MangoHud/bin/mangoplot.py
flightlessmango 6c665653ba mangoplot: look for header further down than 4 rows
Co-authored-by: Adel KARA SLIMANE <adel.ks@zegrapher.com>
2024-06-27 14:53:25 +02:00

449 lines
14 KiB
Python
Executable File

#!/usr/bin/env python
r"""
Script to plot all the MangoHud benchmarks contained in a given folder.
"""
from pathlib import Path
import argparse
import csv
from typing import List, Union
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.widgets import Cursor
from matplotlib.colors import LinearSegmentedColormap
from matplotlib.ticker import EngFormatter
plt.rcParams['font.family'] = "Lato,serif"
plt.rcParams['font.weight'] = "600"
background_color = "#1A1C1D"
legend_facecolor = "#585f63"
legend_textcolor = "#cccbc9"
text_color = "#e8e6e3"
mango_color = "#BB770A"
graphbox_linewidth = 1.5
mango_cmap = LinearSegmentedColormap.from_list("mango_heat", [background_color, mango_color])
def identity(val):
r"""
returns the value as-is
"""
return val
def get_integer(val: str) -> int:
r"""
interprets the str 'val' as an integer and returns it
"""
if is_integer(val):
return int(val)
else:
raise ValueError("Casting a non integer value: ", val)
def is_integer(s: str) -> bool:
r"""
tests if 's' is an integer and returns a bool
"""
try:
int(s)
return True
except ValueError:
return False
def get_float(val):
r"""
interprets the str 'val' as a float and returns it
"""
if is_float(val):
return float(val)
else:
return float("nan")
def is_float(s: str) -> bool:
r"""
tests if 's' is an float and returns a bool
"""
try:
float(s)
return True
except ValueError:
return False
class Database:
r"""
A class that contains all the csv files within
the folder that it is instanced with
"""
def __init__(self,
data_folder_path=None,
csv_separator=" ",
filename_var_separator="|"):
self.datafiles = []
self.result_names_col = None
self.result_values_col = None
self.sim_settings_names_col = None
self.sim_settings_values_col = None
if data_folder_path:
self.load_from_folder(
data_folder_path,
csv_separator,
filename_var_separator)
def load_from_folder(self,
data_folder_path,
csv_separator=" ",
filename_var_separator="|"):
r"""
Load all CSV files form the given folder
"""
filepaths = list(Path(data_folder_path).rglob("*.csv"))
self.datafiles = []
N = len(filepaths)
print(f"Loading {N} benchmark files")
for filepath in filepaths:
try:
datafile = BenchmarkFile(
str(filepath),
csv_separator=csv_separator,
filename_var_separator=filename_var_separator)
self.datafiles.append(datafile)
except Exception:
pass
self.datafiles.sort()
class BenchmarkFile:
r"""
A class that represents a single CSV file, can load CSV files
with arbitrary separators. It can return separately any column
of the file and any mathematical combinations of its columns.
"""
def __init__(self,
filepath="",
filename_var_separator="|",
csv_separator=" "):
self.csv_separator = csv_separator
self.filepath = Path(filepath)
self.filename = self.filepath.name
self.filename_var_separator = filename_var_separator
self.variables = dict()
self.skip_lines = None
self.columns = []
self.column_name_to_index = dict()
self._is_data_loaded = False
if not self.filepath.is_file():
raise Exception("CSV file does not exist")
self._read_column_names()
def __lt__(self, other):
stem = self.filename[:-4] # remove the trailing ".csv"
other_stem = other.filename[:-4]
if stem.startswith(other_stem):
return False
elif stem.startswith(other_stem):
return True
else:
return stem < other_stem
def set_variable(self, name, value):
r"""
Saves a variable within the datafile instance
Note: it will not be saved to disk, it's just a helper method to
attach variables to a given data file.
"""
self.variables[name] = value
def get_variable(self, name):
r"""
Retrieves a saved variable in the instance
"""
return self.variables[name]
def _read_column_names(self):
r"""
Read the first few lines of the benchmark file
to look for the row taht contains the benchmark's
column names i.e. "fps", "frametime", "cpu_load"... etc
and save the columns names and their index
Note: we decide that we found the right row by looking if it
contains "fps"
not the best approach, but it works TM
"""
with open(self.filepath) as open_file:
reader = csv.reader(open_file, delimiter=self.csv_separator)
found_fps_column = False
for row_number, row_content in enumerate(reader):
if row_number > 100:
# did not find the row that starts with the
# 'fps' column up until here. give up.
break
if "fps" in row_content:
self.skip_lines = row_number + 1
found_fps_column = True
for col, col_name in enumerate(row_content):
if col_name in self.column_name_to_index:
raise Exception("Two columns have the same name")
self.column_name_to_index[col_name] = col
if not found_fps_column:
raise Exception("Not a benchmark file")
def _load_data(self):
r"""
Load the benchmark data into memory.
"""
def extend_columns(new_column_num):
current_row_num = 0
if self.columns:
current_row_num = len(self.columns[0])
assert (all([len(column) == current_row_num for column in self.columns]))
current_column_num = len(self.columns)
if new_column_num >= current_column_num:
self.columns += [["" for j in range(current_row_num)] for i in range(new_column_num - current_column_num)]
# no need to load data if it's already loaded
if self._is_data_loaded:
return
with open(self.filepath) as open_file:
reader = csv.reader(open_file, delimiter=self.csv_separator)
self._is_data_loaded = True
for row_number, row_content in enumerate(reader):
if row_number <= self.skip_lines:
continue
extend_columns(len(row_content))
for col, val in enumerate(row_content):
self.columns[col].append(val)
# Delete any eventual empty column
if all([val == "" for val in self.columns[-1]]):
del self.columns[-1]
def get_column_names(self) -> List[str]:
r"""
Returns the list of columns names of the csv file.
"""
return list(self.column_name_to_index.keys())
def get(self, col: str, data_type: str = "float") \
-> Union[List[float], List[str], List[int], List[complex]]:
r"""
Returns the column `col`.
Parameters
----------
col : str
The desired column name to retrieve, or its index
data_type : str
"string", "integer" or "float", the type to cast
the data to before returning it.
Returns
-------
A list of `data_type` containing the column `col`
"""
if not self._is_data_loaded:
self._load_data()
if len(self.columns) == 0:
raise ValueError("Datafile empty, can't return any data")
data_caster_dict = {
"string": identity,
"float": get_float,
"integer": get_integer
}
if data_type not in data_caster_dict:
raise ValueError("the given `data_type' doesn't match any "
"known types. Which are `string', `integer', "
"`float' or `complex'")
if is_integer(col):
# the column's index is given
return [data_caster_dict[data_type](val) for val in self.columns[col]]
if col in self.column_name_to_index:
# a column name has been given
return [data_caster_dict[data_type](val)
for val in self.columns[self.column_name_to_index[col]]]
raise Exception("Column {} does not exist".format(col))
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description='Plot all the MangoHud benchmarks contained in a given folder.')
parser.add_argument('folder', metavar='folder', nargs=1,
help='path the a MangoHud benchmark folder')
args = parser.parse_args()
bench_folder_path = Path(args.folder[0])
if not bench_folder_path.is_dir():
print(f"The path '{bench_folder_path.absolute()}' "
"does not point to an existing folder")
exit(1)
fps_subdivs = 1.0 # one division every fps_subdivs FPS
y_labels = [] # bench files
x_labels = [] # FPS subidivions
database = Database(bench_folder_path, csv_separator=',')
distributions = []
if len(database.datafiles) == 0:
print(f"The folder \n {bench_folder_path.absolute()} \n"
"contains no CSV file "
"(make sure they have the .csv extension)")
exit(1)
for datafile in database.datafiles:
bar_distribution = []
# sort array to get percentiles
fps_array = np.sort(datafile.get("fps"))
# save percentiles
if len(fps_array) < 10000:
print(f"'{datafile.filename}' simulation "
"isn't long enough for precise statistics")
datafile.set_variable("selected", False)
continue
# Save label only if this file has long enough simulation
y_labels.append(datafile.filename[:-4])
datafile.set_variable("selected", True)
# Save percentiles
datafile.set_variable("0.1%", fps_array[int(float(len(fps_array))*0.001)])
datafile.set_variable("1%", fps_array[int(float(len(fps_array))*0.01)])
datafile.set_variable("50%", fps_array[int(float(len(fps_array))*0.5)])
datafile.set_variable("average fps", np.average(fps_array))
for frame_num, fps in enumerate(fps_array):
if fps > 1000:
print("FPS value above 1000, omitting outlier.")
continue
index = int(fps/fps_subdivs)
for i in range(len(bar_distribution), index+1):
bar_distribution.append(0)
bar_distribution[index] += 1
distributions.append(bar_distribution)
if not distributions:
print("Nothing to plot, exiting.")
exit(1)
num_benchs = len(distributions)
max_size = 0
for distrib in distributions:
max_size = max(max_size, len(distrib))
for distrib in distributions:
for i in range(len(distrib), max_size):
distrib.append(0)
for i in range(max_size):
x_labels.append(str(fps_subdivs * i))
fig, ax = plt.subplots()
# change color of the graph box to the same color as the text
for spine in ['left', 'right', 'bottom', 'top']:
ax.spines[spine].set_color(text_color)
ax.spines[spine].set_linewidth(graphbox_linewidth)
im = ax.imshow(distributions,
aspect="auto",
extent=[0, max_size*fps_subdivs, 0, num_benchs],
cmap=mango_cmap)
# draw thick line that separates each benchmark
for i in range(len(y_labels)+1):
ax.axhline(float(i), color=text_color, lw=graphbox_linewidth)
i = 0
for datafile in database.datafiles:
if datafile.get_variable("selected"):
kwargs = dict(ymin=(num_benchs-i-1+0.15)/num_benchs,
ymax=(num_benchs-i-0.15)/num_benchs,
lw=3)
ax.axvline(datafile.get_variable("0.1%"),
color='#35260f',
label=("0.1%" if i == 0 else None), **kwargs)
ax.axvline(datafile.get_variable("1%"),
color='#6E4503',
label=("1%" if i == 0 else None), **kwargs)
ax.axvline(datafile.get_variable("50%"),
color='#0967BA',
label=("50%" if i == 0 else None), **kwargs)
ax.axvline(datafile.get_variable("average fps"),
color='#003A6E',
label=("Average" if i == 0 else None), **kwargs)
i += 1
ax.tick_params(axis='y', colors=text_color)
ax.tick_params(axis='x', colors=text_color)
ax.set_yticks(np.arange(len(y_labels)-0.5, 0, -1), labels=y_labels)
ax.grid(False)
fig.set_facecolor(background_color)
ax.ticklabel_format(axis='x', style='plain')
formatter0 = EngFormatter(unit='FPS')
ax.xaxis.set_major_formatter(formatter0)
plt.tight_layout()
plt.legend(facecolor=legend_facecolor, labelcolor=legend_textcolor)
cursor = Cursor(ax,
horizOn=False,
color='#6c49abff',
linewidth=4,
useblit=True)
plt.show()