Update stats functions

- Added verify thicket structures function into each stats class
- Updated the arguments based on previous discussions within PAVE and UTK-LLNL meetings
- Added a conditional check for arguments that need a column from the performance dataframe to be passed. A value error will be raised if the arg is set equal to `None`.

* Removing .swp file

* Updating path to utils

* Updating stats functions to have: Verify thicket structure, updated args, and a conditional check

thicket/stats/calc_average.py

@@ -5,9 +5,10 @@
 
 import numpy as np
 import pandas as pd
+from ..utils import verify_thicket_structures
 
 
-def calc_average(thicket=None, columns=None):
+def calc_average(thicket, columns=None):
     """Calculate median and mean per node.
 
     Designed to take in a Thicket, and will append a column to the statsframe for
@@ -17,6 +18,13 @@ def calc_average(thicket=None, columns=None):
         thicket (thicket): Thicket object
         columns (list): list of hardware/timing metrics to perform average calculations on
     """
+    if columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.dataframe, index=["node", "profile"], columns=columns
+    )
+
     for column in columns:
         median = []
         mean = []

thicket/stats/calc_corr_nodewise.py

@@ -4,38 +4,57 @@
 # SPDX-License-Identifier: MIT
 
 from scipy import stats
+from ..utils import verify_thicket_structures
 
 
-def calc_corr_nodewise(thicket, base=None, correlate=[], correlation="pearson"):
+def calc_corr_nodewise(
+    thicket, base_column=None, correlate_columns=None, correlation="pearson"
+):
     """Calculate the nodewise correlation on user-specified columns.
 
     Calculates the correlation nodewise for user passed in columns. This can
     either be done for the EnsembleFrame or a super thicket.
 
     Arguments:
         thicket (thicket): Thicket object
-        base (str): base column that you want to compare
-        correlate (list): list of columns to correlate to the passed in base column
+        base_column (str): base column that you want to compare
+        correlate_columns (list): list of columns to correlate to the passed in base column
         correlation (str): correlation test to perform -- pearson (default),
             spearman, and kendall
     """
-    for col in correlate:
+    if base_column is None or correlate_columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    if "profile" in thicket.dataframe.index.names:
+        verify_thicket_structures(
+            thicket.dataframe,
+            index=["node", "profile"],
+            columns=[base_column] + correlate_columns,
+        )
+    else:
+        verify_thicket_structures(
+            thicket.dataframe,
+            index=["node", "thicket"],
+            columns=[base_column] + correlate_columns,
+        )
+
+    for col in correlate_columns:
         correlated = []
         for node in thicket.statsframe.dataframe.index.tolist():
             if correlation == "pearson":
-                pearson_base = thicket.dataframe.loc[node][base]
+                pearson_base = thicket.dataframe.loc[node][base_column]
                 pearson_correlate = thicket.dataframe.loc[node][col]
                 correlated.append(stats.pearsonr(pearson_base, pearson_correlate)[0])
             elif correlation == "spearman":
-                spearman_base = thicket.dataframe.loc[node][base]
+                spearman_base = thicket.dataframe.loc[node][base_column]
                 spearman_correlate = thicket.dataframe.loc[node][col]
                 correlated.append(stats.spearmanr(spearman_base, spearman_correlate)[0])
             elif correlation == "kendall":
-                kendall_base = thicket.dataframe.loc[node][base]
+                kendall_base = thicket.dataframe.loc[node][base_column]
                 kendall_correlate = thicket.dataframe.loc[node][col]
                 correlated.append(stats.kendalltau(kendall_base, kendall_correlate)[0])
             else:
                 raise ValueError("Invalid correlation")
         thicket.statsframe.dataframe[
-            base + "_vs_" + col + " " + correlation
+            base_column + "_vs_" + col + " " + correlation
         ] = correlated

thicket/stats/calc_deviation.py

@@ -5,9 +5,10 @@
 
 import numpy as np
 import pandas as pd
+from ..utils import verify_thicket_structures
 
 
-def calc_deviation(thicket=None, columns=None):
+def calc_deviation(thicket, columns=None):
     """Calculate standard deviation and variance per node.
 
     Designed to take in a Thicket, and will append a column to the statsframe
@@ -20,6 +21,13 @@ def calc_deviation(thicket=None, columns=None):
         thicket (thicket): Thicket object
         columns (list): list of hardware/timing metrics to perform deviation calculations on
     """
+    if columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.dataframe, index=["node", "profile"], columns=columns
+    )
+
     for column in columns:
         var = []
         std = []

thicket/stats/calc_extremum.py

@@ -4,9 +4,10 @@
 # SPDX-License-Identifier: MIT
 
 import pandas as pd
+from ..utils import verify_thicket_structures
 
 
-def calc_extremum(thicket=None, columns=None):
+def calc_extremum(thicket, columns=None):
     """Calculate min and max per node.
 
     Designed to take in a Thicket, and will append a column to the statsframe
@@ -19,6 +20,13 @@ def calc_extremum(thicket=None, columns=None):
         thicket (thicket): Thicket object
         columns (list): list of hardware/timing metrics to perform extremnum calculations on
     """
+    if columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.dataframe, index=["node", "profile"], columns=columns
+    )
+
     for column in columns:
         minimum = []
         maximum = []

thicket/stats/calc_percentiles.py

@@ -5,9 +5,10 @@
 
 import pandas as pd
 import numpy as np
+from ..utils import verify_thicket_structures
 
 
-def calc_percentile(thicket=None, columns=None):
+def calc_percentile(thicket, columns=None):
     """Calculate q-th percentile per node.
 
     Designed to take in a Thicket, and will append a column to the statsframe
@@ -27,6 +28,13 @@ def calc_percentile(thicket=None, columns=None):
         thicket (thicket): Thicket object
         columns (list): list of hardware/timing metrics to perform percentile calculations on
     """
+    if columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.dataframe, index=["node", "profile"], columns=columns
+    )
+
     for column in columns:
         percentiles = []
         for node in pd.unique(thicket.dataframe.reset_index()["node"].tolist()):

thicket/stats/check_normality.py

@@ -5,9 +5,10 @@
 
 import pandas as pd
 from scipy import stats
+from ..utils import verify_thicket_structures
 
 
-def check_normality(thicket=None, columns=None):
+def check_normality(thicket, columns=None):
     """
     Designed to take in a Thicket, and will append a column to the statsframe.
 
@@ -27,6 +28,12 @@ def check_normality(thicket=None, columns=None):
     statsframe: Returns statsframe with appended columns for normality check
 
     """
+    if columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.dataframe, index=["node", "profile"], columns=columns
+    )
 
     for column in columns:
         normality = []

thicket/stats/display_heatmap.py

@@ -4,9 +4,10 @@
 # SPDX-License-Identifier: MIT
 
 import seaborn as sns
+from ..utils import verify_thicket_structures
 
 
-def display_heatmap(thicket=None, columns=None, **kwargs):
+def display_heatmap(thicket, columns=None, **kwargs):
     """Display a heatmap.
 
     Arguments:
@@ -16,6 +17,13 @@ def display_heatmap(thicket=None, columns=None, **kwargs):
     Returns:
         (matplotlib Axes): object for managing plot
     """
+    if columns is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.statsframe.dataframe, index=["node"], columns=columns
+    )
+
     thicket.statsframe.dataframe.index = thicket.statsframe.dataframe.index.map(str)
 
     ax = sns.heatmap(thicket.statsframe.dataframe[columns], **kwargs)

thicket/stats/display_histogram.py

@@ -5,30 +5,38 @@
 
 import pandas as pd
 import seaborn as sns
+from ..utils import verify_thicket_structures
 
 
-def display_histogram(thicket=None, node=None, metric=None, **kwargs):
+def display_histogram(thicket, node=None, column=None, **kwargs):
     """Display a histogram.
 
     Arguments:
         thicket (thicket): Thicket object
         node (str): node object
-        metric (str): metric from ensemble frame
+        column (str): column from ensemble frame
 
     Returns:
         (matplotlib Axes): object for managing plot
     """
+    if column is None or node is None:
+        raise ValueError("To see a list of valid columns run get_perf_columns().")
+
+    verify_thicket_structures(
+        thicket.dataframe, index=["node", "profile"], columns=[column]
+    )
+
     df = pd.melt(
         thicket.dataframe.reset_index(),
         id_vars="node",
-        value_vars=metric,
-        value_name=metric,
+        value_vars=column,
+        value_name=column,
     )
 
     df["node"] = df["node"].astype(str)
 
     filtered_df = df[df["node"] == node]
 
-    ax = sns.displot(filtered_df, x=metric, kind="hist", **kwargs)
+    ax = sns.displot(filtered_df, x=column, kind="hist", **kwargs)
 
     return ax