Vulkan-Docs/doc/specs/vulkan/genRef.py

543 lines
18 KiB
Python
Executable File

#!/usr/bin/python3
#
# Copyright (c) 2016 The Khronos Group Inc.
#
# 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.
# genRef.py - create Vulkan ref pages from spec source files
#
#
# Usage: genRef.py files
from reflib import *
from vkapi import *
import argparse, copy, os, pdb, re, string, sys
# Return True if name is a Vulkan extension name (ends with an upper-case
# author ID). This assumes that author IDs are at least two characters.
def isextension(name):
return name[-2:].isalpha() and name[-2:].isupper()
# Print Khronos CC-BY copyright notice on open file fp. If comment is
# True, print as an asciidoc comment block, which copyrights the source
# file. Otherwise print as an asciidoc include of the copyright in markup,
# which copyrights the outputs.
def printCopyrightBlock(fp, comment=True):
if (comment):
print('// Copyright (c) 2014-2016 Khronos Group. This work is licensed under a', file=fp)
print('// Creative Commons Attribution 4.0 International License; see', file=fp)
print('// http://creativecommons.org/licenses/by/4.0/', file=fp)
print('', file=fp)
else:
print('include::footer.txt[]', file=fp)
print('', file=fp)
# Add a spec asciidoc macro prefix to a Vulkan name, depending on its type
# (protos, structs, enums, etc.)
def macroPrefix(name):
if (name in basetypes.keys()):
return 'basetype:' + name
elif (name in defines.keys()):
return 'slink:' + name
elif (name in enums.keys()):
return 'elink:' + name
elif (name in flags.keys()):
return 'elink:' + name
elif (name in funcpointers.keys()):
return 'tlink:' + name
elif (name in handles.keys()):
return 'slink:' + name
elif (name in protos.keys()):
return 'flink:' + name
elif (name in structs.keys()):
return 'slink:' + name
elif (name == 'TBD'):
return 'No cross-references are available'
else:
return 'UNKNOWN:' + name
# Return an asciidoc string with a list of 'See Also' references for the
# Vulkan entity 'name', based on the relationship mapping in vkapi.py and
# the additional references in explicitRefs. If no relationships are
# available, return None.
def seeAlsoList(apiName, explicitRefs = None):
refs = {}
# Add all the implicit references to refs
if (apiName in mapDict.keys()):
for name in sorted(mapDict[apiName].keys()):
refs[name] = None
# Add all the explicit references
if (explicitRefs != None):
for name in explicitRefs.split():
refs[name] = None
names = [macroPrefix(name) for name in sorted(refs.keys())]
if (len(names) > 0):
return ', '.join(names) + '\n'
else:
return None
# Remap include directives in a list of lines so they can be extracted to a
# different directory. Returns remapped lines.
#
# lines - text to remap
# baseDir - target directory
# specDir - source directory
def remapIncludes(lines, baseDir, specDir):
# This should be compiled only once
includePat = re.compile('^include::(?P<path>.*)\[\]')
newLines = []
for line in lines:
matches = includePat.search(line)
if (matches != None):
path = matches.group('path')
# Relative path to include file from here
incPath = specDir + '/' + path
# Remap to be relative to baseDir
newPath = os.path.relpath(incPath, baseDir)
newLine = 'include::' + newPath + '[]\n'
logDiag('remapIncludes: remapping from:\n\t', line, 'to:\n\t', newLine)
newLines.append(newLine)
else:
newLines.append(line)
return newLines
# Generate header of a reference page
# pageName - string name of the page
# pageDesc - string short description of the page
# specText - string that goes in the "C Specification" section
# fieldName - string heading an additional section following specText, if not None
# fieldText - string that goes in the additional section
# descText - string that goes in the "Description" section
# fp - file to write to
def refPageHead(pageName, pageDesc, specText, fieldName, fieldText, descText, fp):
printCopyrightBlock(fp, comment=True)
s = pageName + '(3)'
print(s,
''.ljust(len(s), '='),
'',
sep='\n', file=fp)
print('Name',
'----',
pageName + ' - ' + pageDesc,
'',
sep='\n', file=fp)
print('C Specification',
'---------------',
'',
specText,
'',
sep='\n', file=fp)
if (fieldName != None):
print(fieldName,
''.ljust(len(fieldName), '-'),
'',
fieldText,
sep='\n', file=fp)
print('Description',
'-----------',
'',
descText,
'',
sep='\n', file=fp)
def refPageTail(pageName, seeAlso, fp, auto = False):
# This is difficult to get working properly in asciidoc
# specURL = 'link:{vkspecpath}/vkspec.html'
# This needs to have the current repository branch path installed in
# place of '1.0'
specURL = 'https://www.khronos.org/registry/vulkan/specs/1.0/xhtml/vkspec.html'
if (seeAlso == None):
seeAlso = 'No cross-references are available\n'
notes = [
'For more information, see the Vulkan Specification at URL',
'',
specURL + '#' + pageName,
'',
]
if (auto):
notes.extend([
'This page is a generated document.',
'Fixes and changes should be made to the generator scripts,'
'not directly.',
])
else:
notes.extend([
'This page is extracted from the Vulkan Specification.',
'Fixes and changes should be made to the Specification,'
'not directly.',
])
print('See Also',
'--------',
'',
seeAlso,
'',
sep='\n', file=fp)
print('Document Notes',
'--------------',
'',
'\n'.join(notes),
'',
sep='\n', file=fp)
printCopyrightBlock(fp, comment=False)
# Extract a single reference page into baseDir
# baseDir - base directory to emit page into
# specDir - directory extracted page source came from
# pi - pageInfo for this page relative to file
# file - list of strings making up the file, indexed by pi
def emitPage(baseDir, specDir, pi, file):
pageName = baseDir + '/' + pi.name + '.txt'
fp = open(pageName, 'w')
# Add a dictionary entry for this page
global genDict
genDict[pi.name] = None
logDiag('emitPage:', pageName)
# Short description
if (pi.desc == None):
pi.desc = '(no short description available)'
# Specification text
lines = remapIncludes(file[pi.begin:pi.include+1], baseDir, specDir)
specText = ''.join(lines)
# Member/parameter list, if there is one
field = None
fieldText = None
if (pi.param != None):
if (pi.type == 'structs'):
field = 'Members'
elif (pi.type in ['protos', 'funcpointers']):
field = 'Parameters'
else:
logWarn('PyOutputGenerator::emitPage: unknown field type:', pi.type,
'for', pi.name)
lines = remapIncludes(file[pi.param:pi.body], baseDir, specDir)
fieldText = ''.join(lines)
# Description text
lines = remapIncludes(file[pi.body:pi.end+1], baseDir, specDir)
descText = ''.join(lines)
refPageHead(pi.name,
pi.desc,
specText,
field, fieldText,
descText,
fp)
refPageTail(pi.name, seeAlsoList(pi.name, pi.refs), fp, auto = False)
fp.close()
# Autogenerate a single reference page in baseDir
# Script only knows how to do this for /enums/ pages, at present
# baseDir - base directory to emit page into
# pi - pageInfo for this page relative to file
# file - list of strings making up the file, indexed by pi
def autoGenEnumsPage(baseDir, pi, file):
pageName = baseDir + '/' + pi.name + '.txt'
fp = open(pageName, 'w')
# Add a dictionary entry for this page
global genDict
genDict[pi.name] = None
logDiag('autoGenEnumsPage:', pageName)
# Short description
if (pi.desc == None):
pi.desc = '(no short description available)'
# Description text
txt = ''.join([
'For more information, see:\n\n',
' * The reference page for ' + macroPrefix(pi.embed) +
', where this interface is defined.\n',
' * The See Also section for other reference pages using this type.\n',
' * The Vulkan Specification.\n' ])
refPageHead(pi.name,
pi.desc,
''.join(file[pi.begin:pi.include+1]),
None, None,
txt,
fp)
refPageTail(pi.name, seeAlsoList(pi.name, pi.refs), fp, auto = True)
fp.close()
# Pattern to break apart a Vk*Flags{authorID} name, used in autoGenFlagsPage.
flagNamePat = re.compile('(?P<name>\w+)Flags(?P<author>[A-Z]*)')
# Autogenerate a single reference page in baseDir for a Vk*Flags type
# baseDir - base directory to emit page into
# flagName - Vk*Flags name
def autoGenFlagsPage(baseDir, flagName):
pageName = baseDir + '/' + flagName + '.txt'
fp = open(pageName, 'w')
# Add a dictionary entry for this page
global genDict
genDict[flagName] = None
logDiag('autoGenFlagsPage:', pageName)
# Short description
matches = flagNamePat.search(flagName)
if (matches != None):
name = matches.group('name')
author = matches.group('author')
logDiag('autoGenFlagsPage: split name into', name, 'Flags', author)
flagBits = name + 'FlagBits' + author
desc = 'Bitmask of ' + flagBits
else:
logWarn('autoGenFlagsPage:', pageName, 'does not not end in "Flags{author ID}". Cannot infer FlagBits type.')
flagBits = None
desc = 'Unknown Vulkan flags type'
# Description text
if (flagBits != None):
txt = ''.join([
'etext:' + flagName,
' is a mask of zero or more elink:' + flagBits + '.\n',
'It is used as a member and/or parameter of the structures and commands\n',
'in the See Also section below.\n' ])
else:
txt = ''.join([
'etext:' + flagName,
' is an unknown Vulkan type, assumed to be a bitmask.\n' ])
refPageHead(flagName,
desc,
'include::../api/flags/' + flagName + '.txt[]\n',
None, None,
txt,
fp)
refPageTail(flagName, seeAlsoList(flagName), fp, auto = True)
fp.close()
# Autogenerate a single handle page in baseDir for a Vk* handle type
# baseDir - base directory to emit page into
# handleName - Vk* handle name
# @@ Need to determine creation function & add handles/ include for the
# @@ interface in generator.py.
def autoGenHandlePage(baseDir, handleName):
pageName = baseDir + '/' + handleName + '.txt'
fp = open(pageName, 'w')
# Add a dictionary entry for this page
global genDict
genDict[handleName] = None
logDiag('autoGenHandlePage:', pageName)
# Short description
desc = 'Vulkan object handle'
descText = ''.join([
'sname:' + handleName,
' is an object handle type, referring to an object used\n',
'by the Vulkan implementation. These handles are created or allocated\n',
'by the vk @@ TBD @@ function, and used by other Vulkan structures\n',
'and commands in the See Also section below.\n' ])
refPageHead(handleName,
desc,
'include::../api/handles/' + handleName + '.txt[]\n',
None, None,
descText,
fp)
refPageTail(handleName, seeAlsoList(handleName), fp, auto = True)
fp.close()
# Extract reference pages from a spec asciidoc source file
# specFile - filename to extract from
# baseDir - output directory to generate page in
#
def genRef(specFile, baseDir):
file = loadFile(specFile)
if (file == None):
return
# Save the path to this file for later use in rewriting relative includes
specDir = os.path.dirname(os.path.abspath(specFile))
pageMap = findRefs(file)
logDiag(specFile + ': found', len(pageMap.keys()), 'potential pages')
sys.stderr.flush()
# Fix up references in pageMap
fixupRefs(pageMap, specFile, file)
# Create each page, if possible
for name in sorted(pageMap.keys()):
pi = pageMap[name]
printPageInfo(pi, file)
if (pi.Warning):
logDiag('genRef:', pi.name + ':', pi.Warning)
if (pi.extractPage):
emitPage(baseDir, specDir, pi, file)
elif (pi.type == 'enums'):
autoGenEnumsPage(baseDir, pi, file)
elif (pi.type == 'flags'):
autoGenFlagsPage(baseDir, pi.name)
else:
# Don't extract this page
logWarn('genRef: Cannot extract or autogenerate:', pi.name)
# Generate baseDir/apispec.txt, the single-page version of the ref pages.
# This assumes there's a page for everything in the vkapi.py dictionaries.
# Extensions (KHR, EXT, etc.) are currently skipped
def genSinglePageRef(baseDir):
pageName = baseDir + '/apispec.txt'
fp = open(pageName, 'w')
printCopyrightBlock(fp, comment=True)
print('Vulkan API Reference Pages',
'==========================',
'include::../specversion.txt[]',
'',
':doctype: book',
':numbered!:',
':toc2:',
':max-width: 200',
':numbered:',
':doctype: book',
':data-uri:',
':asciimath:',
':toclevels: 4',
'',
sep='\n', file=fp)
print('include::khronoscopyright.txt[]', file=fp)
print('', file=fp)
# Inject the table of contents. Asciidoc really ought to be generating
# this for us.
sections = [
[ protos, 'protos', 'Vulkan Commands' ],
[ handles, 'handles', 'Object Handles' ],
[ structs, 'structs', 'Structures' ],
[ enums, 'enums', 'Enumerations' ],
[ flags, 'flags', 'Flags' ],
[ funcpointers, 'funcpointers', 'Function Pointer Types' ],
[ basetypes, 'basetypes', 'Vulkan Scalar types' ],
[ defines, 'defines', 'C Macro Definitions' ] ]
print('Table of Contents', file=fp)
print('-----------------', file=fp)
for (apiDict,label,title) in sections:
print(' * <<' + label + ',' + title + '>>', file=fp)
print('', file=fp)
for (apiDict,label,title) in sections:
anchor = '[[' + label + ',' + title + ']]'
print(anchor,
title,
''.ljust(len(title), '-'),
'',
':leveloffset: 2',
'',
sep='\n', file=fp)
for refPage in sorted(apiDict.keys()):
if (apiDict == defines or not isextension(refPage)):
print('include::' + refPage + '.txt[]', file=fp)
else:
print('// not including ' + refPage, file=fp)
print('\n' + ':leveloffset: 0' + '\n', file=fp)
# An index of pages could be generated here
if __name__ == '__main__':
global genDict
genDict = {}
parser = argparse.ArgumentParser()
parser.add_argument('-diag', action='store', dest='diagFile',
help='Set the diagnostic file')
parser.add_argument('-warn', action='store', dest='warnFile',
help='Set the warning file')
parser.add_argument('-log', action='store', dest='logFile',
help='Set the log file for both diagnostics and warnings')
parser.add_argument('-basedir', action='store', dest='baseDir',
default='man',
help='Set the base directory in which pages are generated')
parser.add_argument('files', metavar='filename', nargs='*',
help='a filename to extract ref pages from')
parser.add_argument('--version', action='version', version='%(prog)s 1.0')
results = parser.parse_args()
setLogFile(True, True, results.logFile)
setLogFile(True, False, results.diagFile)
setLogFile(False, True, results.warnFile)
baseDir = results.baseDir
for file in results.files:
genRef(file, baseDir)
# Now figure out which pages *weren't* generated from the spec.
# This relies on the dictionaries of API constructs in vkapi.py.
# For Flags (e.g. Vk*Flags types), it's easy to autogenerate pages.
for page in flags.keys():
if not (page in genDict.keys()):
logWarn('Autogenerating flags page:', page, 'which should be included in the spec')
autoGenFlagsPage(baseDir, page)
# autoGenHandlePage is no longer needed because they are added to
# the spec sources now.
# for page in structs.keys():
# if (typeCategory[page] == 'handle'):
# autoGenHandlePage(baseDir, page)
sections = [
[ enums, 'Enumerated Types' ],
[ structs, 'Structures' ],
[ protos, 'Prototypes' ],
[ funcpointers, 'Function Pointers' ],
[ basetypes, 'Vulkan Scalar Types' ] ]
for (apiDict,title) in sections:
flagged = False
for page in apiDict.keys():
if not (page in genDict.keys()):
if (not flagged):
logWarn(title, 'with no ref page generated:')
flagged = True
logWarn(' ', page)
genSinglePageRef(baseDir)