/python/helpers/epydoc/docwriter/html.py

Large files files are truncated, but you can click here to view the full file

#
# epydoc -- HTML output generator
# Edward Loper
#
# Created [01/30/01 05:18 PM]
# $Id: html.py 1674 2008-01-29 06:03:36Z edloper $
#

"""
The HTML output generator for epydoc.  The main interface provided by
this module is the L{HTMLWriter} class.

@todo: Add a cache to L{HTMLWriter.url()}?
"""
__docformat__ = 'epytext en'

import re, os, sys, codecs, sre_constants, pprint, base64
import urllib
import __builtin__
from epydoc.apidoc import *
import epydoc.docstringparser
import time, epydoc, epydoc.markup, epydoc.markup.epytext
from epydoc.docwriter.html_colorize import PythonSourceColorizer
from epydoc.docwriter import html_colorize
from epydoc.docwriter.html_css import STYLESHEETS
from epydoc.docwriter.html_help import HTML_HELP
from epydoc.docwriter.dotgraph import *
from epydoc import log
from epydoc.util import plaintext_to_html, is_src_filename
from epydoc.compat import * # Backwards compatibility

######################################################################
## Template Compiler
######################################################################
# The compile_template() method defined in this section is used to
# define several of HTMLWriter's methods.

def compile_template(docstring, template_string,
                     output_function='out', debug=epydoc.DEBUG):
    """
    Given a template string containing inline python source code,
    return a python function that will fill in the template, and
    output the result.  The signature for this function is taken from
    the first line of C{docstring}.  Output is generated by making
    repeated calls to the output function with the given name (which
    is typically one of the function's parameters).

    The templating language used by this function passes through all
    text as-is, with three exceptions:

      - If every line in the template string is indented by at least
        M{x} spaces, then the first M{x} spaces are stripped from each
        line.

      - Any line that begins with '>>>' (with no indentation)
        should contain python code, and will be inserted as-is into
        the template-filling function.  If the line begins a control
        block (such as 'if' or 'for'), then the control block will
        be closed by the first '>>>'-marked line whose indentation is
        less than or equal to the line's own indentation (including
        lines that only contain comments.)

      - In any other line, any expression between two '$' signs will
        be evaluated and inserted into the line (using C{str()} to
        convert the result to a string).

    Here is a simple example:

        >>> TEMPLATE = '''
        ... <book>
        ...   <title>$book.title$</title>
        ...   <pages>$book.count_pages()$</pages>
        ... >>> for chapter in book.chapters:
        ...     <chaptername>$chapter.name$</chaptername>
        ... >>> #endfor
        ... </book>
        >>> write_book = compile_template('write_book(out, book)', TEMPLATE)

    @newfield acknowledgements: Acknowledgements
    @acknowledgements: The syntax used by C{compile_template} is
    loosely based on Cheetah.
    """
    # Extract signature from the docstring:
    signature = docstring.lstrip().split('\n',1)[0].strip()
    func_name = signature.split('(',1)[0].strip()

    # Regexp to search for inline substitutions:
    INLINE = re.compile(r'\$([^\$]+)\$')
    # Regexp to search for python statements in the template:
    COMMAND = re.compile(r'(^>>>.*)\n?', re.MULTILINE)

    # Strip indentation from the template.
    template_string = strip_indent(template_string)

    # If we're debugging, then we'll store the generated function,
    # so we can print it along with any tracebacks that depend on it.
    if debug:
        signature = re.sub(r'\)\s*$', ', __debug=__debug)', signature)

    # Funciton declaration line
    pysrc_lines = ['def %s:' % signature]
    indents = [-1]

    if debug:
        pysrc_lines.append('    try:')
        indents.append(-1)

    commands = COMMAND.split(template_string.strip()+'\n')
    for i, command in enumerate(commands):
        if command == '': continue

        # String literal segment:
        if i%2 == 0:
            pieces = INLINE.split(command)
            for j, piece in enumerate(pieces):
                if j%2 == 0:
                    # String piece
                    pysrc_lines.append('    '*len(indents)+
                                 '%s(%r)' % (output_function, piece))
                else:
                    # Variable piece
                    pysrc_lines.append('    '*len(indents)+
                                 '%s(unicode(%s))' % (output_function, piece))

        # Python command:
        else:
            srcline = command[3:].lstrip()
            # Update indentation
            indent = len(command)-len(srcline)
            while indent <= indents[-1]: indents.pop()
            # Add on the line.
            srcline = srcline.rstrip()
            pysrc_lines.append('    '*len(indents)+srcline)
            if srcline.endswith(':'):
                indents.append(indent)
        
    if debug:
        pysrc_lines.append('    except Exception,e:')
        pysrc_lines.append('        pysrc, func_name = __debug ')
        pysrc_lines.append('        lineno = sys.exc_info()[2].tb_lineno')
        pysrc_lines.append('        print ("Exception in template %s() on "')
        pysrc_lines.append('               "line %d:" % (func_name, lineno))')
        pysrc_lines.append('        print pysrc[lineno-1]')
        pysrc_lines.append('        raise')
        
    pysrc = '\n'.join(pysrc_lines)+'\n'
    #log.debug(pysrc)
    if debug: localdict = {'__debug': (pysrc_lines, func_name)}
    else: localdict = {}
    try: exec pysrc in globals(), localdict
    except SyntaxError:
        log.error('Error in script:\n' + pysrc + '\n')
        raise
    template_func = localdict[func_name]
    template_func.__doc__ = docstring
    return template_func
    
def strip_indent(s):
    """
    Given a multiline string C{s}, find the minimum indentation for
    all non-blank lines, and return a new string formed by stripping
    that amount of indentation from all lines in C{s}.
    """
    # Strip indentation from the template.
    minindent = sys.maxint
    lines = s.split('\n')
    for line in lines:
        stripline = line.lstrip()
        if stripline:
            minindent = min(minindent, len(line)-len(stripline))
    return '\n'.join([l[minindent:] for l in lines])

######################################################################
## HTML Writer
######################################################################

class HTMLWriter:
    #////////////////////////////////////////////////////////////
    # Table of Contents
    #////////////////////////////////////////////////////////////
    #
    # 1. Interface Methods
    #
    # 2. Page Generation -- write complete web page files
    #   2.1. Module Pages
    #   2.2. Class Pages
    #   2.3. Trees Page
    #   2.4. Indices Page
    #   2.5. Help Page
    #   2.6. Frames-based table of contents pages
    #   2.7. Homepage (index.html)
    #   2.8. CSS Stylesheet
    #   2.9. Javascript file
    #   2.10. Graphs
    #   2.11. Images
    #
    # 3. Page Element Generation -- write pieces of a web page file
    #   3.1. Page Header
    #   3.2. Page Footer
    #   3.3. Navigation Bar
    #   3.4. Breadcrumbs
    #   3.5. Summary Tables
    #
    # 4. Helper functions

    def __init__(self, docindex, **kwargs):
        """
        Construct a new HTML writer, using the given documentation
        index.
        
        @param docindex: The documentation index.
        
        @type prj_name: C{string}
        @keyword prj_name: The name of the project.  Defaults to
              none.
        @type prj_url: C{string}
        @keyword prj_url: The target for the project hopeage link on
              the navigation bar.  If C{prj_url} is not specified,
              then no hyperlink is created.
        @type prj_link: C{string}
        @keyword prj_link: The label for the project link on the
              navigation bar.  This link can contain arbitrary HTML
              code (e.g. images).  By default, a label is constructed
              from C{prj_name}.
        @type top_page: C{string}
        @keyword top_page: The top page for the documentation.  This
              is the default page shown main frame, when frames are
              enabled.  C{top} can be a URL, the name of a
              module, the name of a class, or one of the special
              strings C{"trees.html"}, C{"indices.html"}, or
              C{"help.html"}.  By default, the top-level package or
              module is used, if there is one; otherwise, C{"trees"}
              is used.
        @type css: C{string}
        @keyword css: The CSS stylesheet file.  If C{css} is a file
              name, then the specified file's conents will be used.
              Otherwise, if C{css} is the name of a CSS stylesheet in
              L{epydoc.docwriter.html_css}, then that stylesheet will
              be used.  Otherwise, an error is reported.  If no stylesheet 
              is specified, then the default stylesheet is used.
        @type help_file: C{string}
        @keyword help_file: The name of the help file.  If no help file is
              specified, then the default help file will be used.
        @type show_private: C{boolean}
        @keyword show_private: Whether to create documentation for
            private objects.  By default, private objects are documented.
        @type show_frames: C{boolean})
        @keyword show_frames: Whether to create a frames-based table of
              contents.  By default, it is produced.
        @type show_imports: C{boolean}
        @keyword show_imports: Whether or not to display lists of
              imported functions and classes.  By default, they are
              not shown.
        @type variable_maxlines: C{int}
        @keyword variable_maxlines: The maximum number of lines that
              should be displayed for the value of a variable in the
              variable details section.  By default, 8 lines are
              displayed.
        @type variable_linelength: C{int}
        @keyword variable_linelength: The maximum line length used for
              displaying the values of variables in the variable
              details sections.  If a line is longer than this length,
              then it will be wrapped to the next line.  The default
              line length is 70 characters.
        @type variable_summary_linelength: C{int}
        @keyword variable_summary_linelength: The maximum line length
              used for displaying the values of variables in the summary
              section.  If a line is longer than this length, then it
              will be truncated.  The default is 40 characters.
        @type variable_tooltip_linelength: C{int}
        @keyword variable_tooltip_linelength: The maximum line length
              used for tooltips for the values of variables.  If a
              line is longer than this length, then it will be
              truncated.  The default is 600 characters.
        @type property_function_linelength: C{int}
        @keyword property_function_linelength: The maximum line length
              used to dispaly property functions (C{fget}, C{fset}, and
              C{fdel}) that contain something other than a function
              object.  The default length is 40 characters.
        @type inheritance: C{string}
        @keyword inheritance: How inherited objects should be displayed.
              If C{inheritance='grouped'}, then inherited objects are
              gathered into groups; if C{inheritance='listed'}, then
              inherited objects are listed in a short list at the
              end of their group; if C{inheritance='included'}, then
              inherited objects are mixed in with non-inherited
              objects.  The default is 'grouped'.
        @type include_source_code: C{boolean}
        @keyword include_source_code: If true, then generate colorized
              source code files for each python module.
        @type include_log: C{boolean}
        @keyword include_log: If true, the the footer will include an
              href to the page 'epydoc-log.html'.
        @type src_code_tab_width: C{int}
        @keyword src_code_tab_width: Number of spaces to replace each tab
            with in source code listings.
        """
        self.docindex = docindex

        # Process keyword arguments.
        self._show_private = kwargs.get('show_private', 1)
        """Should private docs be included?"""
        
        self._prj_name = kwargs.get('prj_name', None)
        """The project's name (for the project link in the navbar)"""
        
        self._prj_url = kwargs.get('prj_url', None)
        """URL for the project link in the navbar"""
        
        self._prj_link = kwargs.get('prj_link', None)
        """HTML code for the project link in the navbar"""
        
        self._top_page = kwargs.get('top_page', None)
        """The 'main' page"""

        self._css = kwargs.get('css')
        """CSS stylesheet to use"""
        
        self._helpfile = kwargs.get('help_file', None)
        """Filename of file to extract help contents from"""
        
        self._frames_index = kwargs.get('show_frames', 1)
        """Should a frames index be created?"""
        
        self._show_imports = kwargs.get('show_imports', False)
        """Should imports be listed?"""
        
        self._propfunc_linelen = kwargs.get('property_function_linelength', 40)
        """[XXX] Not used!"""
        
        self._variable_maxlines = kwargs.get('variable_maxlines', 8)
        """Max lines for variable values"""
        
        self._variable_linelen = kwargs.get('variable_linelength', 70)
        """Max line length for variable values"""
        
        self._variable_summary_linelen = \
                         kwargs.get('variable_summary_linelength', 65)
        """Max length for variable value summaries"""
        
        self._variable_tooltip_linelen = \
                         kwargs.get('variable_tooltip_linelength', 600)
        """Max length for variable tooltips"""
        
        self._inheritance = kwargs.get('inheritance', 'listed')
        """How should inheritance be displayed?  'listed', 'included',
        or 'grouped'"""

        self._incl_sourcecode = kwargs.get('include_source_code', True)
        """Should pages be generated for source code of modules?"""

        self._mark_docstrings = kwargs.get('mark_docstrings', False)
        """Wrap <span class='docstring'>...</span> around docstrings?"""

        self._graph_types = kwargs.get('graphs', ()) or ()
        """Graphs that we should include in our output."""

        self._include_log = kwargs.get('include_log', False)
        """Are we generating an HTML log page?"""

        self._src_code_tab_width = kwargs.get('src_code_tab_width', 8)
        """Number of spaces to replace each tab with in source code
        listings."""
        
        self._callgraph_cache = {}
        """Map the callgraph L{uid<DotGraph.uid>} to their HTML
        representation."""

        self._redundant_details = kwargs.get('redundant_details', False)
        """If true, then include objects in the details list even if all
        info about them is already provided by the summary table."""

        # For use with select_variables():
        if self._show_private:
            self._public_filter = None
        else:
            self._public_filter = True
        
        # Make sure inheritance has a sane value.
        if self._inheritance not in ('listed', 'included', 'grouped'):
            raise ValueError, 'Bad value for inheritance'

        # Create the project homepage link, if it was not specified.
        if (self._prj_name or self._prj_url) and not self._prj_link:
            self._prj_link = plaintext_to_html(self._prj_name or
                                               'Project Homepage')

        # Add a hyperlink to _prj_url, if _prj_link doesn't already
        # contain any hyperlinks.
        if (self._prj_link and self._prj_url and
            not re.search(r'<a[^>]*\shref', self._prj_link)):
            self._prj_link = ('<a class="navbar" target="_top" href="'+
                              self._prj_url+'">'+self._prj_link+'</a>')

        # Precompute lists & sets of APIDoc objects that we're
        # interested in.
        self.valdocs = valdocs = sorted(docindex.reachable_valdocs(
            imports=False, packages=False, bases=False, submodules=False, 
            subclasses=False, private=self._show_private))
        self.module_list = [d for d in valdocs if isinstance(d, ModuleDoc)]
        """The list of L{ModuleDoc}s for the documented modules."""
        self.module_set = set(self.module_list)
        """The set of L{ModuleDoc}s for the documented modules."""
        self.class_list = [d for d in valdocs if isinstance(d, ClassDoc)]
        """The list of L{ClassDoc}s for the documented classes."""
        self.class_set = set(self.class_list)
        """The set of L{ClassDoc}s for the documented classes."""
        self.routine_list = [d for d in valdocs if isinstance(d, RoutineDoc)]
        """The list of L{RoutineDoc}s for the documented routines."""
        self.indexed_docs = []
        """The list of L{APIDoc}s for variables and values that should
        be included in the index."""

        # URL for 'trees' page
        if self.module_list: self._trees_url = 'module-tree.html'
        else: self._trees_url = 'class-tree.html'

        # Construct the value for self.indexed_docs.
        self.indexed_docs += [d for d in valdocs
                              if not isinstance(d, GenericValueDoc)]
        for doc in valdocs:
            if isinstance(doc, NamespaceDoc):
                # add any vars with generic values; but don't include
                # inherited vars.
                self.indexed_docs += [d for d in doc.variables.values() if
                                      isinstance(d.value, GenericValueDoc)
                                      and d.container == doc]
        self.indexed_docs.sort()

        # Figure out the url for the top page.
        self._top_page_url = self._find_top_page(self._top_page)

        # Decide whether or not to split the identifier index.
        self._split_ident_index = (len(self.indexed_docs) >=
                                   self.SPLIT_IDENT_INDEX_SIZE)
        
        # Figure out how many output files there will be (for progress
        # reporting).
        self.modules_with_sourcecode = set()
        for doc in self.module_list:
            if isinstance(doc, ModuleDoc) and is_src_filename(doc.filename):
                self.modules_with_sourcecode.add(doc)
        self._num_files = (len(self.class_list) + len(self.module_list) +
                           10 + len(self.METADATA_INDICES))
        if self._frames_index:
            self._num_files += len(self.module_list) + 3

        if self._incl_sourcecode:
            self._num_files += len(self.modules_with_sourcecode)
        if self._split_ident_index:
            self._num_files += len(self.LETTERS)
            
    def _find_top_page(self, pagename):
        """
        Find the top page for the API documentation.  This page is
        used as the default page shown in the main frame, when frames
        are used.  When frames are not used, this page is copied to 
        C{index.html}.

        @param pagename: The name of the page, as specified by the
            keyword argument C{top} to the constructor.
        @type pagename: C{string}
        @return: The URL of the top page.
        @rtype: C{string}
        """
        # If a page name was specified, then we need to figure out
        # what it points to.
        if pagename:
            # If it's a URL, then use it directly.
            if pagename.lower().startswith('http:'):
                return pagename

            # If it's an object, then use that object's page.
            try:
                doc = self.docindex.get_valdoc(pagename)
                return self.url(doc)
            except:
                pass

            # Otherwise, give up.
            log.warning('Could not find top page %r; using %s '
                        'instead' % (pagename, self._trees_url))
            return self._trees_url

        # If no page name was specified, then try to choose one
        # automatically.
        else:
            root = [val_doc for val_doc in self.docindex.root
                    if isinstance(val_doc, (ClassDoc, ModuleDoc))]
            if len(root) == 0:
                # No docs??  Try the trees page.
                return self._trees_url
            elif len(root) == 1:
                # One item in the root; use that.
                return self.url(root[0]) 
            else:
                # Multiple root items; if they're all in one package,
                # then use that.  Otherwise, use self._trees_url
                root = sorted(root, key=lambda v:len(v.canonical_name))
                top = root[0]
                for doc in root[1:]:
                    if not top.canonical_name.dominates(doc.canonical_name):
                        return self._trees_url
                else:
                    return self.url(top)
    
    #////////////////////////////////////////////////////////////
    #{ 1. Interface Methods
    #////////////////////////////////////////////////////////////

    def write(self, directory=None):
        """
        Write the documentation to the given directory.

        @type directory: C{string}
        @param directory: The directory to which output should be
            written.  If no directory is specified, output will be
            written to the current directory.  If the directory does
            not exist, it will be created.
        @rtype: C{None}
        @raise OSError: If C{directory} cannot be created.
        @raise OSError: If any file cannot be created or written to.
        """
        # For progress reporting:
        self._files_written = 0.
        
        # Set the default values for ValueDoc formatted representations.
        orig_valdoc_defaults = (ValueDoc.SUMMARY_REPR_LINELEN,
                                ValueDoc.REPR_LINELEN,
                                ValueDoc.REPR_MAXLINES)
        ValueDoc.SUMMARY_REPR_LINELEN = self._variable_summary_linelen
        ValueDoc.REPR_LINELEN = self._variable_linelen
        ValueDoc.REPR_MAXLINES = self._variable_maxlines

        # Use an image for the crarr symbol.
        from epydoc.markup.epytext import ParsedEpytextDocstring
        orig_crarr_html = ParsedEpytextDocstring.SYMBOL_TO_HTML['crarr']
        ParsedEpytextDocstring.SYMBOL_TO_HTML['crarr'] = (
            r'<span class="variable-linewrap">'
            r'<img src="crarr.png" alt="\" /></span>')

        # Keep track of failed xrefs, and report them at the end.
        self._failed_xrefs = {}

        # Create destination directories, if necessary
        if not directory: directory = os.curdir
        self._mkdir(directory)
        self._directory = directory

        # Write the CSS file.
        self._files_written += 1
        log.progress(self._files_written/self._num_files, 'epydoc.css')
        self.write_css(directory, self._css)

        # Write the Javascript file.
        self._files_written += 1
        log.progress(self._files_written/self._num_files, 'epydoc.js')
        self.write_javascript(directory)

        # Write images
        self.write_images(directory)

        # Build the indices.
        indices = {'ident': self.build_identifier_index(),
                   'term': self.build_term_index()}
        for (name, label, label2) in self.METADATA_INDICES:
            indices[name] = self.build_metadata_index(name)

        # Write the identifier index.  If requested, split it into
        # separate pages for each letter.
        ident_by_letter = self._group_by_letter(indices['ident'])
        if not self._split_ident_index:
            self._write(self.write_link_index, directory,
                        'identifier-index.html', indices,
                        'Identifier Index', 'identifier-index.html',
                        ident_by_letter)
        else:
            # Write a page for each section.
            for letter in self.LETTERS:
                filename = 'identifier-index-%s.html' % letter
                self._write(self.write_link_index, directory, filename,
                            indices, 'Identifier Index', filename,
                            ident_by_letter, [letter],
                            'identifier-index-%s.html')
            # Use the first non-empty section as the main index page.
            for letter in self.LETTERS:
                if letter in ident_by_letter:
                    filename = 'identifier-index.html'
                    self._write(self.write_link_index, directory, filename,
                                indices, 'Identifier Index', filename,
                                ident_by_letter, [letter], 
                                'identifier-index-%s.html')
                    break

        # Write the term index.
        if indices['term']:
            term_by_letter = self._group_by_letter(indices['term'])
            self._write(self.write_link_index, directory, 'term-index.html',
                        indices, 'Term Definition Index',
                        'term-index.html', term_by_letter)
        else:
            self._files_written += 1 # (skipped)

        # Write the metadata indices.
        for (name, label, label2) in self.METADATA_INDICES:
            if indices[name]:
                self._write(self.write_metadata_index, directory,
                            '%s-index.html' % name, indices, name,
                            label, label2)
            else:
                self._files_written += 1 # (skipped)

        # Write the trees file (package & class hierarchies)
        if self.module_list:
            self._write(self.write_module_tree, directory, 'module-tree.html')
        else:
            self._files_written += 1 # (skipped)
        if self.class_list:
            self._write(self.write_class_tree, directory, 'class-tree.html')
        else:
            self._files_written += 1 # (skipped)
        
        # Write the help file.
        self._write(self.write_help, directory,'help.html')
        
        # Write the frames-based table of contents.
        if self._frames_index:
            self._write(self.write_frames_index, directory, 'frames.html')
            self._write(self.write_toc, directory, 'toc.html')
            self._write(self.write_project_toc, directory, 'toc-everything.html')
            for doc in self.module_list:
                filename = 'toc-%s' % urllib.unquote(self.url(doc))
                self._write(self.write_module_toc, directory, filename, doc)

        # Write the object documentation.
        for doc in self.module_list:
            filename = urllib.unquote(self.url(doc))
            self._write(self.write_module, directory, filename, doc)
        for doc in self.class_list:
            filename = urllib.unquote(self.url(doc))
            self._write(self.write_class, directory, filename, doc)

        # Write source code files.
        if self._incl_sourcecode:
            # Build a map from short names to APIDocs, used when
            # linking names in the source code.
            name_to_docs = {}
            for api_doc in self.indexed_docs:
                if (api_doc.canonical_name is not None and
                    self.url(api_doc) is not None):
                    name = api_doc.canonical_name[-1]
                    name_to_docs.setdefault(name, []).append(api_doc)
            # Sort each entry of the name_to_docs list.
            for doc_list in name_to_docs.values():
                doc_list.sort()
            # Write the source code for each module.
            for doc in self.modules_with_sourcecode:
                filename = urllib.unquote(self.pysrc_url(doc))
                self._write(self.write_sourcecode, directory, filename, doc,
                            name_to_docs)

        # Write the auto-redirect page.
        self._write(self.write_redirect_page, directory, 'redirect.html')

        # Write the mapping object name -> URL
        self._write(self.write_api_list, directory, 'api-objects.txt')
        
        # Write the index.html files.
        # (this must be done last, since it might copy another file)
        self._files_written += 1
        log.progress(self._files_written/self._num_files, 'index.html')
        self.write_homepage(directory)

        # Don't report references to builtins as missing
        for k in self._failed_xrefs.keys(): # have a copy of keys
            if hasattr(__builtin__, k):
                del self._failed_xrefs[k]

        # Report any failed crossreferences
        if self._failed_xrefs:
            estr = 'Failed identifier crossreference targets:\n'
            failed_identifiers = self._failed_xrefs.keys()
            failed_identifiers.sort()
            for identifier in failed_identifiers:
                names = self._failed_xrefs[identifier].keys()
                names.sort()
                estr += '- %s' % identifier
                estr += '\n'
                for name in names:
                    estr += '      (from %s)\n' % name
            log.docstring_warning(estr)

        # [xx] testing:
        if self._num_files != int(self._files_written):
            log.debug("Expected to write %d files, but actually "
                      "wrote %d files" %
                      (self._num_files, int(self._files_written)))

        # Restore defaults that we changed.
        (ValueDoc.SUMMARY_REPR_LINELEN, ValueDoc.REPR_LINELEN,
         ValueDoc.REPR_MAXLINES) = orig_valdoc_defaults
        ParsedEpytextDocstring.SYMBOL_TO_HTML['crarr'] = orig_crarr_html

    def _write(self, write_func, directory, filename, *args):
        # Display our progress.
        self._files_written += 1
        log.progress(self._files_written/self._num_files, filename)
        
        path = os.path.join(directory, filename)
        f = codecs.open(path, 'w', 'ascii', errors='xmlcharrefreplace')
        write_func(f.write, *args)
        f.close()

    def _mkdir(self, directory):
        """
        If the given directory does not exist, then attempt to create it.
        @rtype: C{None}
        """
        if not os.path.isdir(directory):
            if os.path.exists(directory):
                raise OSError('%r is not a directory' % directory)
            os.mkdir(directory)
        
    #////////////////////////////////////////////////////////////
    #{ 2.1. Module Pages
    #////////////////////////////////////////////////////////////

    def write_module(self, out, doc):
        """
        Write an HTML page containing the API documentation for the
        given module to C{out}.
        
        @param doc: A L{ModuleDoc} containing the API documentation
        for the module that should be described.
        """
        longname = doc.canonical_name
        shortname = doc.canonical_name[-1]

        # Write the page header (incl. navigation bar & breadcrumbs)
        self.write_header(out, str(longname))
        self.write_navbar(out, doc)
        self.write_breadcrumbs(out, doc, self.url(doc))

        # Write the name of the module we're describing.
        if doc.is_package is True: typ = 'Package'
        else: typ = 'Module'
        if longname[0].startswith('script-'):
            shortname = str(longname)[7:]
            typ = 'Script'
        out('<!-- ==================== %s ' % typ.upper() +
            'DESCRIPTION ==================== -->\n')
        out('<h1 class="epydoc">%s %s</h1>' % (typ, shortname))
        out('<p class="nomargin-top">%s</p>\n' % self.pysrc_link(doc))
        
        # If the module has a description, then list it.
        if doc.descr not in (None, UNKNOWN):
            out(self.descr(doc, 2)+'\n\n')

        # Write any standarad metadata (todo, author, etc.)
        if doc.metadata is not UNKNOWN and doc.metadata:
            out('<hr />\n')
        self.write_standard_fields(out, doc)

        # If it's a package, then list the modules it contains.
        if doc.is_package is True:
            self.write_module_list(out, doc)

        # Write summary tables describing the variables that the
        # module defines.
        self.write_summary_table(out, "Classes", doc, "class")
        self.write_summary_table(out, "Functions", doc, "function")
        self.write_summary_table(out, "Variables", doc, "other")

        # Write a list of all imported objects.
        if self._show_imports:
            self.write_imports(out, doc)

        # Write detailed descriptions of functions & variables defined
        # in this module.
        self.write_details_list(out, "Function Details", doc, "function")
        self.write_details_list(out, "Variables Details", doc, "other")

        # Write the page footer (including navigation bar)
        self.write_navbar(out, doc)
        self.write_footer(out)

    #////////////////////////////////////////////////////////////
    #{ 2.??. Source Code Pages
    #////////////////////////////////////////////////////////////

    def write_sourcecode(self, out, doc, name_to_docs):
        #t0 = time.time()
        
        filename = doc.filename
        name = str(doc.canonical_name)
        
        # Header
        self.write_header(out, name)
        self.write_navbar(out, doc)
        self.write_breadcrumbs(out, doc, self.pysrc_url(doc))

        # Source code listing
        out('<h1 class="epydoc">Source Code for %s</h1>\n' %
            self.href(doc, label='%s %s' % (self.doc_kind(doc), name)))
        out('<pre class="py-src">\n')
        out(PythonSourceColorizer(filename, name, self.docindex,
                                  self.url, name_to_docs,
                                  self._src_code_tab_width).colorize())
        out('</pre>\n<br />\n')

        # Footer
        self.write_navbar(out, doc)
        self.write_footer(out)
        
        #log.debug('[%6.2f sec] Wrote pysrc for %s' %
        #          (time.time()-t0, name))

    #////////////////////////////////////////////////////////////
    #{ 2.2. Class Pages
    #////////////////////////////////////////////////////////////

    def write_class(self, out, doc):
        """
        Write an HTML page containing the API documentation for the
        given class to C{out}.
        
        @param doc: A L{ClassDoc} containing the API documentation
        for the class that should be described.
        """
        longname = doc.canonical_name
        shortname = doc.canonical_name[-1]

        # Write the page header (incl. navigation bar & breadcrumbs)
        self.write_header(out, str(longname))
        self.write_navbar(out, doc)
        self.write_breadcrumbs(out, doc, self.url(doc))

        # Write the name of the class we're describing.
        if doc.is_type(): typ = 'Type'
        elif doc.is_exception(): typ = 'Exception'
        else: typ = 'Class'
        out('<!-- ==================== %s ' % typ.upper() +
            'DESCRIPTION ==================== -->\n')
        out('<h1 class="epydoc">%s %s</h1>' % (typ, shortname))
        out('<p class="nomargin-top">%s</p>\n' % self.pysrc_link(doc))

        if ((doc.bases not in (UNKNOWN, None) and len(doc.bases) > 0) or
            (doc.subclasses not in (UNKNOWN,None) and len(doc.subclasses)>0)):
            # Display bases graphically, if requested.
            if 'umlclasstree' in self._graph_types:
                self.write_class_tree_graph(out, doc, uml_class_tree_graph)
            elif 'classtree' in self._graph_types:
                self.write_class_tree_graph(out, doc, class_tree_graph)
                
            # Otherwise, use ascii-art.
            else:
                # Write the base class tree.
                if doc.bases not in (UNKNOWN, None) and len(doc.bases) > 0:
                    out('<pre class="base-tree">\n%s</pre>\n\n' %
                        self.base_tree(doc))

                # Write the known subclasses
                if (doc.subclasses not in (UNKNOWN, None) and
                    len(doc.subclasses) > 0):
                    out('<dl><dt>Known Subclasses:</dt>\n<dd>\n    ')
                    out('  <ul class="subclass-list">\n')
                    for i, subclass in enumerate(doc.subclasses):
                        href = self.href(subclass, context=doc)
                        if self._val_is_public(subclass): css = ''
                        else: css = ' class="private"'
                        if i > 0: href = ', '+href
                        out('<li%s>%s</li>' % (css, href))
                    out('  </ul>\n')
                    out('</dd></dl>\n\n')

            out('<hr />\n')
        
        # If the class has a description, then list it.
        if doc.descr not in (None, UNKNOWN):
            out(self.descr(doc, 2)+'\n\n')

        # Write any standarad metadata (todo, author, etc.)
        if doc.metadata is not UNKNOWN and doc.metadata:
            out('<hr />\n')
        self.write_standard_fields(out, doc)

        # Write summary tables describing the variables that the
        # class defines.
        self.write_summary_table(out, "Nested Classes", doc, "class")
        self.write_summary_table(out, "Instance Methods", doc,
                                 "instancemethod")
        self.write_summary_table(out, "Class Methods", doc, "classmethod")
        self.write_summary_table(out, "Static Methods", doc, "staticmethod")
        self.write_summary_table(out, "Class Variables", doc,
                                 "classvariable")
        self.write_summary_table(out, "Instance Variables", doc,
                                 "instancevariable")
        self.write_summary_table(out, "Properties", doc, "property")

        # Write a list of all imported objects.
        if self._show_imports:
            self.write_imports(out, doc)

        # Write detailed descriptions of functions & variables defined
        # in this class.
        # [xx] why group methods into one section but split vars into two?
        # seems like we should either group in both cases or split in both
        # cases.
        self.write_details_list(out, "Method Details", doc, "method")
        self.write_details_list(out, "Class Variable Details", doc,
                                "classvariable")
        self.write_details_list(out, "Instance Variable Details", doc,
                                "instancevariable")
        self.write_details_list(out, "Property Details", doc, "property")

        # Write the page footer (including navigation bar)
        self.write_navbar(out, doc)
        self.write_footer(out)

    def write_class_tree_graph(self, out, doc, graphmaker):
        """
        Write HTML code for a class tree graph of C{doc} (a classdoc),
        using C{graphmaker} to draw the actual graph.  C{graphmaker}
        should be L{class_tree_graph()}, or L{uml_class_tree_graph()},
        or any other function with a compatible signature.

        If the given class has any private sublcasses (including
        recursive subclasses), then two graph images will be generated
        -- one to display when private values are shown, and the other
        to display when private values are hidden.
        """
        linker = _HTMLDocstringLinker(self, doc)
        private_subcls = self._private_subclasses(doc)
        if private_subcls:
            out('<center>\n'
                '  <div class="private">%s</div>\n'
                '  <div class="public" style="display:none">%s</div>\n'
                '</center>\n' %
                (self.render_graph(graphmaker(doc, linker, doc)),
                 self.render_graph(graphmaker(doc, linker, doc,
                                              exclude=private_subcls))))
        else:
            out('<center>\n%s\n</center>\n' %
                self.render_graph(graphmaker(doc, linker, doc)))

    #////////////////////////////////////////////////////////////
    #{ 2.3. Trees pages
    #////////////////////////////////////////////////////////////

    def write_module_tree(self, out):
        # Header material
        self.write_treepage_header(out, 'Module Hierarchy', 'module-tree.html')
        out('<h1 class="epydoc">Module Hierarchy</h1>\n')

        # Write entries for all top-level modules/packages.
        out('<ul class="nomargin-top">\n')
        for doc in self.module_list:
            if (doc.package in (None, UNKNOWN) or
                doc.package not in self.module_set):
                self.write_module_tree_item(out, doc)
        out('</ul>\n')

        # Footer material
        self.write_navbar(out, 'trees')
        self.write_footer(out)

    def write_class_tree(self, out):
        """
        Write HTML code for a nested list showing the base/subclass
        relationships between all documented classes.  Each element of
        the top-level list is a class with no (documented) bases; and
        under each class is listed all of its subclasses.  Note that
        in the case of multiple inheritance, a class may appear
        multiple times.  
        
        @todo: For multiple inheritance, don't repeat subclasses the
            second time a class is mentioned; instead, link to the
            first mention.
        """
        # [XX] backref for multiple inheritance?
        # Header material
        self.write_treepage_header(out, 'Class Hierarchy', 'class-tree.html')
        out('<h1 class="epydoc">Class Hierarchy</h1>\n')

        # Build a set containing all classes that we should list.
        # This includes everything in class_list, plus any of those
        # class' bases, but not undocumented subclasses.
        class_set = self.class_set.copy()
        for doc in self.class_list:
            if doc.bases != UNKNOWN:
                for base in doc.bases:
                    if base not in class_set:
                        if isinstance(base, ClassDoc):
                            class_set.update(base.mro())
                        else:
                            # [XX] need to deal with this -- how?
                            pass
                            #class_set.add(base)
 
        out('<ul class="nomargin-top">\n')
        for doc in sorted(class_set, key=lambda c:c.canonical_name[-1]):
            if doc.bases != UNKNOWN and len(doc.bases)==0:
                self.write_class_tree_item(out, doc, class_set)
        out('</ul>\n')
        
        # Footer material
        self.write_navbar(out, 'trees')
        self.write_footer(out)

    def write_treepage_header(self, out, title, url):
        # Header material.
        self.write_header(out, title)
        self.write_navbar(out, 'trees')
        self.write_breadcrumbs(out, 'trees', url)
        if self.class_list and self.module_list:
            out('<center><b>\n')
            out(' [ <a href="module-tree.html">Module Hierarchy</a>\n')
            out(' | <a href="class-tree.html">Class Hierarchy</a> ]\n')
            out('</b></center><br />\n')


    #////////////////////////////////////////////////////////////
    #{ 2.4. Index pages
    #////////////////////////////////////////////////////////////

    SPLIT_IDENT_INDEX_SIZE = 3000
    """If the identifier index has more than this number of entries,
    then it will be split into separate pages, one for each
    alphabetical section."""

    LETTERS = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ_'
    """The alphabetical sections that are used for link index pages."""
    
    def write_link_index(self, out, indices, title, url, index_by_section,
                         sections=LETTERS, section_url='#%s'):
        
        # Header
        self.write_indexpage_header(out, indices, title, url)

        # Index title & links to alphabetical sections.
        out('<table border="0" width="100%">\n'
            '<tr valign="bottom"><td>\n')
        out('<h1 class="epydoc">%s</h1>\n</td><td>\n[\n' % title)
        for sec in self.LETTERS:
            if sec in index_by_section:
                out(' <a href="%s">%s</a>\n' % (section_url % sec, sec))
            else:
                out('  %s\n' % sec)
        out(']\n')
        out('</td></table>\n')

        # Alphabetical sections.
        sections = [s for s in sections if s in index_by_section]
        if sections:
            out('<table border="0" width="100%">\n')
            for section in sorted(sections):
                out('<tr valign="top"><td valign="top" width="1%">')
                out('<h2 class="epydoc"><a name="%s">%s</a></h2></td>\n' %
                    (section, section))
                out('<td valign="top">\n')
                self.write_index_section(out, index_by_section[section], True)
                out('</td></tr>\n')
            out('</table>\n<br />')

        # Footer material.
        out('<br />')
        self.write_navbar(out, 'indices')
        self.write_footer(out)
                
        
    def write_metadata_index(self, out, indices, field, title, typ):
        """
        Write an HTML page containing a metadata index.
        """
        index = indices[field]
        
        # Header material.
        self.write_indexpage_header(out, indices, title,
                                    '%s-index.html' % field)

        # Page title.
        out('<h1 class="epydoc"><a name="%s">%s</a></h1>\n<br />\n' %
            (field, title))

        # Index (one section per arg)
        for arg in sorted(index):
            # Write a section title.
            if arg is not None:
                if len([1 for (doc, descrs) in index[arg] if
                        not self._doc_or_ancestor_is_private(doc)]) == 0:
                    out('<div class="private">')
                else:
                    out('<div>')
                self.write_table_header(out, 'metadata-index', arg)
                out('</table>')
            # List every descr for this arg.
            for (doc, descrs) in index[arg]:
                if self._doc_or_ancestor_is_private(doc):
                    out('<div class="private">\n')
                else:
                    out('<div>\n')
                out('<table width="100%" class="metadata-index" '
                    'bgcolor="#e0e0e0"><tr><td class="metadata-index">')
                out('<b>%s in %s</b>' %
                    (typ, self.href(doc, label=doc.canonical_name)))
                out('    <ul class="nomargin">\n')
                for descr in descrs:
                    out('      <li>%s</li>\n' %
                        self.docstring_to_html(descr,doc,4))
                out('    </ul>\n')
                out('</table></div>\n')

        # Footer material.
        out('<br />')
        self.write_navbar(out, 'indices')
        self.write_footer(out)

    def write_indexpage_header(self, out, indices, title, url):
        """
        A helper for the index page generation functions, which
        generates a header that can be used to navigate between the
        different indices.
        """
        self.write_header(out, title)
        self.write_navbar(out, 'indices')
        self.write_breadcrumbs(out, 'indices', url)

        if (indices['term'] or
            [1 for (name,l,l2) in self.METADATA_INDICES if indices[name]]):
            out('<center><b>[\n')
            out(' <a href="identifier-index.html">Identifiers</a>\n')
            if indices['term']:
                out('| <a href="term-index.html">Term Definitions</a>\n')
            for (name, label, label2) in self.METADATA_INDICES:
                if indices[name]:
                    out('| <a href="%s-index.html">%s</a>\n' %
                        (name, label2))
            out(']</b></center><br />\n')

    def write_index_section(self, out, items, add_blankline=False):
        out('<table class="link-index" width="100%" border="1">\n')
        num_rows = (len(items)+2)/3
        for row in range(num_rows):
            out('<tr>\n')
            for col in range(3):
                out('<td width="33%" class="link-index">')
                i = col*num_rows+row
                if i < len(items):
                    name, url, container = items[col*num_rows+row]
                    out('<a href="%s">%s</a>' % (url, name))
                    if container is not None:
                        out('<br />\n')
                        if isinstance(container, ModuleDoc):
                            label = container.canonical_name
                        else:
                            label = container.canonical_name[-1]
                        out('<span class="index-where">(in&nbsp;%s)'
                            '</span>' % self.href(container, label))
                else:
                    out('&nbsp;')
                out('</td>\n')
            out('</tr>\n')
            if add_blankline and num_rows == 1:
                blank_cell = '<td class="link-index">&nbsp;</td>'
                out('<tr>'+3*blank_cell+'</tr>\n')
        out('</table>\n')

    #////////////////////////////////////////////////////////////
    #{ 2.5. Help Page
    #////////////////////////////////////////////////////////////

    def write_help(self, out):
        """
        Write an HTML help file to the given stream.  If
        C{self._helpfile} contains a help file, then use it;
        otherwise, use the default helpfile from
        L{epydoc.docwriter.html_help}.
        """
        # todo: optionally parse .rst etc…