Hitri popravki in navodila

docker-compose.yml

@@ -11,3 +11,8 @@ services:
             - ./swagger_server/classla/classla_resources:/root/classla_resources
         environment:
             - PYTHONUNBUFFERED=1
+    tika:
+        image: apache/tika
+        restart: always
+        ports:
+            - 9998:9998

requirements.txt

@@ -11,3 +11,4 @@ pandas==1.3.3
 python-docx==0.8.11
 lxml==4.8.0
 PyPDF2==2.10.4
+requests

swagger_server/controllers/doc2text_controller.py

@@ -1,6 +1,7 @@
 import connexion
 import six
 # import tika
+import requests
 
 from swagger_server import util
 import pandas as pd
@@ -20,37 +21,25 @@ def datoteka_v_besedilo_post(file=None):  # noqa: E501
 
     :rtype: str
     """
-
+    #zakaj ? tika server se požene zraven v dodatnem containerju in se datoteka samo pošlje tja, nazaj pa se dobi čisto besedilo...
+    #tika ima rest api
     if file is None:
         return "No file provided", 400
-    elif "openxmlformats-officedocument.wordprocessingml.document" in file.content_type:
-        content = [p.text for p in docx.Document(file).paragraphs]
-    elif "text/plain" in file.content_type:
-        content = file.read().decode('utf-8')
-    elif "application/pdf" in file.content_type:
-        reader = PdfReader(file)
-        content = '\n'.join([p.extract_text() for p in reader.pages])
-        content = "ZACASNO UPORABLJEN DRUGI BRALEC KOT TIKA, TA BO SE DODANA KASNEJE...\n\n" + content
-    elif "text/xml" in file.content_type:
-        root = ET.parse(file).getroot()
-        plainText = root.findall('PlainText')
-        if len(plainText) == 0:
-            return "Didn't find anything in PlainText", 400
-        content = '\n'.join([pt.text for pt in plainText])
-    else:
-        return "Currently supporing only docx, txt/plain, pdf, xml*... more will be added later", 501
 
-    return content, 200
+    tika_server="http://tika:9998/tika"
+    response=requests.post(tika_server, data=file)
+    return response.text, 200
 
 
 def get_text_ocr(file=None):  # noqa: E501
-    """Pretvori datoteko formata pdf, doc, docx, ppt, xls,... v besedilo s pomočjo ocr razpoznavanja
     
-     # noqa: E501
 
-    :param file: 
+    if file is None:
-    :type file: strstr
+        return "No file provided", 400
+
+    tika_server="http://tika:9998/tika"
+    #za ocr se zraven samo doda header 
+    response=requests.post(tika_server, data=file,headers={"X-Tika-PDFOcrStrategy":"ocr_only","X-Tika-OCRLanguage":"slv+eng"})
+    return response.text, 200
+
 
-    :rtype: str
-    """
-    return 'Not yet implemented'

swagger_server/controllers/extract_controller.py

@@ -18,4 +18,54 @@ def get_candidates(body):  # noqa: E501
     """
     if connexion.request.is_json:
         body = IzlusciBody.from_dict(connexion.request.get_json())  # noqa: E501
-    return 'do some magic!3'
+
+
+    data={
+        terminoloski_kandidati:[
+              {
+                "POSoznake": "Ncmsn",
+                "kandidat": "vpliv",
+                "kanonicnaoblika": "vpliv",
+                "nosilnautez": 0.8008282,
+                "podporneutezi": [
+                  6.0274563,
+                  6.0274563
+                ],
+                "pogostostpojavljanja": [101,71]
+              },
+              {
+                "POSoznake": "Agpnsg Ncnsg",
+                "kandidat": "bivalen okolje",
+                "kanonicnaoblika": "bivalno okolje",
+                "nosilnautez": 0.60254,
+                "podporneutezi": [
+                  3.263,
+                  2.134
+                ],
+                "pogostostpojavljanja": [27,11]
+              },
+              {
+                "POSoznake": "Agpmsny Ncmsn Ncmpg",
+                "kandidat": "motorični status otrok",
+                "kanonicnaoblika": "motorični status otrok",
+                "nosilnautez": 0.3324,
+                "podporneutezi": [
+                  1.221,
+                  3.323
+                ],
+                "pogostostpojavljanja": [31,51]
+              },
+              {
+                "POSoznake": "Agpnsn Ncnsn",
+                "kandidat": "diplomsko delo",
+                "kanonicnaoblika": "diplomsko delo",
+                "nosilnautez": 0.2008282,
+                "podporneutezi": [
+                  0.883,
+                  1.02
+                ],
+                "pogostostpojavljanja": [1241,111]
+              }
+            ]
+        }
+    return data