config_tools: replace RstCloth library with class.
Currently, Configurator load fail because it needs to download RstCloth related packages. This patch defines Doc class to replace RstCloth library to fix this issue. Tracked-On: #8395 Signed-off-by: Kunhui-Li <kunhuix.li@intel.com> Reviewed-by: Junjie Mao <junjie.mao@intel.com>
This commit is contained in:
parent
e28d6fbfdf
commit
9c2d0f8858
|
@ -4,14 +4,90 @@
|
|||
#
|
||||
# SPDX-License-Identifier: BSD-3-Clause
|
||||
#
|
||||
|
||||
import sys
|
||||
import argparse
|
||||
import logging
|
||||
|
||||
from rstcloth import RstCloth
|
||||
import typing
|
||||
import functools
|
||||
import textwrap
|
||||
from lxml import etree
|
||||
|
||||
t_content = typing.Union[str, typing.List[str]]
|
||||
|
||||
|
||||
class Doc:
|
||||
def __init__(self, stream: typing.TextIO = sys.stdout, line_width: int = 72) -> None:
|
||||
self._stream = stream
|
||||
self._line_width = line_width
|
||||
|
||||
def fill(self, text: str, initial_indent: int = 0, subsequent_indent: int = 0) -> str:
|
||||
return textwrap.fill(
|
||||
text=text,
|
||||
width=self._line_width,
|
||||
initial_indent=" " * initial_indent,
|
||||
subsequent_indent=" " * subsequent_indent,
|
||||
expand_tabs=False,
|
||||
break_long_words=False,
|
||||
break_on_hyphens=False,
|
||||
)
|
||||
|
||||
def _add(self, content: t_content) -> None:
|
||||
if isinstance(content, list):
|
||||
self._stream.write("\n".join(content) + "\n")
|
||||
else:
|
||||
self._stream.write(content + "\n")
|
||||
|
||||
def content(self, content: t_content, indent: int = 0) -> None:
|
||||
if isinstance(content, list):
|
||||
content = " ".join(content)
|
||||
self._add(self.fill(content, indent, indent))
|
||||
|
||||
def note(self):
|
||||
marker = ".. {type}::".format(type='note')
|
||||
self._add(marker)
|
||||
|
||||
def newline(self, count: int = 1) -> None:
|
||||
if count == 1:
|
||||
self._add("")
|
||||
else:
|
||||
self._add("\n" * (count - 1))
|
||||
|
||||
def table(self, header: typing.List, data) -> None:
|
||||
column_widths = list()
|
||||
content = list()
|
||||
data = [header] + data
|
||||
|
||||
for i in range(len(header)):
|
||||
column_widths.append(max(list(map(lambda x: len(str(x[i])), data))))
|
||||
|
||||
for j in range(len(data)):
|
||||
overline = "+" + "+".join(["-" * column_widths[i] for i in range(len(header))]) + "+"
|
||||
underline = "+" + "+".join(["=" * column_widths[i] for i in range(len(header))]) + "+"
|
||||
format_raw = "|" + "|".join([str(data[j][i]).ljust(column_widths[i]) for i in range(len(header))]) + "|"
|
||||
if j == 0:
|
||||
content.extend([overline, format_raw, underline])
|
||||
else:
|
||||
content.extend([format_raw, overline])
|
||||
|
||||
if len(data) == 1:
|
||||
content.append(overline)
|
||||
|
||||
self.newline()
|
||||
self._add(content)
|
||||
self.newline()
|
||||
|
||||
def heading(self, text: str, char: str, overline: bool = False) -> None:
|
||||
underline = char * len(text)
|
||||
content = [text, underline]
|
||||
if overline:
|
||||
content.insert(0, underline)
|
||||
self._add(content)
|
||||
|
||||
h1 = functools.partialmethod(heading, char="#")
|
||||
h2 = functools.partialmethod(heading, char="*")
|
||||
h3 = functools.partialmethod(heading, char="=")
|
||||
title = functools.partialmethod(heading, char="=", overline=True)
|
||||
|
||||
|
||||
class GenerateRst:
|
||||
io_port = {}
|
||||
|
@ -26,7 +102,7 @@ class GenerateRst:
|
|||
self.board_etree = etree.parse(board_file_name)
|
||||
self.scenario_etree = etree.parse(scenario_file_name)
|
||||
self.file = open(rst_file_name, 'w')
|
||||
self.doc = RstCloth(self.file)
|
||||
self.doc = Doc(self.file)
|
||||
|
||||
# The rst content is written in three parts according to the first level title
|
||||
# 1. Hardware Resource Allocation 2. Inter-VM Connections 3. VM info
|
||||
|
@ -298,9 +374,9 @@ class GenerateRst:
|
|||
amount_vm_l3_cache = self.get_amount_l3_cache(vm_node)
|
||||
parameter_dict["Load Order"] = load_order
|
||||
if load_order == "SERVICE_VM":
|
||||
parameter_dict["Number of vCPUs"] = len(self.service_vm_used_pcpu_list)
|
||||
parameter_dict["Number of vCPUs"] = len(self.service_vm_used_pcpu_list)
|
||||
else:
|
||||
parameter_dict["Number of vCPUs"] = len(vm_node.xpath(f"cpu_affinity/pcpu/pcpu_id"))
|
||||
parameter_dict["Number of vCPUs"] = len(vm_node.xpath(f"cpu_affinity/pcpu/pcpu_id"))
|
||||
parameter_dict["Ammount of RAM"] = str(memory_size) + "MB"
|
||||
parameter_dict["Amount of L3 Cache"] = amount_vm_l3_cache
|
||||
data_table.extend(map(list, parameter_dict.items()))
|
||||
|
|
Loading…
Reference in New Issue