Update app.py

app.py

@@ -9,7 +9,6 @@ import zipfile
 import os
 import io
 from gradio_huggingfacehub_search import HuggingfaceHubSearch
-from matheel.similarity import get_sim_list
 
 def calculate_similarity(code1, code2, Ws, Wl, Wj, model_name):
     model = SentenceTransformer(model_name)
@@ -22,8 +21,78 @@ def calculate_similarity(code1, code2, Ws, Wl, Wj, model_name):
 
     return "The similarity score between the two codes is: %.2f" % overall_similarity
 
-def get_sim_list_gradio(zipped_file,Ws, Wl, Wj, model_name,threshold,number_results):
-    result = get_sim_list(zipped_file,Ws, Wl, Wj, model_name,threshold,number_results)
+# Define the function to process the uploaded file and return a DataFrame
+def extract_and_read_compressed_file(file_path):
+    file_names = []
+    codes = []
+
+    # Handle .zip files
+    if file_path.endswith('.zip'):
+        with zipfile.ZipFile(file_path, 'r') as z:
+            file_names = z.namelist()
+            codes = [z.read(file).decode('utf-8', errors='ignore') for file in file_names]
+
+    else:
+        raise ValueError("Unsupported file type. Only .zip is supported.")
+
+    return file_names, codes
+
+def filter_and_return_top(df, similarity_threshold,returned_results):
+    filtered_df = df[df['similarity_score'] > similarity_threshold]
+    return filtered_df.head(returned_results)
+
+# Perform paraphrase mining with the specified weights
+def perform_paraphrase_mining(model, codes_list, weight_semantic, weight_levenshtein, weight_jaro_winkler):
+    return paraphrase_mining_with_combined_score(
+        model,
+        codes_list,
+        weight_semantic=weight_semantic,
+        weight_levenshtein=weight_levenshtein,
+        weight_jaro_winkler=weight_jaro_winkler
+    )
+
+def paraphrase_mining_with_combined_score(
+    model,
+    sentences: List[str],
+    show_progress_bar: bool = False,
+    weight_semantic: float = 1.0,
+    weight_levenshtein: float = 0.0,
+    weight_jaro_winkler: float = 0.0
+):
+    embeddings = model.encode(
+        sentences, show_progress_bar=show_progress_bar, convert_to_tensor=True)
+    paraphrases = util.paraphrase_mining_embeddings(embeddings, score_function=util.cos_sim)
+
+    results = []
+    for score, i, j in paraphrases:
+        lev_ratio = Levenshtein.normalized_similarity(sentences[i], sentences[j])
+        jaro_winkler_ratio = JaroWinkler.normalized_similarity(sentences[i], sentences[j])
+
+        combined_score = (weight_semantic * score) + \
+                         (weight_levenshtein * lev_ratio) + \
+                         (weight_jaro_winkler * jaro_winkler_ratio)
+
+        results.append([combined_score, i, j])
+
+    results = sorted(results, key=lambda x: x[0], reverse=True)
+    return results
+
+def get_sim_list(zipped_file,Ws, Wl, Wj, model_name,threshold,number_results):
+    file_names, codes = extract_and_read_compressed_file(zipped_file)
+    model = SentenceTransformer(model_name)
+    code_pairs = perform_paraphrase_mining(model, codes,Ws, Wl, Wj)
+    pairs_results = []
+
+    for score, i, j in code_pairs:
+      pairs_results.append({
+        'file_name_1': file_names[i],
+        'file_name_2': file_names[j],
+        'similarity_score': score
+    })
+
+    similarity_df = pd.concat([pd.DataFrame(pairs_results)], ignore_index=True)
+    similarity_df = similarity_df.sort_values(by='similarity_score', ascending=False)
+    result = filter_and_return_top(similarity_df,threshold,number_results).round(2)
     return result
 
 # Define the Gradio app