add latex and csv function

app.py

@@ -124,16 +124,19 @@ def plot_radar_chart(df, attributes):
 
     return fig
 
-def pop_latex_table(caption, label, dataframe):
-    table = gen_latex_table(caption, label, dataframe)
+def pop_latex_table(caption, label, lang, dataset, cat):
+    table = gen_latex_table(caption, label, global_df[lang][dataset][cat])
     return gr.Textbox(table, label="LaTeX Table", visible=True)
 
-def generate_csv(df, filename):
-    df.to_csv(filename, index=False)
-    download_link = gr.File(label="Download Link", type="filepath", value=filename,
+def generate_csv(lang, dataset, cat, shot, chosen_dataset):
+    df = global_df[lang][dataset][cat]
+    df.to_csv(f"{chosen_dataset}_{lang}_{cat}_{shot}.csv", index=False)
+    download_link = gr.File(label="Download Link", type="filepath", value=f"{chosen_dataset}_{lang}_{cat}_{shot}.csv",
         visible=True)
     return download_link
 
+global_df = {}
+
 def create_lang_leader_board(lang_dict, lang, dis_lang='en'):
     best_scores = {}
     best_plot_datasets = []
@@ -150,43 +153,104 @@ def create_lang_leader_board(lang_dict, lang, dis_lang='en'):
     # plot = plot_radar_chart(pd.DataFrame(best_scores), best_plot_datasets)
     # gr.Plot(plot)
     tab_list = []
+    
+    latex_btn_list = []
+    latex_textbox_list = []
+
+    csv_btn_list = []
+    csv_textbox_list = []
+
+    export_params = []
+
+    if lang not in global_df:
+        global_df[lang] = {}
 
     for dataset, value in lang_dict.items():
+        
+        if dataset not in global_df[lang]:
+            global_df[lang][dataset] = {}
+
         chosen_dict = dataset_abbr_en_dict if dis_lang == "en" else dataset_abbr_zh_dict
         with gr.Tab(chosen_dict[dataset]) as tab:
             for cat, df in value.items():
+
+                if cat not in global_df[lang][dataset]:
+                    global_df[lang][dataset][cat] = df.copy(deep=True)
+
                 if cat == 'mc':
                     for shot in ['Zeroshot', 'Fewshot']:
                         with gr.Tab(f'Multiple Choice Question ({shot})'):
                             df_component = dataframe_to_gradio(df, is_mc=True, shot=shot)
                             # 加一个latex表格导出按钮， 按一下弹出一个浮动文本窗口
-                            # with gr.Row():
-                            #     latex_button = gr.Button("Export LaTeX Table", variant="primary")
-                            #     csv_button = gr.Button("Export CSV", variant="primary")
-
-                            # latex_textbox = gr.Textbox(label="LaTeX Table", visible=False)
-                            # download_link = gr.File(label="Download Link", type="filepath",
-                            #     visible=False)
-
-                            # latex_button.click(lambda: pop_latex_table(
-                            #     caption=f"{chosen_dict[dataset]} Multiple Choice Question ({shot}, {lang}) Leaderboard",
-                            #     label=f"tab:{dataset}_{shot}_{lang}",
-                            #     dataframe=df,
-                            # ), inputs=[], outputs=[latex_textbox])
-                            # csv_button.click(lambda: generate_csv(df, f"/tmp/opseval-{chosen_dict[dataset]}-mc-{shot}.csv"), inputs=[], outputs=[download_link])
+                            latex_textbox = gr.Textbox(label="LaTeX Table", visible=False)
+                            download_link = gr.File(label="Download Link", type="filepath",
+                                visible=False)
+
+                            latex_caption = gr.Text(value=f"{chosen_dict[dataset]} Multiple Choice Question ({shot}, {lang}) Leaderboard", visible=False)
+                            latex_label = gr.Text(value=f"tab:{dataset}_{shot}_{lang}", visible=False)
+                            param_lang = gr.Text(value=lang, visible=False)
+                            param_dataset = gr.Text(value=dataset, visible=False)
+                            param_cat = gr.Text(value=cat, visible=False)
+                            param_shot = gr.Text(value=shot, visible=False)
+                            dataset_name = gr.Text(value=chosen_dict[dataset], visible=False)
+                            with gr.Row():
+                                latex_button = gr.Button("Export LaTeX Table", variant="primary")
+                                csv_button = gr.Button("Export CSV", variant="primary")
+
+                            latex_btn_list.append(latex_button)
+                            latex_textbox_list.append(latex_textbox)
+
+                            csv_btn_list.append(csv_button)
+                            csv_textbox_list.append(download_link)
+
+                            export_params.append({
+                                "caption": latex_caption,
+                                "label": latex_label,
+                                "lang": param_lang,
+                                "dataset": param_dataset,
+                                "cat": param_cat,
+                                "shot": param_shot,
+                                "dataset_name": dataset_name
+                            })
+
+                            
+                            
                 else:
                     with gr.Tab('Question Answering'):
                         df_component = dataframe_to_gradio(df, is_mc=False)
-                        # df_list.append(df_component)
-                        # button = gr.Button("Export LaTeX Table", variant="primary")
-                        # latex_textbox = gr.Textbox(label="LaTeX Table", visible=False)
-                        # button.click(lambda: pop_latex_table(
-                        #     caption=f"{chosen_dict[dataset]} {shot} {lang} Leaderboard",
-                        #     label=f"tab:{dataset}_{shot}_{lang}",
-                        #     dataframe=df,
-                        # ), inputs=[], outputs=[latex_textbox])
+                        # 加一个latex表格导出按钮， 按一下弹出一个浮动文本窗口
+                        latex_textbox = gr.Textbox(label="LaTeX Table", visible=False)
+                        download_link = gr.File(label="Download Link", type="filepath",
+                            visible=False)
+
+                        latex_caption = gr.Text(value=f"{chosen_dict[dataset]} Question Answering ({lang}) Leaderboard", visible=False)
+                        latex_label = gr.Text(value=f"tab:{dataset}_{lang}", visible=False)
+                        param_lang = gr.Text(value=lang, visible=False)
+                        param_dataset = gr.Text(value=dataset, visible=False)
+                        param_cat = gr.Text(value=cat, visible=False)
+                        param_shot = gr.Text(value=shot, visible=False)
+                        dataset_name = gr.Text(value=chosen_dict[dataset], visible=False)
+                        with gr.Row():
+                            latex_button = gr.Button("Export LaTeX Table", variant="primary")
+                            csv_button = gr.Button("Export CSV", variant="primary")
+
+                        latex_btn_list.append(latex_button)
+                        latex_textbox_list.append(latex_textbox)
+
+                        csv_btn_list.append(csv_button)
+                        csv_textbox_list.append(download_link)
+
+                        export_params.append({
+                            "caption": latex_caption,
+                            "label": latex_label,
+                            "lang": param_lang,
+                            "dataset": param_dataset,
+                            "cat": param_cat,
+                            "shot": param_shot,
+                            "dataset_name": dataset_name
+                        })
         tab_list.append(tab)
-    return tab_list
+    return tab_list, latex_btn_list, latex_textbox_list, csv_btn_list, csv_textbox_list, export_params
     
 def get_latest_modification_date():
     latest = 0
@@ -213,21 +277,23 @@ translation_dict = {
 def get_language_lb(language):
     tab_dict = {'English': None, 'Chinese': None}
     for key, dict in dict_lang.items():
-        tab_list = create_lang_leader_board(dict, key, language)
-        tab_dict[key] = tab_list
-    return [*tab_dict['English'], *tab_dict['Chinese']]
+        tab_list, latex_btn_list, latex_textbox_list, csv_btn_list, csv_textbox_list, export_params = create_lang_leader_board(dict, key, language)
+        tab_dict[key] = (tab_list, latex_btn_list, latex_textbox_list, csv_btn_list, csv_textbox_list, export_params)
+    return tab_dict
 
 def switch_language(language):
     # gr.update(visible=True)
-    return translation_dict[language]['title'], translation_dict[language]['intro'], translation_dict[language]['lb_sec'], *get_language_lb(language), language
+    tab_dict = get_language_lb(language)
+    tab_list = [*(tab_dict['English'][0]), *(tab_dict['Chinese'][0])]
+    return translation_dict[language]['title'], translation_dict[language]['intro'], translation_dict[language]['lb_sec'], *tab_list, language
 
 def get_lb_body(language='en'):
     tab_dict = {'English': None, 'Chinese': None}
     with gr.Blocks() as body:
         for key, dict in dict_lang.items():
             with gr.Tab(key):
-                tab_list = create_lang_leader_board(dict, key, language)
-                tab_dict[key] = tab_list
+                tab_list, latex_btn_list, latex_textbox_list, csv_btn_list, csv_textbox_list, export_params = create_lang_leader_board(dict, key, language)
+                tab_dict[key] = (tab_list, latex_btn_list, latex_textbox_list, csv_btn_list, csv_textbox_list, export_params)
     return body, tab_dict
 
 def launch_gradio():
@@ -247,14 +313,32 @@ def launch_gradio():
         
         lb_body, tab_dict = get_lb_body(language=lang_state.value)
 
-        tab_list = [*tab_dict['English'], *tab_dict['Chinese']]
+        for lang in ['English', 'Chinese']:
+            tab_list, latex_btn_list, latex_textbox_list, csv_btn_list, csv_textbox_list, export_params = tab_dict[lang]
+            for btn, params, textbox in zip(latex_btn_list, export_params, latex_textbox_list):
+                btn.click(pop_latex_table, inputs=[
+                    params["caption"], 
+                    params["label"], 
+                    params["lang"], 
+                    params["dataset"], 
+                    params["cat"]], 
+                    outputs=[textbox])
+            for btn, params, textbox in zip(csv_btn_list, export_params, csv_textbox_list):
+                btn.click(generate_csv, inputs=[
+                    params["lang"], 
+                    params["dataset"], 
+                    params["cat"], 
+                    params["shot"], 
+                    params["dataset_name"]], 
+                    outputs=[textbox])
+        tab_list = [*(tab_dict['English'][0]), *(tab_dict['Chinese'][0])]
         # print(tab_list)
 
         en_button.click(switch_language, inputs=[gr.State("en")], outputs=[title, intro, leaderboard_section, *tab_list, lang_state], postprocess=False)
         zh_button.click(switch_language, inputs=[gr.State("zh")], outputs=[title, intro, leaderboard_section, *tab_list, lang_state], postprocess=False)
         
 
-    demo.launch()
+    return demo
 
 pd.set_option('display.float_format', '{:.02f}'.format)
 
@@ -262,4 +346,5 @@ scheduler = BackgroundScheduler()
 scheduler.add_job(launch_gradio, 'interval', hours=1)
 scheduler.start()
 
-launch_gradio()
+demo = launch_gradio()
+demo.launch()

latex_utils.py

@@ -1,5 +1,6 @@
 import pandas as pd
 from tabulate import tabulate
+import re
 
 # 创建一个示例 DataFrame
 data = {
@@ -33,8 +34,16 @@ latex_table = tabulate(df, headers='keys', tablefmt='latex', showindex=False,
                        colalign='left')
 
 def gen_latex_table(caption, label, dataframe):
-    table = tabulate(dataframe, headers='keys', tablefmt='latex', showindex=False,
-                     colalign='left')
+    # print(caption)
+    # print(label)
+    # print(dataframe)
+    if len(dataframe.columns) > 2:
+        table = tabulate(dataframe, headers='keys', tablefmt='latex', showindex=False, 
+                         missingval='/',
+                        colalign='left')
+    else:
+        table = tabulate(dataframe, headers='keys', tablefmt='latex', showindex=False, 
+                         missingval='/')
     table = (
         "\\begin{table}[]\n"
         f"\\caption{{{caption}}}\n"
@@ -50,7 +59,11 @@ def gen_latex_table(caption, label, dataframe):
     # 将table中的第二个\hline改为\midrule
     table = table.replace("\\hline", "\\midrule", 1)
     # 将table中的最后一个\hline改为\bottomrule
-    table = table.replace("\\hline", "\\bottom", 1)
+    table = table.replace("\\hline", "\\bottomrule", 1)
+
+    # 将table中的所有nan（独立的单词）替换为/
+    table = re.sub(r'\bnan\b', '/', table)
+
     return table
 
 # # 添加表格环境