Web application.

Model downloaded from the Internet if not existing in the local disk.

Author: francisco-ortin

common/languages.py

@@ -0,0 +1,3 @@
+LANGUAGES = ["Assembly", "C", "C++", "C#", "CSS", "Go", "HTML", "Java", "JavaScript", "Kotlin",
+                   "Matlab", "Perl", "PHP", "Python", "R", "Ruby", "Scala", "SQL", "Swift", "TypeScript",
+                   "Unix Shell"]

common/model.py

@@ -1,9 +1,39 @@
+from io import BytesIO
 from typing import Dict
+import requests
+from zipfile import ZipFile
+
 import numpy as np
 import tensorflow as tf
+from keras import Model
+
+MODEL_PATH = '../common/model/BRNN'
+MODEL_URL = 'https://reflection.uniovi.es/bigcode/download/2024/plangrec/BRNN.zip'
+
+
+def download_and_load_model(model_path: str, model_url: str) -> Model:
+    import os
+    import sys
+    if not os.path.exists(model_path):
+        print(f"Model not found in '{model_path}'.")
+        print(f"Downloading model from '{model_url}'. It may take some minutes...")
+        response = requests.get(model_url)
+        if response.status_code == 200:
+            # Extract the zip file content
+            with ZipFile(BytesIO(response.content)) as zip_file:
+                # Create the directory for extraction if it doesn't exist
+                os.makedirs(model_path, exist_ok=True)
+                # Extract all contents to the specified path
+                zip_file.extractall(model_path)
+                print(f"Model successfully downloaded and extracted to '{model_path}'.\n")
+        else:
+            print(f"Failed to download file. Status code: {response.status_code}")
+            sys.exit(-1)
+    return tf.keras.models.load_model(model_path)
+
 
 # Loads the model into memory at startup to go faster upon prediction
-model = tf.keras.models.load_model('../model/BRNN')
+model = download_and_load_model(MODEL_PATH, MODEL_URL)
 
 
 def parse_line(line, allow_short_lines: bool):
@@ -46,7 +76,7 @@ def soft_voting(predictions):
 
 
 def predict(source_code: str) -> Dict[str, float]:
-    from configuration import LANGUAGES
+    from languages import LANGUAGES
     global model
     lines = source_code.split('\n')
     result = {}
@@ -60,4 +90,3 @@ def predict(source_code: str) -> Dict[str, float]:
         result[LANGUAGES[i]] = round(p*100, 2)
     return result
 
-

desktop-app/configuration.py renamed to desktop-app/example_code.py

@@ -1,12 +1,3 @@
-from tkinter import Tk, END, E, W
-from tkinter.ttk import Button, Combobox, Entry, LabelFrame, Treeview, Style
-
-from model import predict
-
-LANGUAGES = ["Assembly", "C", "C++", "C#", "CSS", "Go", "HTML", "Java", "JavaScript", "Kotlin",
-                   "Matlab", "Perl", "PHP", "Python", "R", "Ruby", "Scala", "SQL", "Swift", "TypeScript",
-                   "Unix Shell"]
-
 PROGRAM_EXAMPLES = {
     "Assembly": "ADD CX, [BX+SI*2+10]",
     "C": "int *numbers = malloc(sizeof(int));",

desktop-app/main.py

@@ -1,8 +1,7 @@
 from tkinter import Tk, END, E, W, ttk, StringVar, BooleanVar, NORMAL, DISABLED, Text, Event
 from tkinter.ttk import Button, Combobox, LabelFrame, Treeview, Style
-
-from configuration import PROGRAM_EXAMPLES, LANGUAGES
-from model import predict
+from example_code import PROGRAM_EXAMPLES
+from languages import LANGUAGES
 
 
 def sort_number(table: Treeview, column_name: str, ascending: bool) -> None:
@@ -27,6 +26,16 @@ def sort_string(table: Treeview, column_name: str, asc: bool) -> None:
     table.heading(column_name, command=lambda: sort_string(table, column_name, not asc))
 
 
+def import_model() -> None:
+    """Imports the model component from a sibling directory"""
+    import os
+    import sys
+    if not os.path.exists("../common/model.py"):
+        print("You need to include the common directory as a sibling directory of 'web-api'.", file=sys.stderr)
+        sys.exit(-1)
+    sys.path.append('../common')
+
+
 def main() -> None:
     """Creates the main window and runs the application"""
     # Main window
@@ -76,9 +85,10 @@ def select_lang(_: Event) -> None:
     cmb.bind("<<ComboboxSelected>>", select_lang)
 
     def get_prediction() -> None:
+        import model
         for item in results_tree.get_children():
             results_tree.delete(item)
-        predictions = sorted(predict(text.get("1.0", END)).items(), key=lambda x: x[1])
+        predictions = sorted(model.predict(text.get("1.0", END)).items(), key=lambda x: x[1])
         for data in predictions:
             results_tree.insert(parent="", index=0, values=data)
 
@@ -136,4 +146,5 @@ def text_changed() -> None:
 
 
 if __name__ == "__main__":
+    import_model()
     main()

web-api/main.py

@@ -1,5 +1,8 @@
 import sys
 import logging
+
+from flask_cors import CORS
+
 from services.prediction import brnn_predict_blueprint
 
 from flask import Flask, request, jsonify, Blueprint
@@ -23,8 +26,21 @@ def register_imported_blueprints(app: Flask) -> int:
     return count
 
 
+def import_model() -> None:
+    """Imports the model component from a sibling directory"""
+    import os
+    import sys
+    if not os.path.exists("../common/model.py"):
+        print("You need to include the common directory as a sibling directory of 'web-api'.", file=sys.stderr)
+        sys.exit(-1)
+    sys.path.append('../common')
+    import model  # loads the model into memory
+
+
 def main() -> None:
+    import_model()
     app = Flask(__name__)
+    CORS(app)
     register_imported_blueprints(app)
     app.run(debug=True)  # debug must be False to trace programs in PyCharm
 

web-api/services/prediction.py

@@ -1,8 +1,5 @@
 
 from flask import Blueprint, jsonify, request, Response
-
-from model import predict
-
 brnn_predict_blueprint = Blueprint(f"brnn_predict", __name__)
 
 
@@ -11,7 +8,8 @@ def get_predict() -> Response:
     """One mandatory 'source code' parameter must be passed
         http://127.0.0.1:5000/BRNN/predict?source_code=code
     """
+    import model
     source_code = request.args.get('source_code')
-    predictions = predict(source_code)
+    predictions = model.predict(source_code)
     return jsonify(predictions)