diff options
author | Viliam Luc <vluc@cisco.com> | 2021-12-17 15:26:36 +0100 |
---|---|---|
committer | Tibor Frank <tifrank@cisco.com> | 2022-01-18 09:14:49 +0000 |
commit | 18f2763797ff782f9068550fc2f01cad4d7056af (patch) | |
tree | cb19545a4217ec2190296a3877feae0f59a34463 | |
parent | 5f75d8f5ac0c4c7316dafd75c372b9f6c40aaae6 (diff) |
trending: new view in regressions and progressions
Signed-off-by: Viliam Luc <vluc@cisco.com>
Change-Id: I8524319a215ff551cf67c30d0b08ddae69883f61
-rw-r--r-- | resources/tools/presentation/generator_alerts.py | 10 | ||||
-rw-r--r-- | resources/tools/presentation/generator_cpta.py | 185 | ||||
-rw-r--r-- | resources/tools/presentation/generator_tables.py | 4 |
3 files changed, 145 insertions, 54 deletions
diff --git a/resources/tools/presentation/generator_alerts.py b/resources/tools/presentation/generator_alerts.py index c3bf81d7ff..807f1b035c 100644 --- a/resources/tools/presentation/generator_alerts.py +++ b/resources/tools/presentation/generator_alerts.py @@ -363,7 +363,7 @@ class Alerting: :param idx: Index of the test set as it is specified in the specification file. :param header: The header of the list of [re|pro]gressions. - :param re_pro: 'regression' or 'progression'. + :param re_pro: 'regressions' or 'progressions'. :type alert: dict :type idx: int :type header: str @@ -408,8 +408,8 @@ class Alerting: text = u"" - legend = (f"Legend:\n[ Last trend in Mpps | number of runs for " - f"last trend |") + legend = (f"Legend: Test-name NIC Frame-size Trend[Mpps] Runs[#] " + f"Long-Term change[%]") out_file = ( f"{self.configs[alert[u'way']][u'output-dir']}/" @@ -417,7 +417,7 @@ class Alerting: ) try: with open(out_file, u'w') as reg_file: - reg_file.write(f"{legend} regressions ]") + reg_file.write(legend) except IOError: logging.error(f"Not possible to write the file {out_file}.txt.") @@ -427,7 +427,7 @@ class Alerting: ) try: with open(out_file, u'w') as reg_file: - reg_file.write(f"{legend} progressions ]") + reg_file.write(legend) except IOError: logging.error(f"Not possible to write the file {out_file}.txt.") diff --git a/resources/tools/presentation/generator_cpta.py b/resources/tools/presentation/generator_cpta.py index 1a2dbaa124..fafa8638a4 100644 --- a/resources/tools/presentation/generator_cpta.py +++ b/resources/tools/presentation/generator_cpta.py @@ -1,4 +1,4 @@ -# Copyright (c) 2021 Cisco and/or its affiliates. +# Copyright (c) 2022 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: @@ -838,7 +838,44 @@ def _generate_all_charts(spec, input_data): # Evaluate result: if anomaly_classifications: + test_reg_lst = [] + nic_reg_lst = [] + frmsize_reg_lst = [] + trend_reg_lst = [] + number_reg_lst = [] + ltc_reg_lst = [] + test_prog_lst = [] + nic_prog_lst = [] + frmsize_prog_lst = [] + trend_prog_lst = [] + number_prog_lst = [] + ltc_prog_lst = [] result = u"PASS" + + class MaxLens(): + """Class to store the max lengths of strings displayed in + regressions and progressions. + """ + + def __init__(self, tst, nic, frmsize, trend, run, ltc): + """Initialisation. + + :param tst: Name of the test. + :param nic: NIC used in the test. + :param frmsize: Frame size used in the test. + :param trend: Trend Change. + :param run: Number of runs for last trend. + :param ltc: Regression or Progression + """ + self.tst = tst + self.nic = nic + self.frmsize = frmsize + self.trend = trend + self.run = run + self.ltc = ltc + + max_len = MaxLens(0, 0, 0, 0, 0, 0) + for job_name, job_data in anomaly_classifications.items(): data = [] tb = u"-".join(job_name.split(u"-")[-2:]) @@ -848,57 +885,111 @@ def _generate_all_charts(spec, input_data): file_to_read = f"{spec.cpta[u'output-file']}/{file}" with open(f"{file_to_read}", u"rt") as input: data = data + input.readlines() - file_name = \ - f"{spec.cpta[u'output-file']}/regressions-{job_name}.txt" - with open(file_name, u'w') as txt_file: - for test_name, classification in job_data.items(): - if classification == u"regression": - if u"2n" in test_name: - test_name = test_name.split("-", 2) - tst = test_name[2].split(".")[-1] - nic = test_name[1] - tst_name = f"{nic}-{tst}" - else: - test_name = test_name.split("-", 1) - tst = test_name[1].split(".")[-1] - nic = test_name[0].split(".")[-1] - tst_name = f"{nic}-{tst}" - - for line in data: - if tst_name in line: - line = line.replace(" ", "") - trend = line.split("|")[2] - number = line.split("|")[3] - ltc = line.split("|")[4] - txt_file.write(f"{tst_name} [ {trend}M | " - f"#{number} | {ltc}% ]\n") + + for test_name, classification in job_data.items(): + if classification != u"normal": + if u"2n" in test_name: + test_name = test_name.split("-", 2) + tst = test_name[2].split(".")[-1] + nic = test_name[1] + else: + test_name = test_name.split("-", 1) + tst = test_name[1].split(".")[-1] + nic = test_name[0].split(".")[-1] + frmsize = tst.split("-")[0].upper() + tst = u"-".join(tst.split("-")[1:]) + tst_name = f"{nic}-{frmsize}-{tst}" + if len(tst) > max_len.tst: + max_len.tst = len(tst) + if len(nic) > max_len.nic: + max_len.nic = len(nic) + if len(frmsize) > max_len.frmsize: + max_len.frmsize = len(frmsize) + + for line in data: + if tst_name in line: + line = line.replace(" ", "") + trend = line.split("|")[2] + if len(str(trend)) > max_len.trend: + max_len.trend = len(str(trend)) + number = line.split("|")[3] + if len(str(number)) > max_len.run: + max_len.run = len(str(number)) + ltc = line.split("|")[4] + if len(str(ltc)) > max_len.ltc: + max_len.ltc = len(str(ltc)) + if classification == u'regression': + test_reg_lst.append(tst) + nic_reg_lst.append(nic) + frmsize_reg_lst.append(frmsize) + trend_reg_lst.append(trend) + number_reg_lst.append(number) + ltc_reg_lst.append(ltc) + elif classification == u'progression': + test_prog_lst.append(tst) + nic_prog_lst.append(nic) + frmsize_prog_lst.append(frmsize) + trend_prog_lst.append(trend) + number_prog_lst.append(number) + ltc_prog_lst.append(ltc) if classification in (u"regression", u"outlier"): result = u"FAIL" + + text = u"" + for idx in range(len(test_reg_lst)): + text += ( + f"{test_reg_lst[idx]}" + f"{u' ' * (max_len.tst - len(test_reg_lst[idx]))} " + f"{nic_reg_lst[idx]}" + f"{u' ' * (max_len.nic - len(nic_reg_lst[idx]))} " + f"{frmsize_reg_lst[idx]}" + f"{u' ' * (max_len.frmsize - len(frmsize_reg_lst[idx]))} " + f"{trend_reg_lst[idx]}" + f"{u' ' * (max_len.trend - len(str(trend_reg_lst[idx])))} " + f"{number_reg_lst[idx]}" + f"{u' ' * (max_len.run - len(str(number_reg_lst[idx])))} " + f"{ltc_reg_lst[idx]}" + f"{u' ' * (max_len.ltc - len(str(ltc_reg_lst[idx])))} " + f"\n" + ) + + file_name = \ + f"{spec.cpta[u'output-file']}/regressions-{job_name}.txt" + + try: + with open(f"{file_name}", u'w') as txt_file: + txt_file.write(text) + except IOError: + logging.error( + f"Not possible to write the file {file_name}.") + + text = u"" + for idx in range(len(test_prog_lst)): + text += ( + f"{test_prog_lst[idx]}" + f"{u' ' * (max_len.tst - len(test_prog_lst[idx]))} " + f"{nic_prog_lst[idx]}" + f"{u' ' * (max_len.nic - len(nic_prog_lst[idx]))} " + f"{frmsize_prog_lst[idx]}" + f"{u' ' * (max_len.frmsize - len(frmsize_prog_lst[idx]))} " + f"{trend_prog_lst[idx]}" + f"{u' ' * (max_len.trend -len(str(trend_prog_lst[idx])))} " + f"{number_prog_lst[idx]}" + f"{u' ' * (max_len.run - len(str(number_prog_lst[idx])))} " + f"{ltc_prog_lst[idx]}" + f"{u' ' * (max_len.ltc - len(str(ltc_prog_lst[idx])))} " + f"\n" + ) + file_name = \ f"{spec.cpta[u'output-file']}/progressions-{job_name}.txt" - with open(file_name, u'w') as txt_file: - for test_name, classification in job_data.items(): - if classification == u"progression": - if u"2n" in test_name: - test_name = test_name.split("-", 2) - tst = test_name[2].split(".")[-1] - nic = test_name[1] - tst_name = f"{nic}-{tst}" - else: - test_name = test_name.split("-", 1) - tst = test_name[1].split(".")[-1] - nic = test_name[0].split(".")[-1] - tst_name = f"{nic}-{tst}" - - for line in data: - if tst_name in line: - line = line.replace(" ", "") - trend = line.split("|")[2] - number = line.split("|")[3] - ltc = line.split("|")[4] - txt_file.write(f"{tst_name} [ {trend}M | " - f"#{number} | {ltc}% ]\n") + try: + with open(f"{file_name}", u'w') as txt_file: + txt_file.write(text) + except IOError: + logging.error(f"Not possible to write the file {file_name}.") + else: result = u"FAIL" diff --git a/resources/tools/presentation/generator_tables.py b/resources/tools/presentation/generator_tables.py index c36159825b..2e7d3dee36 100644 --- a/resources/tools/presentation/generator_tables.py +++ b/resources/tools/presentation/generator_tables.py @@ -959,8 +959,8 @@ def table_perf_trending_dash(table, input_data): header = [ u"Test Case", u"Trend [Mpps]", - u"Number of runs [#]", - u"Trend Change [%]", + u"Runs [#]", + u"Long-Term Change [%]", u"Regressions [#]", u"Progressions [#]" ] |