aboutsummaryrefslogtreecommitdiffstats
path: root/resources/tools/robot_output_parser_publish.py
diff options
context:
space:
mode:
authorpmikus <pmikus@cisco.com>2017-02-01 18:54:58 +0100
committerPeter Mikus <pmikus@cisco.com>2017-02-03 20:40:09 +0000
commit35e5a327a66c05a3356d214dcfdece59ca725998 (patch)
treea111bb1d5b9d40f358655187d5bc4a3b899b0d06 /resources/tools/robot_output_parser_publish.py
parent6baa0bc9921bc13b1adff113c20a4db766c3feba (diff)
CSIT 1701 report files and script AD1
CSIT 1701 report files and script addendum 1 Edits to correct and align all Overview sub-sections. Updates in Performance CSIT Release Notes - added more NDR and PDR performance changes. Change-Id: I52b6ee89e9c536fb4ab9d30dc27cca8dbdd88a20 Signed-off-by: pmikus <pmikus@cisco.com> Signed-off-by: Maciek Konstantynowicz <mkonstan@cisco.com> Signed-off-by: pmikus <pmikus@cisco.com>
Diffstat (limited to 'resources/tools/robot_output_parser_publish.py')
-rwxr-xr-xresources/tools/robot_output_parser_publish.py406
1 files changed, 0 insertions, 406 deletions
diff --git a/resources/tools/robot_output_parser_publish.py b/resources/tools/robot_output_parser_publish.py
deleted file mode 100755
index e2bcfa2f0d..0000000000
--- a/resources/tools/robot_output_parser_publish.py
+++ /dev/null
@@ -1,406 +0,0 @@
-#!/usr/bin/python
-
-# Copyright (c) 2017 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.
-
-"""
-Script extracts interested data (name, documentation, message, status) from
-robot framework output file (output.xml) and prints in specified format (wiki,
-html, rst) to defined output file.
-
-Supported formats:
- - html
- - rst
-
-:TODO:
- - wiki
- - md
-
-:Example:
-
-robot_output_parser_publish.py -i output.xml" -o "tests.rst" -f "rst" -s 3 -l 2
-
-The example reads the data from "output.xml", writes the output to "tests.rst"
-in rst format. It will start on the 3rd level of xml structure and the generated
-document hierarchy will start on the 2nd level.
-"""
-
-import argparse
-import re
-import sys
-import json
-import string
-
-from robot.api import ExecutionResult, ResultVisitor
-
-
-class ExecutionChecker(ResultVisitor):
- """Class to traverse through the test suite structure.
-
- The functionality implemented in this class generates a json file. Its
- structure is:
-
- [
- {
- "level": "Level of the suite, type: str",
- "title": "Title of the suite, type: str",
- "doc": "Documentation of the suite, type: str",
- "table": [
- ["TC name", "TC doc", "message or status"],
- ["TC name", "TC doc", "message or status"],
- ... other test cases ...
- ["Name", "Documentation", "Message or Status"]
- ]
- },
- ... other test suites ...
- ]
-
- .. note:: The header of the table with TCs is at the and of the table.
- """
-
- def __init__(self, args):
- self.formatting = args.formatting
-
- def visit_suite(self, suite):
- """Implements traversing through the suite and its direct children.
-
- :param suite: Suite to process.
- :type suite: Suite
- :returns: Nothing.
- """
-
- if self.start_suite(suite) is not False:
- if suite.tests:
- sys.stdout.write(',"tests":[')
- else:
- sys.stdout.write('},')
-
- suite.suites.visit(self)
- suite.tests.visit(self)
-
- if suite.tests:
- if "ndrdisc" in suite.longname.lower():
- hdr = '["Name","Documentation","Message"]'
- else:
- hdr = '["Name","Documentation","Status"]'
- sys.stdout.write(hdr + ']},')
-
- self.end_suite(suite)
-
- def start_suite(self, suite):
- """Called when suite starts.
-
- :param suite: Suite to process.
- :type suite: Suite
- :returns: Nothing.
- """
-
- level = len(suite.longname.split("."))
- sys.stdout.write('{')
- sys.stdout.write('"level":"' + str(level) + '",')
- sys.stdout.write('"title":"' + suite.name.replace('"', "'") + '",')
- sys.stdout.write('"doc":"' + suite.doc.replace('"', "'").
- replace('\n', ' ').replace('\r', '').
- replace('*[', ' |br| *[') + '"')
-
- def end_suite(self, suite):
- """Called when suite ends.
-
- :param suite: Suite to process.
- :type suite: Suite
- :returns: Nothing.
- """
- pass
-
- def visit_test(self, test):
- """Implements traversing through the test.
-
- :param test: Test to process.
- :type test: Test
- :returns: Nothing.
- """
- if self.start_test(test) is not False:
- self.end_test(test)
-
- def start_test(self, test):
- """Called when test starts.
-
- :param test: Test to process.
- :type test: Test
- :returns: Nothing.
- """
-
- name = test.name.replace('"', "'")
- doc = test.doc.replace('"', "'").replace('\n', ' ').replace('\r', '').\
- replace('[', ' |br| [')
- if any("NDRPDRDISC" in tag for tag in test.tags):
- msg = test.message.replace('\n', ' |br| ').replace('\r', ''). \
- replace('"', "'")
-
- sys.stdout.write('["' + name + '","' + doc + '","' + msg + '"]')
- else:
- sys.stdout.write(
- '["' + name + '","' + doc + '","' + test.status + '"]')
-
- def end_test(self, test):
- """Called when test ends.
-
- :param test: Test to process.
- :type test: Test
- :returns: Nothing.
- """
- sys.stdout.write(',')
-
-
-def do_html(data, args):
- """Generation of a html file from json data.
-
- :param data: List of suites from json file.
- :param args: Parsed arguments.
- :type data: list of dict
- :type args: ArgumentParser
- :returns: Nothing.
- """
-
- shift = int(args.level)
- start = int(args.start)
-
- output = open(args.output, 'w')
-
- output.write('<html>')
- for item in data:
- if int(item['level']) < start:
- continue
- level = str(int(item['level']) - start + shift)
- output.write('<h' + level + '>' + item['title'].lower() +
- '</h' + level + '>')
- output.write('<p>' + re.sub(r"(\*)(.*?)(\*)", r"<b>\2</b>", item['doc'],
- 0, flags=re.MULTILINE).
- replace(' |br| ', '<br>') + '</p>')
- try:
- output.write(gen_html_table(item['tests']))
- except KeyError:
- continue
- output.write('</html>')
- output.close()
-
-
-def gen_html_table(data):
- """Generates a table with TCs' names, documentation and messages / statuses
- in html format. There is no css used.
-
- :param data: Json data representing a table with TCs.
- :type data: str
- :returns: Table with TCs' names, documentation and messages / statuses in
- html format.
- :rtype: str
- """
-
- table = '<table width=100% border=1><tr>'
- table += '<th width=30%>Name</th>'
- table += '<th width=50%>Documentation</th>'
- table += '<th width=20%>Status</th></tr>'
-
- for item in data[0:-2]:
- table += '<tr>'
- for element in item:
- table += '<td>' + element.replace(' |br| ', '<br>') + '</td>'
- table += '</tr></table>'
-
- return table
-
-
-def do_rst(data, args):
- """Generation of a rst file from json data.
-
- :param data: List of suites from json file.
- :param args: Parsed arguments.
- :type data: list of dict
- :type args: ArgumentParser
- :returns: Nothing.
- """
-
- hdrs = ['=', '-', '`', "'", '.', '~', '*', '+', '^']
- shift = int(args.level)
- start = int(args.start)
-
- output = open(args.output, 'w')
- output.write('\n.. |br| raw:: html\n\n <br />\n\n')
-
- for item in data:
- if int(item['level']) < start:
- continue
- if 'ndrchk' in item['title'].lower():
- continue
- output.write(item['title'].lower() + '\n' +
- hdrs[int(item['level']) - start + shift] *
- len(item['title']) + '\n\n')
- output.write(item['doc'].replace('*', '**').replace('|br|', '\n\n -') +
- '\n\n')
- try:
- output.write(gen_rst_table(item['tests']) + '\n\n')
- except KeyError:
- continue
- output.close()
-
-
-def gen_rst_table(data):
- """Generates a table with TCs' names, documentation and messages / statuses
- in rst format.
-
- :param data: Json data representing a table with TCs.
- :type data: str
- :returns: Table with TCs' names, documentation and messages / statuses in
- rst format.
- :rtype: str
- """
-
- table = []
- # max size of each column
- lengths = map(max, zip(*[[len(str(elt)) for elt in item] for item in data]))
-
- start_of_line = '| '
- vert_separator = ' | '
- end_of_line = ' |'
- line_marker = '-'
-
- meta_template = vert_separator.join(['{{{{{0}:{{{0}}}}}}}'.format(i)
- for i in range(len(lengths))])
- template = '{0}{1}{2}'.format(start_of_line, meta_template.format(*lengths),
- end_of_line)
- # determine top/bottom borders
- to_separator = string.maketrans('| ', '+-')
- start_of_line = start_of_line.translate(to_separator)
- vert_separator = vert_separator.translate(to_separator)
- end_of_line = end_of_line.translate(to_separator)
- separator = '{0}{1}{2}'.format(start_of_line, vert_separator.
- join([x * line_marker for x in lengths]),
- end_of_line)
- # determine header separator
- th_separator_tr = string.maketrans('-', '=')
- start_of_line = start_of_line.translate(th_separator_tr)
- line_marker = line_marker.translate(th_separator_tr)
- vertical_separator = vert_separator.translate(th_separator_tr)
- end_of_line = end_of_line.translate(th_separator_tr)
- th_separator = '{0}{1}{2}'.format(start_of_line, vertical_separator.
- join([x * line_marker for x in lengths]),
- end_of_line)
- # prepare table
- table.append(separator)
- # set table header
- titles = data[-1]
- table.append(template.format(*titles))
- table.append(th_separator)
- # generate table rows
- for d in data[0:-2]:
- table.append(template.format(*d))
- table.append(separator)
- table.append(template.format(*data[-2]))
- table.append(separator)
- return '\n'.join(table)
-
-
-def do_md(data, args):
- """Generation of a rst file from json data.
-
- :param data: List of suites from json file.
- :param args: Parsed arguments.
- :type data: list of dict
- :type args: ArgumentParser
- :returns: Nothing.
- """
- raise NotImplementedError("Export to 'md' format is not implemented.")
-
-
-def do_wiki(data, args):
- """Generation of a wiki page from json data.
-
- :param data: List of suites from json file.
- :param args: Parsed arguments.
- :type data: list of dict
- :type args: ArgumentParser
- :returns: Nothing.
- """
- raise NotImplementedError("Export to 'wiki' format is not implemented.")
-
-
-def process_robot_file(args):
- """Process data from robot output.xml file and generate defined file type.
-
- :param args: Parsed arguments.
- :type args: ArgumentParser
- :return: Nothing.
- """
-
- old_sys_stdout = sys.stdout
- sys.stdout = open(args.output + '.json', 'w')
-
- result = ExecutionResult(args.input)
- checker = ExecutionChecker(args)
-
- sys.stdout.write('[')
- result.visit(checker)
- sys.stdout.write('{}]')
- sys.stdout.close()
- sys.stdout = old_sys_stdout
-
- with open(args.output + '.json', 'r') as json_file:
- data = json.load(json_file)
- data.pop(-1)
-
- if args.formatting == 'rst':
- do_rst(data, args)
- elif args.formatting == 'wiki':
- do_wiki(data, args)
- elif args.formatting == 'html':
- do_html(data, args)
- elif args.formatting == 'md':
- do_md(data, args)
-
-
-def parse_args():
- """Parse arguments from cmd line.
-
- :return: Parsed arguments.
- :rtype ArgumentParser
- """
-
- parser = argparse.ArgumentParser(description=__doc__,
- formatter_class=argparse.
- RawDescriptionHelpFormatter)
- parser.add_argument("-i", "--input",
- required=True,
- type=argparse.FileType('r'),
- help="Robot XML log file")
- parser.add_argument("-o", "--output",
- type=str,
- required=True,
- help="Output file")
- parser.add_argument("-f", "--formatting",
- required=True,
- choices=['html', 'wiki', 'rst', 'md'],
- help="Output file format")
- parser.add_argument("-s", "--start",
- type=int,
- default=1,
- help="The first level to be taken from xml file")
- parser.add_argument("-l", "--level",
- type=int,
- default=1,
- help="The level of the first chapter in generated file")
-
- return parser.parse_args()
-
-
-if __name__ == "__main__":
- sys.exit(process_robot_file(parse_args()))