node can be other type and
  ;; this piece of code may raise an error
  ;; but for above structure this works :)
  (equal? "name" (json:map-entry-key node))))
 (json:child-nodes json))
;; -> <json:nodeset>
All JSON selectors return JSON nodeset type which contains sets of JSON node. Followings are the JSON node types;
  • JSON map
  • JSON map entry
  • JSON array
  • JSON string
  • JSON number
  • JSON boolean
  • JSON null
All types are sub type of JSON node. The reason why I introduced them is that there was no way to tell the difference between array and map entry which contains array. To avoid ambiguous results, I needed to do it.

To retrieve the result set as S-expression, you can simply call the `json:nodeset->list` procedure like this;
(json:nodeset->list ((json:filter (lambda (node)
                                    (equal? "name" (json:map-entry-key node))))
                     (json:child-nodes json1)))
;; --> (("name" . #(("first" . "Lloyd") ("last" . "Hilaiel"))))
Not sure if the procedure name is proper. (I also though `json:nodeset->sexp`.) To casual use, `json:filter`, `json:child-nodes`, `json:child-nodes-as-list`, `json:child` or `json:child-as-list` are easy to use. The rest of selectors are a bit tricky.

JSON Select

This part of the library is for usability. You can use query language to select  particular nodes from JSON. The example use is like this;
(import (text json select))

;; use the same JSON structure defined above
((json:select ".languagesSpoken") json)
;; --> <json:nodeset>

(json:nodeset->list ((json:select ".languagesSpoken") json))
;; --> '(("languagesSpoken"
;;        #(("lang" . "Bulgarian") ("level" . "advanced"))
;;        #(("lang" . "English")
;;          ("level" . "native")
;;          ("preferred" . #t))
;;        #(("lang" . "Spanish") ("level" . "beginner"))))
The returning value is JSON nodeset the same as tools. So again to retrieve the S-exp result, you need to call the procedure `json:nodeset->list`. Currently, not all query language are implemented but it would be a matter of time.

As I mentioned before, the state is still under development so your opinions or testing results are very welcome :) Of course, your pull requests are much appreciated :)

No comments:

Post a Comment