aboutsummaryrefslogtreecommitdiffstats
path: root/resources/tools/presentation/new/generator_plots.py
diff options
context:
space:
mode:
Diffstat (limited to 'resources/tools/presentation/new/generator_plots.py')
-rw-r--r--resources/tools/presentation/new/generator_plots.py399
1 files changed, 399 insertions, 0 deletions
diff --git a/resources/tools/presentation/new/generator_plots.py b/resources/tools/presentation/new/generator_plots.py
new file mode 100644
index 0000000000..aaee31f53b
--- /dev/null
+++ b/resources/tools/presentation/new/generator_plots.py
@@ -0,0 +1,399 @@
+# Copyright (c) 2018 Cisco and/or its affiliates.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Algorithms to generate plots.
+"""
+
+
+import logging
+import pandas as pd
+import plotly.offline as ploff
+import plotly.graph_objs as plgo
+
+from plotly.exceptions import PlotlyError
+
+from utils import mean
+
+
+def generate_plots(spec, data):
+ """Generate all plots specified in the specification file.
+
+ :param spec: Specification read from the specification file.
+ :param data: Data to process.
+ :type spec: Specification
+ :type data: InputData
+ """
+
+ logging.info("Generating the plots ...")
+ for index, plot in enumerate(spec.plots):
+ try:
+ logging.info(" Plot nr {0}:".format(index + 1))
+ eval(plot["algorithm"])(plot, data)
+ except NameError as err:
+ logging.error("Probably algorithm '{alg}' is not defined: {err}".
+ format(alg=plot["algorithm"], err=repr(err)))
+ logging.info("Done.")
+
+
+def plot_performance_box(plot, input_data):
+ """Generate the plot(s) with algorithm: plot_performance_box
+ specified in the specification file.
+
+ :param plot: Plot to generate.
+ :param input_data: Data to process.
+ :type plot: pandas.Series
+ :type input_data: InputData
+ """
+
+ logging.info(" Generating the plot {0} ...".
+ format(plot.get("title", "")))
+
+ # Transform the data
+ logging.info(" Creating the data set for the {0} '{1}'.".
+ format(plot.get("type", ""), plot.get("title", "")))
+ data = input_data.filter_data(plot)
+ if data is None:
+ logging.error("No data.")
+ return
+
+ # Prepare the data for the plot
+ y_vals = dict()
+ for job in data:
+ for build in job:
+ for test in build:
+ if y_vals.get(test["parent"], None) is None:
+ y_vals[test["parent"]] = list()
+ try:
+ y_vals[test["parent"]].append(test["throughput"]["value"])
+ except (KeyError, TypeError):
+ y_vals[test["parent"]].append(None)
+
+ # Add None to the lists with missing data
+ max_len = 0
+ for val in y_vals.values():
+ if len(val) > max_len:
+ max_len = len(val)
+ for key, val in y_vals.items():
+ if len(val) < max_len:
+ val.extend([None for _ in range(max_len - len(val))])
+
+ # Add plot traces
+ traces = list()
+ df = pd.DataFrame(y_vals)
+ df.head()
+ for i, col in enumerate(df.columns):
+ name = "{0}. {1}".format(i + 1, col.lower().replace('-ndrpdrdisc', ''))
+ traces.append(plgo.Box(x=[str(i + 1) + '.'] * len(df[col]),
+ y=df[col],
+ name=name,
+ **plot["traces"]))
+
+ try:
+ # Create plot
+ plpl = plgo.Figure(data=traces, layout=plot["layout"])
+
+ # Export Plot
+ logging.info(" Writing file '{0}{1}'.".
+ format(plot["output-file"], plot["output-file-type"]))
+ ploff.plot(plpl,
+ show_link=False, auto_open=False,
+ filename='{0}{1}'.format(plot["output-file"],
+ plot["output-file-type"]))
+ except PlotlyError as err:
+ logging.error(" Finished with error: {}".
+ format(str(err).replace("\n", " ")))
+ return
+
+ logging.info(" Done.")
+
+
+def plot_latency_box(plot, input_data):
+ """Generate the plot(s) with algorithm: plot_latency_box
+ specified in the specification file.
+
+ :param plot: Plot to generate.
+ :param input_data: Data to process.
+ :type plot: pandas.Series
+ :type input_data: InputData
+ """
+
+ logging.info(" Generating the plot {0} ...".
+ format(plot.get("title", "")))
+
+ # Transform the data
+ logging.info(" Creating the data set for the {0} '{1}'.".
+ format(plot.get("type", ""), plot.get("title", "")))
+ data = input_data.filter_data(plot)
+ if data is None:
+ logging.error("No data.")
+ return
+
+ # Prepare the data for the plot
+ y_tmp_vals = dict()
+ for job in data:
+ for build in job:
+ for test in build:
+ if y_tmp_vals.get(test["parent"], None) is None:
+ y_tmp_vals[test["parent"]] = [
+ list(), # direction1, min
+ list(), # direction1, avg
+ list(), # direction1, max
+ list(), # direction2, min
+ list(), # direction2, avg
+ list() # direction2, max
+ ]
+ try:
+ y_tmp_vals[test["parent"]][0].append(
+ test["latency"]["direction1"]["50"]["min"])
+ y_tmp_vals[test["parent"]][1].append(
+ test["latency"]["direction1"]["50"]["avg"])
+ y_tmp_vals[test["parent"]][2].append(
+ test["latency"]["direction1"]["50"]["max"])
+ y_tmp_vals[test["parent"]][3].append(
+ test["latency"]["direction2"]["50"]["min"])
+ y_tmp_vals[test["parent"]][4].append(
+ test["latency"]["direction2"]["50"]["avg"])
+ y_tmp_vals[test["parent"]][5].append(
+ test["latency"]["direction2"]["50"]["max"])
+ except (KeyError, TypeError):
+ pass
+
+ y_vals = dict()
+ for key, values in y_tmp_vals.items():
+ y_vals[key] = list()
+ for val in values:
+ if val:
+ average = mean(val)
+ else:
+ average = None
+ y_vals[key].append(average)
+ y_vals[key].append(average) # Twice for plot.ly
+
+ # Add plot traces
+ traces = list()
+ try:
+ df = pd.DataFrame(y_vals)
+ df.head()
+ except ValueError as err:
+ logging.error(" Finished with error: {}".
+ format(str(err).replace("\n", " ")))
+ return
+
+ for i, col in enumerate(df.columns):
+ name = "{0}. {1}".format(i + 1, col.lower().replace('-ndrpdrdisc', ''))
+ traces.append(plgo.Box(x=['TGint1-to-SUT1-to-SUT2-to-TGint2',
+ 'TGint1-to-SUT1-to-SUT2-to-TGint2',
+ 'TGint1-to-SUT1-to-SUT2-to-TGint2',
+ 'TGint1-to-SUT1-to-SUT2-to-TGint2',
+ 'TGint1-to-SUT1-to-SUT2-to-TGint2',
+ 'TGint1-to-SUT1-to-SUT2-to-TGint2',
+ 'TGint2-to-SUT2-to-SUT1-to-TGint1',
+ 'TGint2-to-SUT2-to-SUT1-to-TGint1',
+ 'TGint2-to-SUT2-to-SUT1-to-TGint1',
+ 'TGint2-to-SUT2-to-SUT1-to-TGint1',
+ 'TGint2-to-SUT2-to-SUT1-to-TGint1',
+ 'TGint2-to-SUT2-to-SUT1-to-TGint1'],
+ y=df[col],
+ name=name,
+ **plot["traces"]))
+
+ try:
+ # Create plot
+ logging.info(" Writing file '{0}{1}'.".
+ format(plot["output-file"], plot["output-file-type"]))
+ plpl = plgo.Figure(data=traces, layout=plot["layout"])
+
+ # Export Plot
+ ploff.plot(plpl,
+ show_link=False, auto_open=False,
+ filename='{0}{1}'.format(plot["output-file"],
+ plot["output-file-type"]))
+ except PlotlyError as err:
+ logging.error(" Finished with error: {}".
+ format(str(err).replace("\n", " ")))
+ return
+
+ logging.info(" Done.")
+
+
+def plot_throughput_speedup_analysis(plot, input_data):
+ """Generate the plot(s) with algorithm: plot_throughput_speedup_analysis
+ specified in the specification file.
+
+ :param plot: Plot to generate.
+ :param input_data: Data to process.
+ :type plot: pandas.Series
+ :type input_data: InputData
+ """
+
+ logging.info(" Generating the plot {0} ...".
+ format(plot.get("title", "")))
+
+ # Transform the data
+ logging.info(" Creating the data set for the {0} '{1}'.".
+ format(plot.get("type", ""), plot.get("title", "")))
+ data = input_data.filter_data(plot)
+ if data is None:
+ logging.error("No data.")
+ return
+
+ throughput = dict()
+ for job in data:
+ for build in job:
+ for test in build:
+ if throughput.get(test["parent"], None) is None:
+ throughput[test["parent"]] = {"1": list(),
+ "2": list(),
+ "4": list()}
+ try:
+ if "1T1C" in test["tags"]:
+ throughput[test["parent"]]["1"].\
+ append(test["throughput"]["value"])
+ elif "2T2C" in test["tags"]:
+ throughput[test["parent"]]["2"]. \
+ append(test["throughput"]["value"])
+ elif "4T4C" in test["tags"]:
+ throughput[test["parent"]]["4"]. \
+ append(test["throughput"]["value"])
+ except (KeyError, TypeError):
+ pass
+
+ if not throughput:
+ logging.warning("No data for the plot '{}'".
+ format(plot.get("title", "")))
+ return
+
+ for test_name, test_vals in throughput.items():
+ for key, test_val in test_vals.items():
+ if test_val:
+ throughput[test_name][key] = sum(test_val) / len(test_val)
+
+ names = ['1 core', '2 cores', '4 cores']
+ x_vals = list()
+ y_vals_1 = list()
+ y_vals_2 = list()
+ y_vals_4 = list()
+
+ for test_name, test_vals in throughput.items():
+ if test_vals["1"]:
+ x_vals.append("-".join(test_name.split('-')[1:-1]))
+ y_vals_1.append(1)
+ if test_vals["2"]:
+ y_vals_2.append(
+ round(float(test_vals["2"]) / float(test_vals["1"]), 2))
+ else:
+ y_vals_2.append(None)
+ if test_vals["4"]:
+ y_vals_4.append(
+ round(float(test_vals["4"]) / float(test_vals["1"]), 2))
+ else:
+ y_vals_4.append(None)
+
+ y_vals = [y_vals_1, y_vals_2, y_vals_4]
+
+ y_vals_zipped = zip(names, y_vals)
+ traces = list()
+ for val in y_vals_zipped:
+ traces.append(plgo.Bar(x=x_vals,
+ y=val[1],
+ name=val[0]))
+
+ try:
+ # Create plot
+ logging.info(" Writing file '{0}{1}'.".
+ format(plot["output-file"], plot["output-file-type"]))
+ plpl = plgo.Figure(data=traces, layout=plot["layout"])
+
+ # Export Plot
+ ploff.plot(plpl,
+ show_link=False, auto_open=False,
+ filename='{0}{1}'.format(plot["output-file"],
+ plot["output-file-type"]))
+ except PlotlyError as err:
+ logging.error(" Finished with error: {}".
+ format(str(err).replace("\n", " ")))
+ return
+
+ logging.info(" Done.")
+
+
+def plot_http_server_performance_box(plot, input_data):
+ """Generate the plot(s) with algorithm: plot_http_server_performance_box
+ specified in the specification file.
+
+ :param plot: Plot to generate.
+ :param input_data: Data to process.
+ :type plot: pandas.Series
+ :type input_data: InputData
+ """
+
+ logging.info(" Generating the plot {0} ...".
+ format(plot.get("title", "")))
+
+ # Transform the data
+ logging.info(" Creating the data set for the {0} '{1}'.".
+ format(plot.get("type", ""), plot.get("title", "")))
+ data = input_data.filter_data(plot)
+ if data is None:
+ logging.error("No data.")
+ return
+
+ # Prepare the data for the plot
+ y_vals = dict()
+ for job in data:
+ for build in job:
+ for test in build:
+ if y_vals.get(test["name"], None) is None:
+ y_vals[test["name"]] = list()
+ try:
+ y_vals[test["name"]].append(test["result"]["value"])
+ except (KeyError, TypeError):
+ y_vals[test["name"]].append(None)
+
+ # Add None to the lists with missing data
+ max_len = 0
+ for val in y_vals.values():
+ if len(val) > max_len:
+ max_len = len(val)
+ for key, val in y_vals.items():
+ if len(val) < max_len:
+ val.extend([None for _ in range(max_len - len(val))])
+
+ # Add plot traces
+ traces = list()
+ df = pd.DataFrame(y_vals)
+ df.head()
+ for i, col in enumerate(df.columns):
+ name = "{0}. {1}".format(i + 1, col.lower().replace('-cps', '').
+ replace('-rps', ''))
+ traces.append(plgo.Box(x=[str(i + 1) + '.'] * len(df[col]),
+ y=df[col],
+ name=name,
+ **plot["traces"]))
+ try:
+ # Create plot
+ plpl = plgo.Figure(data=traces, layout=plot["layout"])
+
+ # Export Plot
+ logging.info(" Writing file '{0}{1}'.".
+ format(plot["output-file"], plot["output-file-type"]))
+ ploff.plot(plpl,
+ show_link=False, auto_open=False,
+ filename='{0}{1}'.format(plot["output-file"],
+ plot["output-file-type"]))
+ except PlotlyError as err:
+ logging.error(" Finished with error: {}".
+ format(str(err).replace("\n", " ")))
+ return
+
+ logging.info(" Done.")