aboutsummaryrefslogtreecommitdiffstats
path: root/resources/tools/presentation/generator_files.py
diff options
context:
space:
mode:
Diffstat (limited to 'resources/tools/presentation/generator_files.py')
-rw-r--r--resources/tools/presentation/generator_files.py436
1 files changed, 0 insertions, 436 deletions
diff --git a/resources/tools/presentation/generator_files.py b/resources/tools/presentation/generator_files.py
deleted file mode 100644
index ef06cad519..0000000000
--- a/resources/tools/presentation/generator_files.py
+++ /dev/null
@@ -1,436 +0,0 @@
-# Copyright (c) 2023 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 files.
-"""
-
-import re
-
-from os.path import join
-from collections import OrderedDict
-
-import logging
-
-from pal_utils import get_files, get_rst_title_char
-
-
-RST_INCLUDE_TABLE = (u"\n.. only:: html\n\n"
- u" .. csv-table::\n"
- u" :header-rows: 1\n"
- u" :widths: auto\n"
- u" :align: center\n"
- u" :file: {file_html}\n"
- u"\n.. only:: latex\n\n"
- u"\n .. raw:: latex\n\n"
- u" \\csvautolongtable{{{file_latex}}}\n\n")
-
-REGEX_NIC_SHORT = re.compile(r'(\d*ge\dp\d)([a-z]*\d*[a-z]*)-')
-
-
-def generate_files(spec, data):
- """Generate all files specified in the specification file.
-
- :param spec: Specification read from the specification file.
- :param data: Data to process.
- :type spec: Specification
- :type data: InputData
- """
-
- generator = {
- u"file_details_split": file_details_split,
- u"file_details_split_html": file_details_split_html,
- u"file_test_results": file_test_results,
- u"file_test_results_html": file_test_results_html
- }
-
- logging.info(u"Generating the files ...")
- for file_spec in spec.files:
- try:
- generator[file_spec[u"algorithm"]](file_spec, data)
- except (NameError, KeyError) as err:
- logging.error(
- f"Probably algorithm {file_spec[u'algorithm']} is not defined: "
- f"{repr(err)}"
- )
- logging.info(u"Done.")
-
-
-def _tests_in_suite(suite_name, tests):
- """Check if the suite includes tests.
-
- :param suite_name: Name of the suite to be checked.
- :param tests: Set of tests
- :type suite_name: str
- :type tests: pandas.Series
- :returns: True if the suite includes tests.
- :rtype: bool
- """
-
- for key in tests.keys():
- if suite_name == tests[key][u"parent"]:
- return True
- return False
-
-
-def file_details_split(file_spec, input_data, frmt=u"rst"):
- """Generate the file(s) with algorithms
- - file_details_split
- specified in the specification file.
-
- :param file_spec: File to generate.
- :param input_data: Data to process.
- :param frmt: Format can be: rst or html
- :type file_spec: pandas.Series
- :type input_data: InputData
- :type frmt: str
- """
-
- fileset_file_name = f"{file_spec[u'output-file']}"
- rst_header = (
- u"\n"
- u".. |br| raw:: html\n\n <br />\n\n\n"
- u".. |prein| raw:: html\n\n <pre>\n\n\n"
- u".. |preout| raw:: html\n\n </pre>\n\n"
- )
- start_lvl = file_spec.get(u"data-start-level", 4)
-
- logging.info(f" Generating the file set {fileset_file_name} ...")
-
- data_sets = file_spec.get(u"data", None)
- if not data_sets:
- logging.error(
- f" No data sets specified for {file_spec[u'output-file']}, exit."
- )
- return
-
- table_sets = file_spec.get(u"dir-tables", None)
- if not table_sets:
- logging.error(
- f" No table sets specified for {file_spec[u'output-file']}, exit."
- )
- return
-
- if len(data_sets) != len(table_sets):
- logging.error(
- f" The number of data sets and the number of table sets for "
- f"{file_spec[u'output-file']} are not equal, exit."
- )
- return
-
- chapters = OrderedDict()
- for data_set, table_set in zip(data_sets, table_sets):
-
- logging.info(f" Processing the table set {table_set}...")
-
- table_lst = None
- if frmt == u"html":
- table_lst = get_files(table_set, u".rst", full_path=True)
- elif frmt == u"rst":
- table_lst = get_files(table_set, u".csv", full_path=True)
-
- if not table_lst:
- logging.error(
- f" No tables to include in {table_set}. Skipping."
- )
- continue
-
- logging.info(u" Creating the test data set...")
- tests = input_data.filter_data(
- element=file_spec,
- params=[u"name", u"parent", u"doc", u"type", u"level"],
- data=data_set,
- data_set=u"tests",
- continue_on_error=True
- )
- if tests.empty:
- continue
- tests = input_data.merge_data(tests)
- tests.sort_index(inplace=True)
-
- logging.info(u" Creating the suite data set...")
- suites = input_data.filter_data(
- element=file_spec,
- data=data_set,
- continue_on_error=True,
- data_set=u"suites"
- )
- if suites.empty:
- continue
- suites = input_data.merge_data(suites)
- suites.sort_index(inplace=True)
-
- logging.info(u" Generating files...")
-
- chapter_l1 = u""
- chapter_l2 = u"-".join(table_set.split(u"_")[-2:])
- for suite_longname, suite in suites.items():
-
- suite_lvl = len(suite_longname.split(u"."))
- if suite_lvl < start_lvl:
- # Not interested in this suite
- continue
-
- if suite_lvl == start_lvl:
- # Our top-level suite
- chapter_l1 = suite_longname.split(u'.')[-1]
- if chapters.get(chapter_l1, None) is None:
- chapters[chapter_l1] = OrderedDict()
- if chapters[chapter_l1].get(chapter_l2, None) is None:
- chapters[chapter_l1][chapter_l2] = OrderedDict()
- continue
-
- if _tests_in_suite(suite[u"name"], tests):
- groups = re.search(REGEX_NIC_SHORT, suite[u"name"])
- nic = groups.group(2) if groups else None
- if nic is None:
- continue
- if chapters[chapter_l1][chapter_l2].get(nic, None) is None:
- chapters[chapter_l1][chapter_l2][nic] = dict(
- rst_file=f"{join(table_set, chapter_l1)}_{nic}.rst".
- replace(u"2n1l-", u"").replace(u"1n1l-", u""),
- tables=list()
- )
- for idx, tbl_file in enumerate(table_lst):
- if suite[u"name"] in tbl_file:
- chapters[chapter_l1][chapter_l2][nic][u"tables"].append(
- (table_lst.pop(idx), suite[u"doc"])
- )
- break
- titles = {
- # VPP Perf, MRR
- u"container_memif": u"LXC/DRC Container Memif",
- u"crypto": u"IPsec IPv4 Routing",
- u"hoststack": u"Hoststack Testing",
- u"ip4": u"IPv4 Routing",
- u"ip4_tunnels": u"IPv4 Tunnels",
- u"ip6": u"IPv6 Routing",
- u"ip6_tunnels": u"IPv6 Tunnels",
- u"l2": u"L2 Ethernet Switching",
- u"lb": u"LoadBalancer",
- u"nfv_density": u"NFV Service Density",
- u"srv6": u"SRv6 Routing",
- u"vm_vhost": u"KVM VMs vhost-user",
- u"vts": u"Virtual Topology System",
- # VPP Device
- u"interfaces": u"Interfaces",
- u"l2bd": u"L2 Bridge-domain",
- u"l2patch": u"L2 Patch",
- u"l2xc": u"L2 Cross-connect",
- }
-
- order_chapters = file_spec.get(u"order-chapters", None)
-
- if order_chapters:
- order_1 = order_chapters.get(u"level-1", None)
- order_2 = order_chapters.get(u"level-2", None)
- order_3 = order_chapters.get(u"level-3", None)
- if not order_1:
- order_1 = chapters.keys()
- else:
- order_1 = None
- order_2 = None
- order_3 = None
-
- for chapter_l1 in order_1:
- content_l1 = chapters.get(chapter_l1, None)
- if not content_l1:
- continue
- with open(f"{fileset_file_name}/index.rst", u"a") as file_handler:
- file_handler.write(f" {chapter_l1}\n")
- l1_file_name = f"{join(fileset_file_name, chapter_l1)}.rst"
- title = titles.get(chapter_l1, chapter_l1)
- logging.info(f" Generating {title} ...")
- with open(l1_file_name, u"w") as file_handler:
- file_handler.write(
- f"{title}\n"
- f"{get_rst_title_char(1) * len(title)}\n\n"
- f".. toctree::\n\n"
- )
-
- if not order_2:
- order_2 = chapters[chapter_l1].keys()
- for chapter_l2 in order_2:
- content_l2 = content_l1.get(chapter_l2, None)
- if not content_l2:
- continue
- if not order_3:
- order_3 = chapters[chapter_l1][chapter_l2].keys()
- for chapter_l3 in order_3:
- content_l3 = content_l2.get(chapter_l3, None)
- if not content_l3:
- continue
- with open(l1_file_name, u"a") as file_handler:
- item = u"/".join(content_l3[u'rst_file'].split(u'/')[-2:])
- file_handler.write(f" ../{item}\n")
- logging.info(f" Writing the file {content_l3[u'rst_file']}")
- with open(content_l3[u'rst_file'], u"w+") as file_handler:
- title = f"{chapter_l2}-{chapter_l3}"
- file_handler.write(
- f"{rst_header}\n"
- f"{title}\n"
- f"{get_rst_title_char(2) * len(title)}\n"
- )
- for table in content_l3[u'tables']:
- title = table[0].split(u"/")[-1].split(u".")[0]
- file_handler.write(
- f"\n{title}\n"
- f"{get_rst_title_char(3) * len(title)}\n"
- )
- file_handler.write(f"\n{table[1]}\n")
- if frmt == u"html":
- file_handler.write(
- f"\n.. include:: {table[0].split(u'/')[-1]}"
- f"\n"
- )
- elif frmt == u"rst":
- file_handler.write(
- RST_INCLUDE_TABLE.format(
- file_latex=table[0],
- file_html=table[0].split(u"/")[-1])
- )
-
-
-def file_details_split_html(file_spec, input_data):
- """Generate the file(s) with algorithms
- - file_details_split_html
- specified in the specification file.
-
- :param file_spec: File to generate.
- :param input_data: Data to process.
- :type file_spec: pandas.Series
- :type input_data: InputData
- """
- file_details_split(file_spec, input_data, frmt=u"html")
-
-
-def file_test_results(file_spec, input_data, frmt=u"rst"):
- """Generate the file(s) with algorithms
- - file_test_results
- specified in the specification file.
-
- :param file_spec: File to generate.
- :param input_data: Data to process.
- :param frmt: Format can be: rst or html
- :type file_spec: pandas.Series
- :type input_data: InputData
- :type frmt: str
- """
-
- base_file_name = f"{file_spec[u'output-file']}"
- rst_header = (
- u"\n"
- u".. |br| raw:: html\n\n <br />\n\n\n"
- u".. |prein| raw:: html\n\n <pre>\n\n\n"
- u".. |preout| raw:: html\n\n </pre>\n\n"
- )
- start_lvl = file_spec.get(u"data-start-level", 4)
-
- logging.info(f" Generating the file {base_file_name} ...")
-
- if frmt == u"html":
- table_lst = get_files(file_spec[u"dir-tables"], u".rst", full_path=True)
- elif frmt == u"rst":
- table_lst = get_files(file_spec[u"dir-tables"], u".csv", full_path=True)
- else:
- return
- if not table_lst:
- logging.error(
- f" No tables to include in {file_spec[u'dir-tables']}. Skipping."
- )
- return
-
- logging.info(
- f" Creating the tests data set for the "
- f"{file_spec.get(u'type', u'')} {file_spec.get(u'title', u'')}."
- )
-
- tests = input_data.filter_data(
- file_spec,
- params=[u"name", u"parent", u"doc", u"type", u"level"],
- continue_on_error=True
- )
- if tests.empty:
- return
- tests = input_data.merge_data(tests)
-
- suites = input_data.filter_data(
- file_spec,
- continue_on_error=True,
- data_set=u"suites"
- )
- if suites.empty:
- return
- suites = input_data.merge_data(suites)
- suites.sort_index(inplace=True)
-
- file_name = u""
- for suite_longname, suite in suites.items():
-
- suite_lvl = len(suite_longname.split(u"."))
- if suite_lvl < start_lvl:
- # Not interested in this suite
- continue
-
- if suite_lvl == start_lvl:
- # Our top-level suite
- chapter = suite_longname.split(u'.')[-1]
- file_name = f"{base_file_name}/{chapter}.rst"
- logging.info(f" Writing file {file_name}")
- with open(f"{base_file_name}/index.rst", u"a") as file_handler:
- file_handler.write(f" {chapter}\n")
- with open(file_name, u"a") as file_handler:
- file_handler.write(rst_header)
-
- title_line = get_rst_title_char(suite[u"level"] - start_lvl + 2) * \
- len(suite[u"name"])
- with open(file_name, u"a") as file_handler:
- if not (u"-ndrpdr" in suite[u"name"] or
- u"-mrr" in suite[u"name"] or
- u"-dev" in suite[u"name"]):
- file_handler.write(f"\n{suite[u'name']}\n{title_line}\n")
-
- if _tests_in_suite(suite[u"name"], tests):
- for tbl_file in table_lst:
- if suite[u"name"] in tbl_file:
- file_handler.write(
- f"\n{suite[u'name']}\n{title_line}\n"
- )
- file_handler.write(
- f"\n{suite[u'doc']}\n".replace(u'|br|', u'\n\n -')
- )
- if frmt == u"html":
- file_handler.write(
- f"\n.. include:: {tbl_file.split(u'/')[-1]}\n"
- )
- elif frmt == u"rst":
- file_handler.write(
- RST_INCLUDE_TABLE.format(
- file_latex=tbl_file,
- file_html=tbl_file.split(u"/")[-1])
- )
- break
-
- logging.info(u" Done.")
-
-
-def file_test_results_html(file_spec, input_data):
- """Generate the file(s) with algorithms
- - file_test_results_html
- specified in the specification file.
-
- :param file_spec: File to generate.
- :param input_data: Data to process.
- :type file_spec: pandas.Series
- :type input_data: InputData
- """
- file_test_results(file_spec, input_data, frmt=u"html")