aggregating GBIF occurrence data by scientific name

gistfile1.clj

;; see https://github.com/MapofLife/gbifer

;; use with https://s3.amazonaws.com/gbifsource/occ.txt

(use 'gulo.gbif)
(in-ns 'gulo.gbif)


(defn read-occurrences
  ([]
     (read-occurrences "/mnt/hgfs/Dropbox/code/github/MapofLife/gbifer/resources/gbif/occ.txt")) 
  ([path]
     (let [src (hfs-textline path)]
       (<- [?scientificname ?occurrenceid ?latitude ?longitude]
           (src ?line)
           (split-line ?line :>> occ-fields)))))

(defbufferop wide-format
  "Aggregate occurrence id and lat/lon into wide format, from long"
  [tuples]
  (let [occ-ids (map first tuples)
        lats (map #(nth % 1) tuples)
        lons (map #(nth % 2) tuples)]
    [[occ-ids [lats lons]]]))

(let [occ-src (read-occurrences)]
  (??<- [?sci-name ?occ-ids ?latlons]
      (occ-src ?sci-name ?occ-id ?lat-str ?lon-str)
      (valid-name? ?sci-name)
      (latlon-valid? ?lat-str ?lon-str)
      (wide-format ?occ-id ?lat-str ?lon-str :> ?occ-ids ?latlons)))

;; results in this:

(["Acidobacteria" ("242135095" "244666043" "244664083" "244662123" "244661391" "244662763" "242136127" "242135147" "242135541" "244661001") 
                  [("9.93166666667" "47.166668" "-40.8747" "15.509722" "73.967" "-43.288" "31.41666667" "41.526565" "41.576233" "53.49027") 
                   ("0.896666666667" "19.583334" "170.851" "73.88306" "-140.088" "-175.5532" "35.41666667" "-70.670762" "-70.6336" "6.144433")]])