3
0
mirror of https://github.com/triqs/dft_tools synced 2024-12-27 23:03:51 +01:00
dft_tools/foreignlibs/doxylink/doxylink.py

455 lines
18 KiB
Python
Raw Normal View History

# -*- coding: utf-8 -*-
import os
import xml.etree.ElementTree as ET
import urlparse
import re
import itertools
from docutils import nodes, utils
from sphinx.util.nodes import split_explicit_title
from sphinx.util.console import bold, standout
from parsing import normalise, ParseException
def find_url(doc, symbol):
"""
Return the URL for a given symbol.
This is where the magic happens.
This function could be a lot more clever. At present it required the passed symbol to be almost exactly the same as the entries in the Doxygen tag file.
.. todo::
Maybe print a list of all possible matches as a warning (but still only return the first)
:Parameters:
doc : xml.etree.ElementTree
The XML DOM object
symbol : string
The symbol to lookup in the file. E.g. something like 'PolyVox::Array' or 'tidyUpMemory'
:return: String representing the filename part of the URL
"""
#First check for an exact match with a top-level object (namespaces, objects etc.)
#env = inliner.document.settings.env
matches = []
for compound in doc.findall('.//compound'):
if compound.find('name').text == symbol:
matches += [{'file':compound.find('filename').text, 'kind':compound.get('kind')}]
if len(matches) > 1:
pass
#env.warn(env.docname, 'There were multiple matches for `%s`: %s' % (symbol, matches))
if len(matches) == 1:
return matches[0]
#Strip off first namespace bit of the compound name so that 'ArraySizes' can match 'PolyVox::ArraySizes'
for compound in doc.findall('.//compound'):
symbol_list = compound.find('name').text.split('::', 1)
if len(symbol_list) == 2:
reducedsymbol = symbol_list[1]
if reducedsymbol == symbol:
return {'file':compound.find('filename').text, 'kind':compound.get('kind')}
#Now split the symbol by '::'. Find an exact match for the first part and then a member match for the second
#So PolyVox::Array::operator[] becomes like {namespace: "PolyVox::Array", endsymbol: "operator[]"}
symbol_list = symbol.rsplit('::', 1)
if len(symbol_list) == 2:
namespace = symbol_list[0]
endsymbol = symbol_list[1]
for compound in doc.findall('.//compound'):
if compound.find('name').text == namespace:
for member in compound.findall('member'):
#If this compound object contains the matching member then return it
if member.find('name').text == endsymbol:
return {'file':(member.findtext('anchorfile') or compound.findtext('filename')) + '#' + member.find('anchor').text, 'kind':member.get('kind')}
#Then we'll look at unqualified members
for member in doc.findall('.//member'):
if member.find('name').text == symbol:
return {'file':(member.findtext('anchorfile') or compound.findtext('filename')) + '#' + member.find('anchor').text, 'kind':member.get('kind')}
return None
def parse_tag_file(doc):
"""
Takes in an XML tree from a Doxygen tag file and returns a dictionary that looks something like:
.. code-block:: python
{'PolyVox': {'file': 'namespace_poly_vox.html',
'kind': 'namespace'},
'PolyVox::Array': {'file': 'class_poly_vox_1_1_array.html',
'kind': 'class'},
'PolyVox::Array1DDouble': {'file': 'namespace_poly_vox.html#a7a1f5fd5c4f7fbb4258a495d707b5c13',
'kind': 'typedef'},
'PolyVox::Array1DFloat': {'file': 'namespace_poly_vox.html#a879a120e49733eba1905c33f8a7f131b',
'kind': 'typedef'},
'PolyVox::Array1DInt16': {'file': 'namespace_poly_vox.html#aa1463ece448c6ebed55ab429d6ae3e43',
'kind': 'typedef'},
'QScriptContext::throwError': {'arglist': {'( Error error, const QString & text )': 'qscriptcontext.html#throwError',
'( const QString & text )': 'qscriptcontext.html#throwError-2'},
'kind': 'function'},
'QScriptContext::toString': {'arglist': {'()': 'qscriptcontext.html#toString'},
'kind': 'function'}}
Note the different form for functions. This is required to allow for 'overloading by argument type'.
To access a filename for a symbol you do:
.. code-block:: python
symbol_mapping = mapping[symbol]
if symbol_mapping['kind'] == 'function':
url = symbol_mapping['arglist'][argument_string]
else:
url = symbol_mapping['file']
:Parameters:
doc : xml.etree.ElementTree
The XML DOM object
:return: a dictionary mapping fully qualified symbols to files
"""
mapping = {}
function_list = [] #This is a list of function to be parsed and inserted into mapping at the end of the function.
for compound in doc.findall("./compound"):
compound_kind = compound.get('kind')
if compound_kind != 'namespace' and compound_kind != 'class' and compound_kind!= 'struct' and compound_kind != 'file':
continue #Skip everything that isn't a namespace, class, struct or file
compound_name = compound.findtext('name')
compound_filename = compound.findtext('filename')
#TODO The following is a hack bug fix I think
#Doxygen doesn't seem to include the file extension to <compound kind="file"><filename> entries
#If it's a 'file' type, check if it _does_ have an extension, if not append '.html'
if compound_kind == 'file' and not os.path.splitext(compound_filename)[1]:
compound_filename = join(compound_filename, '.html')
#If it's a compound we can simply add it
mapping[compound_name] = {'kind' : compound_kind, 'file' : compound_filename}
for member in compound.findall('member'):
#If the member doesn't have an <anchorfile> element, use the parent compounds <filename> instead
#This is the way it is in the qt.tag and is perhaps an artefact of old Doxygen
anchorfile = member.findtext('anchorfile') or compound_filename
member_symbol = join(compound_name, '::', member.findtext('name'))
member_kind = member.get('kind')
arglist_text = member.findtext('./arglist') #If it has an <arglist> then we assume it's a function. Empty <arglist> returns '', not None. Things like typedefs and enums can have empty arglists
if arglist_text and member_kind != 'variable' and member_kind != 'typedef' and member_kind != 'enumeration':
function_list.append((member_symbol, arglist_text, member_kind, join(anchorfile,'#',member.findtext('anchor'))))
else:
mapping[member_symbol] = {'kind' : member.get('kind'), 'file' : join(anchorfile,'#',member.findtext('anchor'))}
for old_tuple, normalised_tuple in zip(function_list, itertools.imap(normalise, (member_tuple[1] for member_tuple in function_list))):
member_symbol = old_tuple[0]
original_arglist = old_tuple[1]
kind = old_tuple[2]
anchor_link = old_tuple[3]
normalised_arglist = normalised_tuple[1]
if normalised_tuple[1] is not None: #This is a 'flag' for a ParseException having happened
if mapping.get(member_symbol):
mapping[member_symbol]['arglist'][normalised_arglist] = anchor_link
else:
mapping[member_symbol] = {'kind' : kind, 'arglist' : {normalised_arglist : anchor_link}}
else:
print 'Skipping %s %s%s. Error reported from parser was: %s' % (old_tuple[2], old_tuple[0], old_tuple[1], normalised_tuple[0])
#from pprint import pprint; pprint(mapping)
return mapping
def find_url2(mapping, symbol):
"""
Return the URL for a given symbol.
This is where the magic happens.
.. todo::
Maybe print a list of all possible matches as a warning (but still only return the first)
:Parameters:
mapping : dictionary
A dictionary of the form returned by :py:func:`parse_tag_file`
symbol : string
The symbol to lookup in the file. E.g. something like 'PolyVox::Array' or 'tidyUpMemory'
:return: String representing the filename part of the URL
:raises:
LookupError
Raised if the symbol could not be matched in the file
"""
#print "\n\nSearching for", symbol
try:
symbol, normalised_arglist = normalise(symbol)
except ParseException as error:
raise LookupError(error)
#print symbol, normalised_arglist
#If we have an exact match then return it.
if mapping.get(symbol):
#print ('Exact match')
return return_from_mapping(mapping[symbol], normalised_arglist)
#If the user didn't pass in any arguments, i.e. `arguments == ''` then they don't care which version of the overloaded funtion they get.
#First we check for any mapping entries which even slightly match the requested symbol
#endswith_list = {}
#for item, data in mapping.items():
# if item.endswith(symbol):
#print symbol + ' : ' + item
# endswith_list[item] = data
# mapping[item]['file']
#If we only find one then we return it.
#if len(endswith_list) is 1:
# return endswith_list.values()[0]['file']
#print("Still", len(endswith_list), 'possible matches')
piecewise_list = find_url_piecewise(mapping, symbol)
#If there is only one match, return it.
if len(piecewise_list) is 1:
return return_from_mapping(piecewise_list.values()[0], normalised_arglist)
#print("Still", len(piecewise_list), 'possible matches')
#If there is more than one item in piecewise_list then there is an ambiguity
#Often this is due to the symbol matching the name of the constructor as well as the class name itself
classes_list = find_url_classes(piecewise_list, symbol)
#If there is only one by here we return it.
if len(classes_list) is 1:
return classes_list.values()[0]
#print("Still", len(classes_list), 'possible matches')
#If we exhaused the list by requiring classes, use the list from before the filter.
if len(classes_list) == 0:
classes_list = piecewise_list
no_templates_list = find_url_remove_templates(classes_list, symbol)
if len(no_templates_list) is 1:
return return_from_mapping(no_templates_list.values()[0], normalised_arglist)
#print("Still", len(no_templates_list), 'possible matches')
#If not found by now, just return the first one in the list
if len(no_templates_list) != 0:
#TODO return a warning here?
return return_from_mapping(no_templates_list.values()[0], normalised_arglist)
#Else return None if the list is empty
else:
LookupError('Could not find a match')
def return_from_mapping(mapping_entry, normalised_arglist=''):
"""
Return a mapping to a single URL in the form. This is needed since mapping entries for functions are more complicated due to function overriding.
If the mapping to be returned is not a function, this will simply return the mapping entry intact. If the entry is a function it will attempt to get the right version based on the function signature.
:Parameters:
mapping_entry : dict
should be a single entry from the large mapping file corresponding to a single symbol. If the symbol is a function, then ``mappingentry['arglist']`` will be a dictionary mapping normalised signatures to URLs
normalised_arglist : string
the normalised form of the arglist that the user has requested. This can be empty in which case the function will return just the first element of ``mappingentry['arglist']``. This parameter is ignored if ``mappingentry['kind'] != 'function'``
:return: dictionary something like:
.. code-block:: python
{'kind' : 'function', 'file' : 'something.html#foo'}
"""
#If it's a function we need to grab the right signature from the arglist.
if mapping_entry['kind'] == 'function':
#If the user has requested a specific function through specifying an arglist then get the right anchor
if normalised_arglist:
filename = mapping_entry['arglist'].get(normalised_arglist)
if not filename: #If we didn't get the filename because it's not in the mapping then we will just return a random one?
#TODO return a warning here!
filename = mapping_entry['arglist'].values()[0]
else:
#Otherwise just return the first entry (if they don't care they get whatever comes first)
filename = mapping_entry['arglist'].values()[0]
return {'kind' : 'function', 'file' : filename}
elif mapping_entry.get('arglist'):
#This arglist should only be one entry long and that entry should have '' as its key
return {'kind' : mapping_entry['kind'], 'file' : mapping_entry['arglist']['']}
#If it's not a function, then return it raw
return mapping_entry
def find_url_piecewise(mapping, symbol):
"""
Match the requested symbol reverse piecewise (split on ``::``) against the tag names to ensure they match exactly (modulo ambiguity)
So, if in the mapping there is ``PolyVox::Volume::FloatVolume`` and ``PolyVox::Volume`` they would be split into:
.. code-block:: python
['PolyVox', 'Volume', 'FloatVolume'] and ['PolyVox', 'Volume']
and reversed:
.. code-block:: python
['FloatVolume', 'Volume', 'PolyVox'] and ['Volume', 'PolyVox']
and truncated to the shorter of the two:
.. code-block:: python
['FloatVolume', 'Volume'] and ['Volume', 'PolyVox']
If we're searching for the ``PolyVox::Volume`` symbol we would compare:
.. code-block:: python
['Volume', 'PolyVox'] to ['FloatVolume', 'Volume', 'PolyVox'].
That doesn't match so we look at the next in the mapping:
.. code-block:: python
['Volume', 'PolyVox'] to ['Volume', 'PolyVox'].
Good, so we add it to the list
"""
piecewise_list = {}
for item, data in mapping.items():
split_symbol = symbol.split('::')
split_item = item.split('::')
split_symbol.reverse()
split_item.reverse()
min_length = min(len(split_symbol), len(split_item))
split_symbol = split_symbol[:min_length]
split_item = split_item[:min_length]
#print split_symbol, split_item
if split_symbol == split_item:
#print symbol + ' : ' + item
piecewise_list[item] = data
return piecewise_list
def find_url_classes(mapping, symbol):
"""Prefer classes over names of constructors"""
classes_list = {}
for item, data in mapping.items():
if data['kind'] == 'class':
#print symbol + ' : ' + item
classes_list[item] = data
return classes_list
def find_url_remove_templates(mapping, symbol):
"""Now, to disambiguate between ``PolyVox::Array< 1, ElementType >::operator[]`` and ``PolyVox::Array::operator[]`` matching ``operator[]``, we will ignore templated (as in C++ templates) tag names by removing names containing ``<``"""
no_templates_list = {}
for item, data in mapping.items():
if '<' not in item:
#print symbol + ' : ' + item
no_templates_list[item] = data
return no_templates_list
def join(*args):
return ''.join(args)
def create_role(app, tag_filename, rootdir):
#Tidy up the root directory path
if not rootdir.endswith(('/', '\\')):
rootdir = join(rootdir, os.sep)
try:
tag_file = ET.parse(tag_filename)
cache_name = os.path.basename(tag_filename)
app.info(bold('Checking tag file cache for %s: ' % cache_name), nonl=True)
if not hasattr(app.env, 'doxylink_cache'):
# no cache present at all, initialise it
app.info('No cache at all, rebuilding...')
mapping = parse_tag_file(tag_file)
app.env.doxylink_cache = { cache_name : {'mapping' : mapping, 'mtime' : os.path.getmtime(tag_filename)}}
elif not app.env.doxylink_cache.get(cache_name):
# Main cache is there but the specific sub-cache for this tag file is not
app.info('Sub cache is missing, rebuilding...')
mapping = parse_tag_file(tag_file)
app.env.doxylink_cache[cache_name] = {'mapping' : mapping, 'mtime' : os.path.getmtime(tag_filename)}
elif app.env.doxylink_cache[cache_name]['mtime'] < os.path.getmtime(tag_filename):
# tag file has been modified since sub-cache creation
app.info('Sub-cache is out of date, rebuilding...')
mapping = parse_tag_file(tag_file)
app.env.doxylink_cache[cache_name] = {'mapping' : mapping, 'mtime' : os.path.getmtime(tag_filename)}
else:
#The cache is up to date
app.info('Sub-cache is up-to-date')
except IOError:
tag_file = None
app.warn(standout('Could not open tag file %s. Make sure your `doxylink` config variable is set correctly.' % tag_filename))
def find_doxygen_link(name, rawtext, text, lineno, inliner, options={}, content=[]):
text = utils.unescape(text)
# from :name:`title <part>`
has_explicit_title, title, part = split_explicit_title(text)
warning_messages = []
if tag_file:
url = find_url(tag_file, part)
# If there are several doxy projects within the same sphinx instance then the cache is always OK
# and one cannot regenerate a new cache for a new project... So I comment these 4 lines...
#try:
# url = find_url2(app.env.doxylink_cache[cache_name]['mapping'], part)
#except LookupError as error:
# warning_messages.append('Error while parsing `%s`. Is not a well-formed C++ function call or symbol. If this is not the case, it is a doxylink bug so please report it. Error reported was: %s' % (part, error))
if url:
#If it's an absolute path then the link will work regardless of the document directory
#Also check if it is a URL (i.e. it has a 'scheme' like 'http' or 'file')
if os.path.isabs(rootdir) or urlparse.urlparse(rootdir).scheme:
full_url = join(rootdir, url['file'])
#But otherwise we need to add the relative path of the current document to the root source directory to the link
else:
relative_path_to_docsrc = os.path.relpath(app.env.srcdir, os.path.dirname(inliner.document.current_source))
full_url = join(relative_path_to_docsrc, '/', rootdir, url['file']) #We always use the '/' here rather than os.sep since this is a web link avoids problems like documentation/.\../library/doc/ (mixed slashes)
if url['kind'] == 'function' and app.config.add_function_parentheses and not normalise(title)[1]:
title = join(title, '()')
pnode = nodes.reference(title, title, internal=False, refuri=full_url)
return [pnode], []
#By here, no match was found
warning_messages.append('Could not find match for `%s` in `%s` tag file' % (part, tag_filename))
else:
warning_messages.append('Could not find match for `%s` because tag file not found' % (part))
pnode = nodes.inline(rawsource=title, text=title)
return [pnode], [inliner.reporter.warning(message, line=lineno) for message in warning_messages]
return find_doxygen_link
def setup_doxylink_roles(app):
for name, [tag_filename, rootdir] in app.config.doxylink.iteritems():
app.add_role(name, create_role(app, tag_filename, rootdir))
def setup(app):
app.add_config_value('doxylink', {}, 'env')
app.connect('builder-inited', setup_doxylink_roles)