docfixer.py

reduced python source for embedded apps

        # we may need to split off trailing white space:
        child = children[after - 1]
        data = child.data
        if string.rstrip(data) != data:
            have_last = 0
            child.splitText(len(string.rstrip(data)))
    para = doc.createElement(PARA_ELEMENT)
    prev = None
    indexes = range(start, after)
    indexes.reverse()
    for j in indexes:
        node = parent.childNodes[j]
        parent.removeChild(node)
        para.insertBefore(node, prev)
        prev = node
    if have_last:
        parent.appendChild(para)
        parent.appendChild(doc.createTextNode("\n\n"))
        return len(parent.childNodes)
    else:
        nextnode = parent.childNodes[start]
        if nextnode.nodeType == TEXT:
            if nextnode.data and nextnode.data[0] != "\n":
                nextnode.data = "\n" + nextnode.data
        else:
            newnode = doc.createTextNode("\n")
            parent.insertBefore(newnode, nextnode)
            nextnode = newnode
            start = start + 1
        parent.insertBefore(para, nextnode)
        return start + 1


def skip_leading_nodes(children, start=0):
    """Return index into children of a node at which paragraph building should
    begin or a recursive call to fixup_paras_helper() should be made (for
    subsections, etc.).

    When the return value >= len(children), we've built all the paras we can
    from this list of children.
    """
    i = len(children)
    while i > start:
        # skip over leading comments and whitespace:
        child = children[start]
        nodeType = child.nodeType
        if nodeType == TEXT:
            data = child.data
            shortened = string.lstrip(data)
            if shortened:
                if data != shortened:
                    # break into two nodes: whitespace and non-whitespace
                    child.splitText(len(data) - len(shortened))
                    return start + 1
                return start
            # all whitespace, just skip
        elif nodeType == ELEMENT:
            tagName = child.get_tagName()
            if tagName in RECURSE_INTO_PARA_CONTAINERS:
                return start
            if tagName not in PARA_LEVEL_ELEMENTS + PARA_LEVEL_PRECEEDERS:
                return start
        start = start + 1
    return start


def fixup_rfc_references(doc, fragment):
    for rfcnode in find_all_elements(fragment, "rfc"):
        rfcnode.appendChild(doc.createTextNode(
            "RFC " + rfcnode.getAttribute("num")))


def fixup_signatures(doc, fragment):
    for child in fragment.childNodes:
        if child.nodeType == ELEMENT:
            args = child.getElementsByTagName("args")
            for arg in args:
                fixup_args(doc, arg)
                arg.normalize()
            args = child.getElementsByTagName("constructor-args")
            for arg in args:
                fixup_args(doc, arg)
                arg.normalize()


def fixup_args(doc, arglist):
    for child in arglist.childNodes:
        if child.get_nodeName() == "optional":
            # found it; fix and return
            arglist.insertBefore(doc.createTextNode("["), child)
            optkids = child.childNodes
            while optkids:
                k = optkids[0]
                child.removeChild(k)
                arglist.insertBefore(k, child)
            arglist.insertBefore(doc.createTextNode("]"), child)
            arglist.removeChild(child)
            return fixup_args(doc, arglist)


def fixup_sectionauthors(doc, fragment):
    for sectauth in find_all_elements(fragment, "sectionauthor"):
        section = sectauth.parentNode
        section.removeChild(sectauth)
        sectauth._node.name = "author"
        sectauth.appendChild(doc.createTextNode(
            sectauth.getAttribute("name")))
        sectauth.removeAttribute("name")
        after = section.childNodes[2]
        title = section.childNodes[1]
        if title.get_nodeName() != "title":
            after = section.childNodes[0]
        section.insertBefore(doc.createTextNode("\n  "), after)
        section.insertBefore(sectauth, after)


def fixup_verbatims(doc):
    for verbatim in find_all_elements(doc, "verbatim"):
        child = verbatim.childNodes[0]
        if child.nodeType == TEXT \
           and string.lstrip(child.data)[:3] == ">>>":
            verbatim._node.name = "interactive-session"


def add_node_ids(fragment, counter=0):
    fragment._node.node_id = counter
    for node in fragment.childNodes:
        counter = counter + 1
        if node.nodeType == ELEMENT:
            counter = add_node_ids(node, counter)
        else:
            node._node.node_id = counter
    return counter + 1


REFMODINDEX_ELEMENTS = ('refmodindex', 'refbimodindex',
                        'refexmodindex', 'refstmodindex')

def fixup_refmodindexes(fragment):
    # Locate <ref*modindex>...</> co-located with <module>...</>, and
    # remove the <ref*modindex>, replacing it with index=index on the
    # <module> element.
    nodes = find_all_elements_from_set(fragment, REFMODINDEX_ELEMENTS)
    d = {}
    for node in nodes:
        parent = node.parentNode
        d[parent._node.node_id] = parent
    del nodes
    map(fixup_refmodindexes_chunk, d.values())


def fixup_refmodindexes_chunk(container):
    # node is probably a <para>; let's see how often it isn't:
    if container.get_tagName() != PARA_ELEMENT:
        bwrite("--- fixup_refmodindexes_chunk(%s)\n" % container)
    module_entries = find_all_elements(container, "module")
    if not module_entries:
        return
    index_entries = find_all_elements_from_set(container, REFMODINDEX_ELEMENTS)
    removes = []
    for entry in index_entries:
        children = entry.childNodes
        if len(children) != 0:
            bwrite("--- unexpected number of children for %s node:\n"
                   % entry.get_tagName())
            ewrite(entry.toxml() + "\n")
            continue
        found = 0
        module_name = entry.getAttribute("module")
        for node in module_entries:
            if len(node.childNodes) != 1:
                continue
            this_name = node.childNodes[0].data
            if this_name == module_name:
                found = 1
                node.setAttribute("index", "yes")
        if found:
            removes.append(entry)
    for node in removes:
        container.removeChild(node)


def fixup_bifuncindexes(fragment):
    nodes = find_all_elements(fragment, 'bifuncindex')
    d = {}
    # make sure that each parent is only processed once:
    for node in nodes:
        parent = node.parentNode
        d[parent._node.node_id] = parent
    del nodes
    map(fixup_bifuncindexes_chunk, d.values())


def fixup_bifuncindexes_chunk(container):
    removes = []
    entries = find_all_child_elements(container, "bifuncindex")
    function_entries = find_all_child_elements(container, "function")
    for entry in entries:
        function_name = entry.getAttribute("name")
        found = 0
        for func_entry in function_entries:
            t2 = func_entry.childNodes[0].data
            if t2[-2:] != "()":
                continue
            t2 = t2[:-2]
            if t2 == function_name:
                func_entry.setAttribute("index", "yes")
                func_entry.setAttribute("module", "__builtin__")
                if not found:
                    found = 1
                    removes.append(entry)
    for entry in removes:
        container.removeChild(entry)


def join_adjacent_elements(container, gi):
    queue = [container]
    while queue:
        parent = queue.pop()
        i = 0
        children = parent.get_childNodes()
        nchildren = len(children)
        while i < (nchildren - 1):
            child = children[i]
            if child.nodeName == gi:
                if children[i+1].nodeName == gi:
                    ewrite("--- merging two <%s/> elements\n" % gi)
                    child = children[i]
                    nextchild = children[i+1]
                    nextchildren = nextchild.get_childNodes()
                    while len(nextchildren):
                        node = nextchildren[0]
                        nextchild.removeChild(node)
                        child.appendChild(node)
                    parent.removeChild(nextchild)
                    continue
            if child.nodeType == ELEMENT:
                queue.append(child)
            i = i + 1


_token_rx = re.compile(r"[a-zA-Z][a-zA-Z0-9.-]*$")

def write_esis(doc, ofp, knownempty):
    for node in doc.childNodes:
        nodeType = node.nodeType
        if nodeType == ELEMENT:
            gi = node.get_tagName()
            if knownempty(gi):
                if node.hasChildNodes():
                    raise ValueError, \
                          "declared-empty node <%s> has children" % gi
                ofp.write("e\n")
            for k, v in node.attributes.items():
                value = v.value
                if _token_rx.match(value):
                    dtype = "TOKEN"
                else:
                    dtype = "CDATA"
                ofp.write("A%s %s %s\n" % (k, dtype, esistools.encode(value)))
            ofp.write("(%s\n" % gi)
            write_esis(node, ofp, knownempty)
            ofp.write(")%s\n" % gi)
        elif nodeType == TEXT:
            ofp.write("-%s\n" % esistools.encode(node.data))
        elif nodeType == ENTITY_REFERENCE:
            ofp.write("&%s\n" % node.get_nodeName())
        else:
            raise RuntimeError, "unsupported node type: %s" % nodeType


def convert(ifp, ofp):
    p = esistools.ExtendedEsisBuilder()
    p.feed(ifp.read())
    doc = p.document
    fragment = p.fragment
    normalize(fragment)
    simplify(doc, fragment)
    handle_labels(doc, fragment)
    handle_appendix(doc, fragment)
    fixup_trailing_whitespace(doc, {
        "abstract": "\n",
        "title": "",
        "chapter": "\n\n",
        "section": "\n\n",
        "subsection": "\n\n",
        "subsubsection": "\n\n",
        "paragraph": "\n\n",
        "subparagraph": "\n\n",
        })
    cleanup_root_text(doc)
    cleanup_trailing_parens(fragment, ["function", "method", "cfunction"])
    cleanup_synopses(doc, fragment)
    fixup_descriptors(doc, fragment)
    fixup_verbatims(fragment)
    normalize(fragment)
    fixup_paras(doc, fragment)
    fixup_sectionauthors(doc, fragment)
    fixup_table_structures(doc, fragment)
    fixup_rfc_references(doc, fragment)
    fixup_signatures(doc, fragment)
    add_node_ids(fragment)
    fixup_refmodindexes(fragment)
    fixup_bifuncindexes(fragment)
    # Take care of ugly hacks in the LaTeX markup to avoid LaTeX and
    # LaTeX2HTML screwing with GNU-style long options (the '--' problem).
    join_adjacent_elements(fragment, "option")
    #
    d = {}
    for gi in p.get_empties():
        d[gi] = gi
    if d.has_key("rfc"):
        del d["rfc"]
    knownempty = d.has_key
    #
    try:
        write_esis(fragment, ofp, knownempty)
    except IOError, (err, msg):
        # Ignore EPIPE; it just means that whoever we're writing to stopped
        # reading.  The rest of the output would be ignored.  All other errors
        # should still be reported,
        if err != errno.EPIPE:
            raise


def main():
    if len(sys.argv) == 1:
        ifp = sys.stdin
        ofp = sys.stdout
    elif len(sys.argv) == 2:
        ifp = open(sys.argv[1])
        ofp = sys.stdout
    elif len(sys.argv) == 3:
        ifp = open(sys.argv[1])
        ofp = open(sys.argv[2], "w")
    else:
        usage()
        sys.exit(2)
    convert(ifp, ofp)


if __name__ == "__main__":
    main()