From 4e2e7a60c7d0bf82751b3437f468fcc739460fc2 Mon Sep 17 00:00:00 2001
From: Tim Jammer <tim.jammer@tu-darmstadt.de>
Date: Wed, 31 Jul 2024 10:39:55 +0200
Subject: [PATCH] changed noise table to csv output
---
MBI.py | 60 +++++++++++++++++++++++++++-------------------------------
1 file changed, 28 insertions(+), 32 deletions(-)
diff --git a/MBI.py b/MBI.py
index 48900c88e..6ffd7dbf6 100755
--- a/MBI.py
+++ b/MBI.py
@@ -638,36 +638,24 @@ def cmd_helpfulness(rootdir, toolnames):
pass
-def cmd_csv(rootdir, toolnames):
+def cmd_csv(rootdir, toolnames,print_to_console=False):
here = os.getcwd()
os.chdir(rootdir)
- used_toolnames = []
outpath = f'{rootdir}/csv/'
# Create directory for output if not present
pathlib.Path(outpath).mkdir(parents=True, exist_ok=True)
-
- # select the tools for which we have some results
- print("Produce the stats for:", end='')
- for toolname in toolnames:
- if not toolname in tools:
- raise Exception(f"Tool {toolname} does not seem to be a valid name.")
-
- if os.path.exists(f'{args.logs_dir}/{toolname}'):
- used_toolnames.append(toolname)
- print(f' {toolname}', end="")
-
- print(".")
-
-
+ df_noise_ratio = pd.DataFrame(columns=toolnames)
+ df_overall_noise_ratio = pd.DataFrame(columns=toolnames)
pd.set_option('display.max_columns', 14)
for toolname in toolnames:
df = read_tool_reports(rootdir, toolname)
df.to_csv(f'{outpath}/{toolname}_raw.csv', index=False)
- print(f"=== {toolname} ===")
+ if print_to_console:
+ print(f"=== {toolname} ===")
# Output for each type of TP
for (colname) in ["base", "class", "line", "class_line", "class_line_no_line_noise", "class_line_no_line_noise", "class_line_no_class_noise"]:
@@ -676,27 +664,35 @@ def cmd_csv(rootdir, toolnames):
df_classified.to_csv(f'{outpath}/{toolname}_{colname}_full.csv', index=False)
df_result = aggregate_metrics_per_category(df_classified)
df_result.to_csv(f'{outpath}/{toolname}_{colname}.csv', index=True)
- print(f"\n{colname}:")
- print(df_result[["CE", "RE", "TP", "TN", "FP", "FN", "coverage", "conclusiveness", "specificity", "recall", "precision", "f1", "overallaccuracy"]])
- df_result[["CE", "RE", "TP", "TN", "FP", "FN", "coverage", "conclusiveness", "specificity", "recall", "precision", "f1", "overallaccuracy"]].style.format(precision=2).to_latex(f'{outpath}/{toolname}_{colname}.tex')
+ if print_to_console:
+ print(f"\n{colname}:")
+ print(df_result[["CE", "RE", "TP", "TN", "FP", "FN", "coverage", "conclusiveness", "specificity", "recall", "precision", "f1", "overallaccuracy"]])
+ df_result[["CE", "RE", "TP", "TN", "FP", "FN", "coverage", "conclusiveness", "specificity", "recall", "precision", "f1", "overallaccuracy"]].style.format(precision=2).to_latex(f'{outpath}/{toolname}_{colname}.tex')
plot_helpfulness(df, outpath, toolname)
- df_plot = df.groupby("category").sum()
- df_plot.loc["ALL"] = df_plot.sum(axis=0)
- df_plot.drop("other", axis=0, inplace=True)
- df_plot["noise_ratio"] = df_plot["num_noise_line"] / df_plot["num_error_reports"]
- print("overall_noise")
- print(df_plot["noise_ratio"])
+ df_noise_per_tool = df.groupby("category").sum()
+ df_noise_per_tool.loc["ALL"] = df_noise_per_tool.sum(axis=0)
+ df_noise_per_tool.drop("other", axis=0, inplace=True)
+ df_noise_per_tool["noise_ratio"] = df_noise_per_tool["num_noise_line"] / df_noise_per_tool["num_error_reports"]
+ if print_to_console:
+ print("overall_noise")
+ print(df_noise_per_tool["noise_ratio"])
+ df_overall_noise_ratio[toolname] =df_noise_per_tool["noise_ratio"]
df_copy = df.copy()
df_copy.loc[df_copy['ERROR_EXPECTED'] == False, ['num_noise_class_line', 'num_error_reports']] = 0
- df_plot = df_copy.groupby("category").sum()
- df_plot.loc["ALL"] = df_plot.sum(axis=0)
- df_plot.drop("other", axis=0, inplace=True)
- df_plot["noise_ratio"] = df_plot["num_noise_line"] / df_plot["num_error_reports"]
- print("noise_in_cases_where_errors_are_present")
- print(df_plot[["noise_ratio", "num_noise_class_line", "num_error_reports"]])
+ df_noise_per_tool = df_copy.groupby("category").sum()
+ df_noise_per_tool.loc["ALL"] = df_noise_per_tool.sum(axis=0)
+ df_noise_per_tool.drop("other", axis=0, inplace=True)
+ df_noise_per_tool["noise_ratio"] = df_noise_per_tool["num_noise_line"] / df_noise_per_tool["num_error_reports"]
+ if print_to_console:
+ print("noise_in_cases_where_errors_are_present")
+ print(df_noise_per_tool[["noise_ratio", "num_noise_class_line", "num_error_reports"]])
+ df_noise_ratio[toolname] = df_noise_per_tool["noise_ratio"]
+
+ df_noise_ratio.to_csv(f'{outpath}/noise.csv')
+ df_overall_noise_ratio.to_csv(f'{outpath}/overall_noise_including_unexpected.csv')
def plot_helpfulness(df, outpath, toolname):
--
GitLab