init

README.md

@@ -1,8 +1,8 @@
 ---
 title: RelevancyApp
-emoji: 🏆
-colorFrom: green
-colorTo: red
+emoji: 💻
+colorFrom: red
+colorTo: yellow
 sdk: streamlit
 sdk_version: 1.35.0
 app_file: app.py

app.py

@@ -0,0 +1,197 @@
+from utils import (
+    container_emphasis_css,
+    container_regular_css,
+    question_font,
+    modify_sql_guid_injection
+)
+
+import streamlit as st
+import clipboard
+import textwrap as tw
+import pandas as pd
+from streamlit_extras.stylable_container import stylable_container
+
+@st.cache_data
+def load_data(file):
+    return pd.read_csv(file)
+
+st.set_page_config(layout="wide", initial_sidebar_state="expanded")
+st.title("Relevancy Compare App")
+st.divider()
+st.markdown(question_font, unsafe_allow_html=True)
+
+
+def edit_df_data_change():
+    edited_rows: dict = st.session_state.edit_df_data['edited_rows']
+    st.session_state.previous_row_index = st.session_state.previous_row_index
+    st.session_state.current_row_index = next(iter(edited_rows))
+    st.session_state.df = st.session_state.df.assign(selected=False)
+    update_dict = {idx: values for idx, values in edited_rows.items()}
+    st.session_state.df.update(pd.DataFrame.from_dict(update_dict, orient='index'))
+
+
+st.need_rerun = False
+if "df" not in st.session_state:
+    uploaded_file = st.file_uploader("Choose a CSV file", type="csv")
+    st.need_rerun = True
+else:
+    uploaded_file = st.session_state.df
+if uploaded_file is not None:
+    with st.container(height=250):
+        if "df" not in st.session_state:
+            df = load_data(uploaded_file)
+            st.session_state.df_shape = df.shape
+            # selected load
+            select_values = [False] * len(df)
+            
+            if "Comment" not in df.columns:
+                comments = [''] * len(df)
+                df.insert(0, 'Comment', comments)
+                st.session_state.default_row_index = 0
+            else:
+                df['Comment'] = df['Comment'].astype(str)
+                comments = [c if c != 'nan' else '' for c in df['Comment'].tolist()]
+                df['Comment'] = comments
+                first_nan = 0
+                for i, c in enumerate(comments):
+                    if not c:
+                        first_nan = i
+                        break
+                st.session_state.default_row_index = first_nan
+            
+            st.session_state.current_row_index = st.session_state.default_row_index
+            st.session_state.previous_row_index = st.session_state.current_row_index
+
+            df.insert(0, 'Select', select_values)
+            df.at[st.session_state.current_row_index, 'Select'] = True
+            st.session_state.comments_add = dict((idx, t) for idx, t in enumerate(df["Comment"]))
+            st.session_state['df'] = df
+        
+        df_col, = st.columns(1)
+
+    st.session_state.previous_row_index = st.session_state.current_row_index
+    with st.container():
+        bcol1, bcol2, _ = st.columns([1, 1, 5])
+        with bcol1:
+            previous_row = st.button("Prev Row")
+        with bcol2:
+            next_row = st.button("Next Row")
+            
+
+    if "current_row_index" in st.session_state and previous_row:
+        if st.session_state.current_row_index > 0:
+            st.session_state.previous_row_index = st.session_state.current_row_index
+            st.session_state.current_row_index -= 1
+
+    if "current_row_index" in st.session_state and next_row:
+        if st.session_state.current_row_index < st.session_state.df_shape[0] - 1:
+            st.session_state.previous_row_index = st.session_state.current_row_index
+            st.session_state.current_row_index += 1
+
+    row_index = st.session_state.current_row_index
+    row_data = st.session_state.df.loc[row_index]
+    question = row_data['question']
+    pred_relevant_choice = row_data['pred_relevant_choice']
+    correct_response = row_data['correct_response']
+    first_css = container_regular_css
+    second_css = container_regular_css
+    first_color = ':black'
+    second_color = ':black'
+    
+        
+    st.write(f"##### Row {row_index}/{st.session_state.df_shape[0]-1}")
+    with st.container(border=True):
+        # st.write(f"Question: {question}")
+        st.markdown(f'<p class="question-font">[Q]: {question}</p>', unsafe_allow_html=True)
+        select_by = st.selectbox("Mark By: ", ["pred_relevant_choice", "correct_response"])
+        col1_res, col2_res, _ = st.columns([0.2, 0.2, 0.6])
+        if select_by == 'pred_relevant_choice':
+            col1_res.write(f":blue[pred_relevant_choice: {pred_relevant_choice}]")
+            col2_res.write(f"correct_response: {correct_response}")
+        else:
+            col1_res.write(f"pred_relevant_choice: {pred_relevant_choice}")
+            col2_res.write(f":blue[correct_response: {correct_response}]")
+
+
+        if row_data[select_by].lower() == 'first' or row_data[select_by].lower() == 'both':
+            first_css = container_emphasis_css
+            first_color = ':blue'
+
+            if row_data[select_by].lower() == 'both':
+                second_css = container_emphasis_css
+                second_color = ':blue'
+        else:
+            second_css = container_emphasis_css
+            second_color = ':blue'
+
+        col1, col2 = st.columns(2)
+        with col1:
+            st.write("Target SQL:", use_container_width=True)
+            with stylable_container(
+                key="first_sql_container",
+                css_styles=first_css
+                ):
+                target_sql_text = modify_sql_guid_injection(row_data['target_sql'].strip())
+                #st.write(f"{first_color}[{target_sql_text}]", use_container_width=True)
+                col1, _ = st.columns([0.99, 0.01])
+                with col1:
+                    st.code(
+                        target_sql_text,
+                        language="sql"
+                    )
+                    # st.code(
+                    #     "\n".join(
+                    #         tw.wrap(
+                    #         target_sql_text,
+                    #         width=60,
+                    #         )
+                    #     ),
+                    #     language="sql"
+                    # )
+
+        with col2:
+            st.write("Predicted SQL:", use_container_width=True)
+            with stylable_container(
+                key="second_sql_container",
+                css_styles=second_css
+                ):
+                predicted_sql_text = modify_sql_guid_injection(row_data['predicted_sql'].strip())
+                # st.write(f"{second_color}[{predicted_sql_text}]")
+                col1, _ = st.columns([0.99, 0.01])
+                with col1:
+                     st.code(
+                        predicted_sql_text,
+                        language="sql"
+                    )
+        comment = st.text_area("Comment", value=row_data['Comment'])
+    
+    with st.container():
+        bcol1, bcol2, _, _, _, _, _, _, _, _ = st.columns(10)
+        with bcol1:
+            save = st.button("Save")
+        with bcol2:
+            download = st.button("Download")
+
+    if save:
+        st.session_state.comments_add[row_index] = comment
+        st.session_state.df.at[row_index, 'Comment'] = comment
+        df = st.session_state.df.drop('Select', axis=1)
+        df.to_csv('modified.csv', index=False)
+        # st.rerun()
+    
+    if download:
+        with open('modified.csv') as f:
+            st.download_button('Download modified CSV', f, file_name='modified.csv')
+    
+    st.session_state.df['Select'] = False
+    st.session_state.df.at[st.session_state.current_row_index, 'Select'] = True
+
+    edited_df = df_col.data_editor(st.session_state.df, num_rows="fixed", use_container_width=True, key='edit_df_data', on_change=edit_df_data_change)
+
+    #st.session_state.mannul_select = edited_df[edited_df['Select']].index
+    st.session_state.df = edited_df
+
+    if st.need_rerun:
+        st.need_rerun = False
+        st.rerun()
+

requirements.txt

@@ -0,0 +1 @@
+streamlit-extras

utils.py

@@ -0,0 +1,41 @@
+container_emphasis_css = """
+{
+    border: 5px solid rgba(128, 128, 128, 1);
+    border-radius: 0.5rem;
+    padding: 0em 1em 1em 0.5em;
+    background-color: rgba(0, 128, 0, 0.2); 
+    overflow-wrap: break-word; /* Added to ensure long words break */
+    word-wrap: break-word; /* Added for compatibility with older browsers */
+    word-break: break-all; /* Added to handle breaking of long strings */
+    white-space: pre-wrap; /* Added to respect new lines and wrap text */
+}
+"""
+
+container_regular_css = """
+{
+    border: 5px solid rgba(0, 0, 0, 0.2);
+    border-radius: 0.5rem;
+    padding: 0em 1em 1em 0.5em;
+    background-color: rgba(255, 255, 255, 0); 
+    overflow-wrap: break-word; /* Added to ensure long words break */
+    word-wrap: break-word; /* Added for compatibility with older browsers */
+    word-break: break-all; /* Added to handle breaking of long strings */
+    white-space: pre-wrap; /* Added to respect new lines and wrap text */
+}
+"""
+
+
+question_font = """
+<style>
+.question-font {
+    font-size: 16px !important;
+    font-weight: bold;
+    font-family: 'Arial', sans-serif;
+}
+</style>
+"""
+
+
+def modify_sql_guid_injection(sql):
+    injected = "where UPPER(CompanyGuid) = '14A328EE-1ABF-448C-9F86-8D4E05838F8D')"
+    return sql.replace(injected, '\n' + injected)