]> rtime.felk.cvut.cz Git - opencv.git/blobdiff - opencv/doc/latex2sphinx/latex.py
new test category python_fragments from documentation
[opencv.git] / opencv / doc / latex2sphinx / latex.py
index f9c98e10aba26ac545af40ef7a1839c914be36d2..3edbb2cf3faf7f3dc4fe8bb17672adf4c1aff77e 100644 (file)
 import sys
-from pyparsing import Word, CharsNotIn, Optional, OneOrMore, ZeroOrMore, Group, ParseException, Literal, replaceWith, StringEnd, lineno, QuotedString
-
-class Argument:
-    def __init__(self, s, loc, toks):
-        self.str = toks[1]
-    def __repr__(self):
-        return "[%s]" % self.str
-def argfun(s, loc, toks):
-    return Argument(s, loc, toks)
-
-class Parameter:
-    def __init__(self, s, loc, toks):
-        self.str = toks[0]
-    def __repr__(self):
-        return "{%s}" % self.str
-def paramfun(s, loc, toks):
-    return Parameter(s, loc, toks)
-
-class TexCmd:
-    def __init__(self, s, loc, toks):
-        self.cmd = str(toks[0])[1:]
-        #print 'cmd', self.cmd
-        self.args = toks[1].asList()
-        self.params = toks[2].asList()
-    def __repr__(self):
-        return self.cmd + "".join([repr(a) for a in self.args]) + "".join([repr(p) for p in self.params])
-
-class ZeroOrMoreAsList(ZeroOrMore):
-    def __init__(self, *args):
-        ZeroOrMore.__init__(self, *args)
-        def listify(s, loc, toks):
-            return [toks]
-        self.setParseAction(listify)
-
-arg = '[' + CharsNotIn("]") + ']'
-arg.setParseAction(argfun)
-# param = '{' + Optional(CharsNotIn("}")) + '}'
-param = QuotedString('{', endQuoteChar = '}', multiline = True)
-param.setParseAction(paramfun)
-texcmd = Word("\\", "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789") + ZeroOrMoreAsList(arg) + ZeroOrMoreAsList(param)
-def texcmdfun(s, loc, toks):
-    if str(toks[0])[1:] == 'input':
-        filename = "../" + toks[2].asList()[0].str + ".tex"
-        print 'Now parsing', filename, loc
-        return parsefile(filename, lineno(loc, s))
-    else:
-        return TexCmd(s, loc, toks)
-texcmd.setParseAction(texcmdfun)
-
-backslash = chr(92)
-legal = "".join([chr(x) for x in set(range(32, 127)) - set(backslash)])
-filler = CharsNotIn(backslash)
-filler = Word(legal)
-document = ZeroOrMore(texcmd | filler) + StringEnd().suppress()
-
-if 0:
-    print document.parseString('\\hello{foo}')
-    sys.exit(0)
-
-def parsefile(filename, startline):
-    f = open(filename, "rt")
-
-    lines = list(f)
-    def uncomment(s):
-        if '%' in s:
-            return s[:s.index('%')] + '\n'
+from latexparser import latexparser, TexCmd
+import distutils.dep_util
+import os
+import cPickle as pickle
+import pyparsing as pp
+import StringIO
+from qfile import QOpen
+from string import Template
+
+# useful things for pyparsing
+def returnList(x):
+    def listify(s, loc, toks):
+        return [toks]
+    x.setParseAction(listify)
+    return x
+def returnTuple(x):
+    def listify(s, loc, toks):
+        return [tuple(toks)]
+    x.setParseAction(listify)
+    return x
+def CommaList(word):
+    return returnList(pp.Optional(word + pp.ZeroOrMore(pp.Suppress(',') + word)))
+def sl(s):
+    return pp.Suppress(pp.Literal(s))
+
+import pythonapi
+
+python_api = pythonapi.reader("../../interfaces/python/api")
+
+
+class SphinxWriter:
+    def __init__(self, filename, language, abspath):
+        assert language in ['py', 'c', 'cpp']
+        self.language = language
+
+        self.abspath = abspath
+        os.path.abspath(os.path.dirname(filename))
+
+        self.f_index = QOpen(os.path.join(self.language, filename), 'wt')
+        self.f = self.f_index
+        self.f_chapter = None
+        self.f_section = None
+        self.indent = 0
+        self.state = None
+        self.envstack = []
+        self.tags = {}
+        self.errors = open('errors.%s' % language, 'wt')
+        self.unhandled_commands = set()
+        self.freshline = True
+        self.function_props = {}
+        self.covered = set()        # covered functions, used for error report
+        self.description = ""
+
+    def write(self, s):
+        self.freshline = len(s) > 0 and (s[-1] == '\n')
+        self.f.write(s.replace('\n', '\n' + self.indent * "    "))
+
+    def appendspace(self):
+        """ append a space to the output - if we're not at the start of a line """
+        if not self.freshline:
+            self.write(' ')
+
+    def doplain(self, s):
+        if (len(s) > 1) and (s[0] == '$' and s[-1] == '$') and self.state != 'math':
+            s = ":math:`%s`" % s[1:-1].strip()
+        elif self.state != 'math':
+            s.replace('\\_', '_')
+        if len(s) > 0 and s[-1] == '\n':
+            s = s[:-1]
+        if self.state == 'fpreamble':
+            self.description += s
+        else:
+            self.write(s)
+
+    def docmd(self, c):
+        if self.state == 'math':
+            if c.cmd != ']':
+                self.default_cmd(c)
+            else:
+                self.indent -= 1
+                self.state = None
+                self.write('\n\n')
+        else:
+            if c.cmd == '\n':
+                self.write('\\\n')
+            else:
+                if c.cmd == '[':
+                    meth = self.cmd_gomath
+                else:
+                    cname = "cmd_" + c.cmd
+                    meth = getattr(self, cname, self.unrecognized_cmd)
+                meth(c)
+
+    def cmd_gomath(self, c):
+        self.state = 'math'
+        print >>self, "\n\n.. math::"
+        self.indent += 1
+        print >>self
+
+    def cmd_chapter(self, c):
+        filename = str(c.params[0]).lower().replace(' ', '_').replace('/','_')
+        self.f_index.write("    %s\n" % filename)
+        self.f_chapter = QOpen(os.path.join(self.language, filename + '.rst'), 'wt')
+        self.f_section = None
+        self.f = self.f_chapter
+        self.indent = 0
+        title = str(c.params[0])
+        print >>self, '*' * len(title)
+        print >>self, title
+        print >>self, '*' * len(title)
+        print >>self
+        self.chapter_intoc = False
+
+    def cmd_section(self, c):
+        filename = str(c.params[0]).lower().replace(' ', '_').replace('/','_')
+        if not self.chapter_intoc:
+            self.chapter_intoc = True
+            print >>self.f_chapter
+            print >>self.f_chapter, '.. toctree::'
+            print >>self.f_chapter, '    :maxdepth: 2'
+            print >>self.f_chapter
+        self.f_chapter.write("    %s\n" % filename)
+        self.f_section = QOpen(os.path.join(self.language, filename + '.rst'), 'wt')
+        self.f = self.f_section
+        self.indent = 0
+        title = self.render(c.params[0].str)
+        print >>self, title
+        print >>self, '=' * len(title)
+        print >>self
+        print >>self, '.. highlight:: %s' % {'c': 'c', 'cpp': 'cpp', 'py': 'python'}[self.language]
+        print >>self
+
+    def cmd_subsection(self, c):
+        print >>self
+        print >>self, str(c.params[0])
+        print >>self, '-' * len(str(c.params[0]))
+        print >>self
+        self.function_props = {}
+
+    def cmd_includegraphics(self, c):
+        filename = os.path.join('..', '..', str(c.params[0]))
+        print >>self, "\n\n.. image:: %s\n\n" % filename
+
+    def cmd_cvCppCross(self, c):
+        self.write(":ref:`%s`" % str(c.params[0]))
+
+    def cmd_cvCPyCross(self, c):
+        self.write(":ref:`%s`" % str(c.params[0]))
+
+    def cmd_cross(self, c):
+        self.write(":ref:`%s`" % str(c.params[0]))
+
+    def cmd_cvCross(self, c):
+        self.write(":ref:`%s`" % str(c.params[0]))
+
+    def cmd_cvclass(self, c):
+        self.indent = 0
+        self.state = None
+        nm = self.render(list(c.params[0].str))
+        print >>self, "\n.. index:: %s\n" % nm
+        print >>self, ".. _%s:\n" % nm
+        print >>self, nm
+        print >>self, '-' * len(nm)
+        print >>self
+        if self.language == 'py':
+            print >>self, ".. class:: " + nm + "\n"
         else:
-            return s
-
-    lines = [uncomment(l) for l in lines]
-    print len(lines), "lines"
-
-    docstr = "".join(lines)
-    # document.setFailAction(None)
-    try:
-        return document.parseString(docstr)
-    except ParseException, pe:
-        print 'Fatal problem at %s line %d col %d' % (filename, pe.lineno, pe.col)
-        print pe.line
-        sys.exit(1)
-
-doc = parsefile(sys.argv[1], 0)
-
-raw = open('raw', 'w')
-for x in doc:
-    print >>raw, repr(x)
-raw.close()
-
-for x in doc:
-    if isinstance(x, TexCmd):
-        if x.cmd == 'chapter':
-            print repr(x)
-        if x.cmd == 'section':
-            print ' ' * 4, repr(x)
+            print >>self, ".. ctype:: " + nm + "\n"
+        print >>self
+        self.addtag(nm, c)
+        self.state = 'class'
+
+    def addtag(self, nm, c):
+        if nm == "":
+            self.report_error(c, "empty name")
+        self.tags[nm] = "%s\t%s\t%d" % (nm, os.path.join(os.getcwd(), c.filename), c.lineno)
+
+    def cmd_cvfunc(self, c):
+        self.cmd_cvCPyFunc(c)
+
+    def cmd_cvCPyFunc(self, c):
+        self.indent = 0
+        nm = self.render(c.params[0].str)
+        print >>self, "\n.. index:: %s\n" % nm
+        print >>self, ".. _%s:\n" % nm
+        print >>self, nm
+        print >>self, '-' * len(nm)
+        print >>self
+        self.state = 'fpreamble'
+        if self.description != "":
+            self.report_error(c, "overflow - preceding cvfunc (starting %s) not terminated?" % repr(self.description[:30]))
+        self.description = ""
+        self.addtag(nm, c)
+
+        self.function_props = {'name' : nm}
+        self.covered.add(nm)
+
+    def cmd_cvCppFunc(self, c):
+        self.indent = 0
+        nm = self.render(c.params[0].str)
+        print >>self, "\n.. index:: %s\n" % nm
+        print >>self, ".. _%s:\n" % nm
+        print >>self, nm
+        print >>self, '-' * len(nm)
+        print >>self
+        self.state = 'fpreamble'
+        if self.description != "":
+            self.report_error(c, "overflow - preceding cvfunc (starting %s) not terminated?" % repr(self.description[:30]))
+        self.description = ""
+        self.addtag(nm, c)
+
+        self.function_props = {'name' : nm}
+        self.covered.add(nm)
+
+    def cmd_cvdefC(self, c):
+        if self.language != 'c':
+            return
+        s = str(c.params[0]).replace('\\_', '_')
+        s = s.replace('\\par', '')
+        s = s.replace('\n', ' ')
+        s = s.replace(';', '')
+        self.indent = 0
+        for proto in s.split('\\newline'):
+            if proto.strip() != "":
+                print >>self, "\n\n.. cfunction:: " + proto.strip() + "\n"
+        # print >>self, "=", repr(c.params[0].str)
+        print >>self, '    ' + self.description
+        self.description = ""
+        print >>self
+        self.state = None
+        self.function_props['defpy'] = s
+
+    def cmd_cvdefCpp(self, c):
+        if self.language != 'cpp':
+            return
+        s = str(c.params[0]).replace('\\_', '_')
+        s = s.replace('\\par', '')
+        s = s.replace('\n', ' ')
+        s = s.replace(';', '')
+        self.indent = 0
+        for proto in s.split('\\newline'):
+            if proto.strip() != "":
+                print >>self, "\n\n.. cfunction:: " + proto.strip() + "\n"
+        # print >>self, "=", repr(c.params[0].str)
+        if self.description != "":
+            print >>self, '    ' + self.description
+        else:
+            self.report_error(c, 'empty description')
+        self.description = ""
+        print >>self
+        self.state = None
+        self.function_props['defpy'] = s
+
+    def cmd_cvdefPy(self, c):
+        if self.language != 'py':
+            return
+        s = str(c.params[0]).replace('\\_', '_')
+        self.indent = 0
+        print >>self, ".. function:: " + s + "\n"
+        # print >>self, "=", repr(c.params[0].str)
+        print >>self, '    ' + self.description
+        print >>self
+        self.description = ""
+        self.state = None
+        self.function_props['defpy'] = s
+
+        pp.ParserElement.setDefaultWhitespaceChars(" \n\t")
+
+        ident = pp.Word(pp.alphanums + "_.+-")
+        ident_or_tuple = ident | (sl('(') + CommaList(ident) + sl(')'))
+        initializer = ident_or_tuple
+        arg = returnList(ident + pp.Optional(sl('=') + initializer))
+
+        decl = ident + sl('(') + CommaList(arg) + sl(')') + sl("->") + ident_or_tuple + pp.StringEnd()
+
+        try:
+            l = decl.parseString(s)
+            if str(l[0]) != self.function_props['name']:
+                self.report_error(c, 'Decl "%s" does not match function name "%s"' % (str(l[0]), self.function_props['name']))
+            self.function_props['signature'] = l
+            if l[0] in python_api:
+                (ins, outs) = python_api[l[0]]
+                ins = [a for a in ins if not 'O' in a.flags]
+                if outs != None:
+                    outs = outs.split(',')
+                if len(ins) != len(l[1]):
+                    self.report_error(c, "function %s documented arity %d, code arity %d" % (l[0], len(l[1]), len(ins)))
+                if outs == None:
+                    if l[2] != 'None':
+                        self.report_error(c, "function %s documented None, but code has %s" % (l[0], l[2]))
+                else:
+                    if isinstance(l[2], str):
+                        doc_outs = [l[2]]
+                    else:
+                        doc_outs = l[2]
+                    if len(outs) != len(doc_outs):
+                        self.report_error(c, "function %s output documented tuple %d, code %d" % (l[0], len(outs), len(doc_outs)))
+            else:
+                # self.report_error(c, "function %s documented but not found in code" % l[0])
+                pass
+        except pp.ParseException, pe:
+            self.report_error(c, str(pe))
+            print s
+            print pe
+
+    def report_error(self, c, msg):
+        print >>self.errors, "%s:%d: [%s] Error %s" % (c.filename, c.lineno, self.language, msg)
+
+    def cmd_begin(self, c):
+        if len(c.params) == 0:
+            self.report_error(c, "Malformed begin")
+            return
+        self.write('\n')
+        s = str(c.params[0])
+        self.envstack.append((s, (c.filename, c.lineno)))
+        if s == 'description':
+            if self.language == 'py' and 'name' in self.function_props and not 'defpy' in self.function_props:
+                self.report_error(c, "No cvdefPy for function %s" % self.function_props['name'])
+            self.indent += 1
+        elif s == 'lstlisting':
+            # Set indent to zero while collecting code; so later write will not double-indent
+            self.saved_f = self.f
+            self.saved_indent = self.indent
+            self.f = StringIO.StringIO()
+            self.indent = 0
+        elif s in ['itemize', 'enumerate']:
+            self.indent += 1
+        elif s == 'tabular':
+            self.f = StringIO.StringIO()
+        else:
+            self.default_cmd(c)
+
+    def cmd_item(self, c):
+        if len(self.ee()) == 0:
+            self.report_error(c, "item without environment")
+            return
+        self.indent -= 1
+        markup = {'itemize' : '*', 'enumerate' : '#.', 'description' : '*'}[self.ee()[-1]]
+        if len(c.args) > 0:
+            markup += " " + self.render([c.args[0].str])
+        if len(c.params) > 0:
+            markup += " " + self.render(c.params[0].str)
+        self.write("\n\n" + markup)
+        self.indent += 1
+
+    def cmd_end(self, c):
+        if len(c.params) != 1:
+            self.report_error(c, "Malformed end")
+            return
+        if len(self.envstack) == 0:
+            self.report_error(c, "end with no env")
+            return
+        self.write('\n')
+        s = str(c.params[0])
+        if self.envstack == []:
+            print "Cannot pop at", (c.filename, c.lineno)
+        if self.envstack[-1][0] != s:
+            self.report_error(c, "end{%s} does not match current stack %s" % (s, repr(self.envstack)))
+        self.envstack.pop()
+        if s == 'description':
+            self.indent -= 1
+            if self.indent == 0:
+                self.function_props['done'] = True
+        elif s in ['itemize', 'enumerate']:
+            self.indent -= 1
+        elif s == 'tabular':
+            tabletxt = self.f.getvalue()
+            self.f = self.f_section
+            self.f.write(self.handle_table(tabletxt))
+        elif s == 'lstlisting':
+            listing = self.f.getvalue()
+
+            self.f = self.saved_f
+            self.indent = self.saved_indent
+            print >>self
+            if self.language == 'py':
+                ckeys = ['#define', 'void', '#include', ';\n']
+                found = [repr(k) for k in ckeys if k in listing]
+                if len(found) > 0:
+                    self.report_error(c, 'listing is probably C, found %s' % ",".join(found))
+            if (self.language == 'py') and ('>>>' in listing):
+                print >>self, "\n.. doctest::\n"
+            else:
+                print >>self, "\n::\n"
+            self.indent += 1
+            print >>self
+            self.write(listing)
+            self.indent -= 1
+            print >>self
+            print >>self
+            print >>self, ".."      # otherwise a following :param: gets treated as more listing
+        elif s == 'document':
+            pass
+        else:
+            self.default_cmd(c)
+        
+    def cmd_label(self, c):
+        pass
+
+    def cmd_lstinputlisting(self, c):
+        s = str(c.params[0])
+        print >>self.f, ".. include:: %s" % os.path.normpath(os.path.join(self.abspath, s))
+        print >>self.f, "    :literal:"
+        print >>self.f
+
+    # Conditionals
+    def cmd_cvC(self, c):
+        self.do_conditional(['c'], c)
+    def cmd_cvCpp(self, c):
+        self.do_conditional(['cpp'], c)
+    def cmd_cvPy(self, c):
+        self.do_conditional(['py'], c)
+    def cmd_cvCPy(self, c):
+        self.do_conditional(['c', 'py'], c)
+    def do_conditional(self, langs, c):
+        if self.language in langs:
+            self.doL(c.params[0].str, False)
+
+    def render(self, L):
+        """ return L rendered as a string """
+        save = self.f
+        self.f = StringIO.StringIO()
+        for x in L:
+            if isinstance(x, TexCmd):
+                self.docmd(x)
+            else:
+                self.doplain(x)
+        r = self.f.getvalue()
+        self.f = save
+        return r
+
+    def cmd_cvarg(self, c):
+        if len(c.params) != 2:
+            self.report_error(c, "Malformed cvarg")
+            return
+        e = self.ee()
+        if self.state == 'class':
+            nm = self.render(c.params[0].str)
+            if '->' in nm:
+                print >>self, "\n\n.. method:: %s\n\n" % nm
+            else:
+                print >>self, "\n\n.. attribute:: %s\n\n" % nm
+            self.indent += 1
+            print >>self
+            self.doL(c.params[1].str, False)
+            self.indent -= 1
+            print >>self
+            return
+        is_func_arg = (e == ['description']) and (not 'done' in self.function_props)
+        if is_func_arg:
+            nm = self.render(c.params[0].str)
+            print >>self, '\n:param %s: ' % nm,
+            type = None         # Try to figure out the argument type
+            # For now, multiple args get a pass
+            if (self.language == 'py') and ('signature' in self.function_props) and (not ',' in nm):
+                sig = self.function_props['signature']
+                argnames = [a[0] for a in sig[1]]
+                if isinstance(sig[2], str):
+                    resnames = [sig[2]]
+                else:
+                    resnames = list(sig[2])
+                if not nm in argnames + resnames:
+                    self.report_error(c, "Argument %s is not mentioned in signature (%s) (%s)" % (nm, ", ".join(argnames), ", ".join(resnames)))
+
+                api = python_api.get(self.function_props['name'], None)
+                if api:
+                    (ins, outs) = api
+                    adict = dict([(a.nm, a) for a in ins])
+                    arg = adict.get(nm, None)
+                    if arg:
+                        type = arg.ty
+                    else:
+                        self.report_error(c, 'cannot find arg %s in code' % nm)
+        elif len(e) > 0 and e[-1] == 'description':
+            print >>self, '\n* **%s** ' % self.render(c.params[0].str),
+        else:
+            self.report_error(c, "unexpected env (%s) for cvarg" % ",".join(e))
+        self.indent += 1
+        self.doL(c.params[1].str, False)
+        self.indent -= 1
+        print >>self
+        if is_func_arg and type:
+            type = type.replace('*', '')
+            translate = {
+                "ints" : "sequence of int",
+                "floats" : "sequence of int",
+                "IplImages" : "sequence of :class:`IplImage`",
+                "double" : "float",
+                "int" : "int",
+                "float" : "float",
+                "char" : "str",
+                "cvarrseq" : ":class:`CvArr` or :class:`CvSeq`",
+                "CvPoint2D32fs" : "sequence of (float, float)",
+                "pts_npts_contours" : "list of lists of (x,y) pairs",
+                "CvSeqOfCvSURFPoint" : ":class:`CvSeq` of :class:`CvSURFPoint`",
+                "CvSeqOfCvSURFDescriptor" : ":class:`CvSeq` of list of float",
+                "cvpoint2d32f_count" : "int",
+            }
+            print >>self, "\n:type %s: %s" % (nm, translate.get(type, ':class:`%s`' % type))
+
+    def cmd_genc(self, c): pass 
+    def cmd_genpy(self, c): pass 
+    def cmd_author(self, c): pass 
+    def cmd_date(self, c): pass
+    def cmd_def(self, c): pass
+    def cmd_documentclass(self, c): pass
+    def cmd_maketitle(self, c): pass
+    def cmd_newcommand(self, c): pass
+    def cmd_newline(self, c): pass
+    def cmd_setcounter(self, c): pass
+    def cmd_tableofcontents(self, c): pass
+    def cmd_targetlang(self, c): pass
+    def cmd_usepackage(self, c): pass
+    def cmd_title(self, c): pass
+    def cmd_par(self, c): pass
+    def cmd_hline(self, c):
+        print >>self, "\\hline"
+
+    def cmd_cite(self, c):
+        self.write("[%s]_" % str(c.params[0]))
+
+    def cmd_href(self, c):
+        if len(c.params) == 2:
+            self.write("`%s <%s>`_" % (str(c.params[1]), self.render(c.params[0].str)))
+        else:
+            self.report_error(c, "href should have two params")
+
+    def cmd_url(self, c):
+        self.write(str(c.params[0]))
+
+    def cmd_emph(self, c):
+        self.write("*" + self.render(c.params[0].str) + "*")
+
+    def cmd_textit(self, c):
+        self.write("*" + self.render(c.params[0].str) + "*")
+
+    def cmd_textbf(self, c):
+        self.write("**" + self.render(c.params[0].str) + "**")
+
+    def cmd_texttt(self, c):
+        self.write("``" + self.render(c.params[0].str) + "``")
+
+    def default_cmd(self, c):
+        if self.f == self.f_section:
+            self.write(repr(c))
+
+    def unrecognized_cmd(self, c):
+        # if writing the index or chapter heading, anything goes
+        if not self.f in [self.f_index, self.f_chapter]:
+            self.write(c.cmd)
+            if (not 'lstlisting' in self.ee()) and (not c.cmd in "#{}%&*\\_"):
+                if not c.cmd in self.unhandled_commands:
+                    self.report_error(c, 'unhandled command %s' % c.cmd)
+                    self.unhandled_commands.add(c.cmd)
+
+    def doL(self, L, newlines = True):
+        for x in L:
+            pos0 = self.f.tell()
+            if isinstance(x, TexCmd):
+                self.docmd(x)
+            else:
+                if 'lstlisting' in self.ee() or not newlines:
+                    self.doplain(x)
+                else:
+                    self.doplain(x.lstrip())
+            pos1 = self.f.tell()
+            if pos0 != pos1:
+                if self.state in ['math'] or not newlines:
+                    self.appendspace()
+                else:
+                    if not 'lstlisting' in self.ee():
+                        self.write('\n')
+
+    def handle_table(self, s):
+        oneline = s.replace('\n', ' ').strip()
+        rows = [r.strip() for r in oneline.split('\\hline')]
+        tab = []
+        for r in rows:
+            if r != "":
+                cols = [c.strip() for c in r.split('&')]
+                tab.append(cols)
+        widths = [max([len(r[i]) for r in tab]) for i in range(len(tab[0]))]
+
+        st = ""         # Sphinx table
+
+        if 0:
+            sep = "+" + "+".join(["-" * w for w in widths]) + "+"
+            st += sep + '\n'
+            for r in tab:
+                st += "|" + "|".join([c.center(w) for (c, w) in zip(r, widths)]) + "|" + '\n'
+                st += sep + '\n'
+
+        st = '.. table::\n\n'
+        sep = "  ".join(["=" * w for w in widths])
+        st += '    ' + sep + '\n'
+        for y,r in enumerate(tab):
+            st += '    ' + "  ".join([c.ljust(w) for (c, w) in zip(r, widths)]) + '\n'
+            if y == 0:
+                st += '    ' + sep + '\n'
+        st += '    ' + sep + '\n'
+        return st
+
+    def ee(self):
+        """ Return tags of the envstack.  envstack[0] is 'document', so skip it """
+        return [n for (n,_) in self.envstack[1:]]
+
+    def get_tags(self):
+        return self.tags
+
+    def close(self):
+
+        if self.envstack != []:
+            print >>self.errors, "Error envstack not empty at end of doc: " + repr(self.envstack)
+        print >>self.errors, "Unrecognized commands:"
+        for c in sorted(self.unhandled_commands):
+            print >>self.errors, "\n    " + c
+        print >>self.errors
+        if self.language == 'py':
+            print >>self.errors, "The following functions are undocumented"
+            for f in sorted(set(python_api) - self.covered):
+                print >>self.errors, '    ', f
+
+        print >>self.f_index, "    bibliography"
+        print >>self.f_index, """
+
+Indices and tables
+==================
+
+* :ref:`genindex`
+* :ref:`search`
+"""
+
+# Quick and dirty bibtex parser
+
+def parseBib(filename, language):
+    pp.ParserElement.setDefaultWhitespaceChars(" \n\t")
+    entry = returnList(pp.Word('@', pp.alphanums) + sl('{') +
+        pp.Word(pp.alphanums + "_") + sl(',') +
+        CommaList(returnTuple(pp.Word(pp.alphanums) + sl('=') + pp.QuotedString('{', endQuoteChar = '}'))) +
+        pp.Suppress(pp.Optional(',')) +
+        sl('}'))
+    r = (pp.ZeroOrMore(entry) | pp.Suppress('#' + pp.ZeroOrMore(pp.CharsNotIn('\n'))) + pp.StringEnd()).parseFile(filename)
+
+    bibliography = QOpen(os.path.join(language, "bibliography.rst"), 'wt')
+    print >>bibliography, "Bibliography"
+    print >>bibliography, "============"
+    print >>bibliography
+
+    for _,e in sorted([(str(x[1]), x) for x in r]):
+        (etype, tag, attrs) = str(e[0][1:]), str(e[1]), dict([(str(a), str(b)) for (a,b) in e[2]])
+        
+        representations = {
+            'article' :         '$author, "$title". $journal $volume $number, pp $pages ($year)',
+            'inproceedings' :   '$author "$title", $booktitle, $year',
+            'misc' :            '$author "$title", $year',
+            'techreport' :      '$author "$title", $edition, $edition ($year)',
+        }
+        if etype in representations:
+            if 0:
+                print >>bibliography, tag
+                print >>bibliography, "^" * len(tag)
+                print >>bibliography
+
+            print >>bibliography, ".. [%s] %s" % (tag, Template(representations[etype]).safe_substitute(attrs))
+            print >>bibliography
+    bibliography.close()
+
+if 1:
+    fulldoc = latexparser(sys.argv[1])
+
+    abspath = os.path.abspath(os.path.dirname(sys.argv[1]))
+
+    raw = open('raw.full', 'w')
+    for x in fulldoc:
+        print >>raw, repr(x)
+    raw.close()
+
+    # Filter on target language
+    def preprocess_conditionals(fd, conditionals):
+        r = []
+        ifstack = []
+        for x in fd:
+            if isinstance(x, TexCmd):
+                ll = x.cmd.rstrip()
+                loc = (x.filename, x.lineno)
+                if ll.startswith("if"):
+                    # print " " * len(ifstack), '{', loc
+                    ifstack.append((conditionals.get(ll[2:], False), loc))
+                elif ll.startswith("else"):
+                    ifstack[-1] = (not ifstack[-1][0], ifstack[-1][1])
+                elif ll.startswith("fi"):
+                    ifstack.pop()
+                    # print " " * len(ifstack), '}', loc
+                elif not False in [p for (p,_) in ifstack]:
+                    r.append(x)
+            else:
+                if not False in [p for (p,_) in ifstack]:
+                    r.append(x)
+        if ifstack != []:
+            print "unterminated if", ifstack
+            sys.exit(0)
+        return r
+
+    tags = {}
+    for language in sys.argv[2:]:
+        doc = preprocess_conditionals(fulldoc, {
+                                              'C' : language=='c',
+                                              'Python' : language=='py',
+                                              'Py' : language=='py',
+                                              'CPy' : (language=='py' or language == 'c'),
+                                              'Cpp' : language=='cpp',
+                                              'plastex' : True})
+
+        raw = open('raw.%s' % language, 'w')
+        for x in doc:
+            print >>raw, repr(x)
+        raw.close()
+        sr = SphinxWriter('index.rst', language, abspath)
+        print >>sr, """
+OpenCV |version| %s Reference
+=================================
+
+The OpenCV Wiki is here: http://opencv.willowgarage.com/
+
+Contents:
+
+.. toctree::
+    :maxdepth: 2
+
+""" % {'c': 'C', 'cpp': 'C++', 'py': 'Python'}[language]
+        sr.doL(doc)
+        sr.close()
+        parseBib('../opencv.bib', language)
+        tags.update(sr.get_tags())
+    open('TAGS', 'w').write("\n".join(sorted(tags.values())) + "\n")
+