Convert PDF to text with pdf.js

pdf-to-text.cljs

(ns pdf-to-txt
  (:require
   ["pdfjs-dist" :as pdfjs]))

(assoc!
 pdfjs.GlobalWorkerOptions
 :workerSrc "https://cdn.jsdelivr.net/npm/[email protected]/build/pdf.worker.mjs")

(defn cleanup-text [txt]
  (.replaceAll txt "\"" "'"))

(defn ^:async page->txt [pdf page-num]
  (let* [page (js-await (pdf.getPage page-num))
         text-content (js-await (.getTextContent page))
         items (.-items text-content)]

    (.map items #(-> % .-str (str " ")))))

(defn ^:async pdf->txt [pdf]
  (console.log "pdf loaded")
  (let* [pages (range 1 pdf._pdfInfo.numPages)
         promises (doall (map #(page->txt pdf %) pages))]

    (-> (js-await (Promise.all promises))
        (.join "\n")
        cleanup-text)))

(defn handle [{:keys [reader file buffer finish]}]
  (let [task (pdfjs.getDocument {:data reader.result})]
    (->
     task.promise
     (.then pdf->txt)
     (.then finish))))

(defn handle-onload [finish reader file]
  (let [fun (if finish
              finish
              (fn [x] nil))]
    #(pdf/handle {:reader reader :file file :buffer % :finish fun})))

(defn file-changed-listener [finish]
  (fn [ev]
    (let [file (get ev.target.files 0)
          reader (new js/FileReader)]
      (assoc! reader :onloadend
              (handle-onload finish reader file))

      (.readAsBinaryString reader file)))))

(defn attach-pdf-file-handler [{:keys [id finish]}]
  (console.log "attached at element" id)
  (-> (js/document.getElementById id)
      (.addEventListener "change" (file-changed-listener finish))))

;; make CvFileHandler globally available
(set! js/window.PdfHandler attach-pdf-file-handler)