prepare.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-

"""Format Identification for Digital Objects."""

from __future__ import print_function

from argparse import ArgumentParser
import hashlib
import sys
from xml.dom import minidom
from xml.etree import ElementTree as ET
import zipfile

from six.moves import cStringIO
from six.moves.urllib.request import urlopen
from six.moves.urllib.parse import urlparse
from six.moves.urllib.error import HTTPError

from .versions import get_local_versions
from .char_handler import escape

FLG_INCOMPATIBLE = '__INCOMPATIBLE_SIG__'


class NS:
    """
    Helper class for XML name spaces in ElementTree.

    Use like MYNS=NS("{http://some/uri}") and then MYNS(tag1/tag2).
    """

    def __init__(self, uri):
        """Instantiate class with `uri` argument."""
        self.uri = uri

    def __getattr__(self, tag):
        """Append URI to the class attributes."""
        return self.uri + tag

    def __call__(self, path):
        """Define behavior when the instant is used as a function."""
        return "/".join(getattr(self, tag) for tag in path.split("/"))


XHTML = NS("{http://www.w3.org/1999/xhtml}")  # XHTML namespace
TNA = NS("{http://pronom.nationalarchives.gov.uk}")  # TNA namespace


def get_text_tna(element, tag, default=''):
    """Helper function to return the text for a tag or path using the TNA namespace."""
    part = element.find(TNA(tag))
    if part is None or part.text is None:
        return default
    return part.text.strip()


def prettify(elem):
    """Return a pretty-printed XML string for the Element."""
    rough_string = ET.tostring(elem, 'UTF-8')
    reparsed = minidom.parseString(rough_string)
    return reparsed.toprettyxml(indent="  ")


class FormatInfo:
    """Convert PRONOM formats into FIDO signatures."""

    def __init__(self, pronom_files, format_list=None):
        """Instantiate class, take a list of PRONOM files and an optional list of formats."""
        format_list = format_list if format_list else []
        self.info = {}
        self.formats = []
        self.pronom_files = pronom_files
        for f in format_list:
            self.add_format(f)  # FIXME: add_format is undefined!

    def save(self, dst=sys.stdout):
        """Write the fido XML format definitions to @param dst."""
        tree = ET.ElementTree(ET.Element('formats', {
            'version': '0.3',
            'xmlns:xsi': "http://www.w3.org/2001/XMLSchema-instance",
            'xsi:noNamespaceSchemaLocation': "fido-formats.xsd",
            'xmlns:dc': "http://purl.org/dc/elements/1.1/",
            'xmlns:dcterms': "http://purl.org/dc/terms/"
        }))
        root = tree.getroot()
        for f in self.formats:
            # MdR: this skipped puids without sig, but we want them ALL
            # because puid might be matched on extension
            # if f.find('signature'):
            root.append(f)
        self.indent(root)
        with open(dst, 'wb') as file_:
            file_.write(ET.tostring(root))

    def indent(self, elem, level=0):
        """Indent output."""
        if len(elem):
            self._indent_ele(elem, level)
        else:
            if level and (not elem.tail or not elem.tail.strip()):
                elem.tail = self._indent_text(level)

    def _indent_ele(self, elem, level):
        """Indent the element."""
        if not elem.text or not elem.text.strip():
            elem.text = self._indent_text(level) + "  "
        if not elem.tail or not elem.tail.strip():
            elem.tail = self._indent_text(level)
        for elem in elem:
            self.indent(elem, level + 1)
        if not elem.tail or not elem.tail.strip():
            elem.tail = self._indent_text(level)

    def _indent_text(self, level):
        return "\n" + level * "  "

    def load_pronom_xml(self, puid_filter=None):
        """
        Load the pronom XML from self.pronom_files and convert it to fido XML.

        As a side-effect, set self.formats to a list of ElementTree.Element.
        If a @param puid is specified, only that one will be loaded.
        """
        formats = []
        try:
            zip = zipfile.ZipFile(self.pronom_files, 'r')
            for item in zip.infolist():
                try:
                    stream = zip.open(item)
                    # Work is done here!
                    format_ = self.parse_pronom_xml(stream, puid_filter)
                    if format_ is not None:
                        formats.append(format_)
                finally:
                    stream.close()
        finally:
            try:
                zip.close()
            except Exception as e:
                print("An error occured loading '{0}' (exception: {1})".format(self.pronom_files, e), file=sys.stderr)
                sys.exit()
        # Replace the formatID with puids in has_priority_over
        if puid_filter is None:
            id_map = {}
            for element in formats:
                puid = element.find('puid').text
                # print('working on puid:{}'.format(puid))
                pronom_id = element.find('pronom_id').text
                id_map[pronom_id] = puid
            for element in formats:
                for rel in element.findall('has_priority_over'):
                    try:
                        rel.text = id_map[rel.text]
                    except KeyError:
                        print("Error looking up priority over PRONOM ID {0} for format {1}".format(rel.text, element.find('puid').text), file=sys.stderr)

        self._sort_formats(formats)
        self.formats = formats

    def parse_pronom_xml(self, source, puid_filter=None):
        """
        Parse PRONOM XML and convert into FIDO XML.

        If a @param puid is specified, only that one will be loaded.
        @return ET.ElementTree Element representing it.
        """
        pronom_xml = ET.parse(source)
        pronom_root = pronom_xml.getroot()
        pronom_format = pronom_root.find(TNA('report_format_detail/FileFormat'))
        fido_format = ET.Element('format')
        # Get the base Format information
        for id in pronom_format.findall(TNA('FileFormatIdentifier')):
            type = get_text_tna(id, 'IdentifierType')
            if type == 'PUID':
                puid = get_text_tna(id, 'Identifier')
                ET.SubElement(fido_format, 'puid').text = puid
                if puid_filter and puid != puid_filter:
                    return None
        # A bit clumsy.  I want to have puid first, then mime, then container.
        for id in pronom_format.findall(TNA('FileFormatIdentifier')):
            type = get_text_tna(id, 'IdentifierType')
            if type == 'MIME':
                ET.SubElement(fido_format, 'mime').text = get_text_tna(id, 'Identifier')
            elif type == 'PUID':
                puid = get_text_tna(id, 'Identifier')
                if puid == 'x-fmt/263':
                    ET.SubElement(fido_format, 'container').text = 'zip'
                elif puid == 'x-fmt/265':
                    ET.SubElement(fido_format, 'container').text = 'tar'
        ET.SubElement(fido_format, 'name').text = get_text_tna(pronom_format, 'FormatName')
        ET.SubElement(fido_format, 'version').text = get_text_tna(pronom_format, 'FormatVersion')
        ET.SubElement(fido_format, 'alias').text = get_text_tna(pronom_format, 'FormatAliases')
        ET.SubElement(fido_format, 'pronom_id').text = get_text_tna(pronom_format, 'FormatID')
        # Get the extensions from the ExternalSignature
        for x in pronom_format.findall(TNA('ExternalSignature')):
            ET.SubElement(fido_format, 'extension').text = get_text_tna(x, 'Signature')
        for id in pronom_format.findall(TNA('FileFormatIdentifier')):
            type = get_text_tna(id, 'IdentifierType')
            if type == 'Apple Uniform Type Identifier':
                ET.SubElement(fido_format, 'apple_uti').text = get_text_tna(id, 'Identifier')
        # Handle the relationships
        for x in pronom_format.findall(TNA('RelatedFormat')):
            rel = get_text_tna(x, 'RelationshipType')
            if rel == 'Has priority over':
                ET.SubElement(fido_format, 'has_priority_over').text = get_text_tna(x, 'RelatedFormatID')
        # Get the InternalSignature information
        for pronom_sig in pronom_format.findall(TNA('InternalSignature')):
            fido_sig = ET.SubElement(fido_format, 'signature')
            ET.SubElement(fido_sig, 'name').text = get_text_tna(pronom_sig, 'SignatureName')
            # There are some funny chars in the notes, which caused me trouble and it is a unicode string,
            ET.SubElement(fido_sig, 'note').text = get_text_tna(pronom_sig, 'SignatureNote')
            for pronom_pat in pronom_sig.findall(TNA('ByteSequence')):
                # print('Parsing ID:{}'.format(puid))
                fido_pat = ET.SubElement(fido_sig, 'pattern')
                pos = fido_position(get_text_tna(pronom_pat, 'PositionType'))
                byte_seq = get_text_tna(pronom_pat, 'ByteSequenceValue')
                offset = get_text_tna(pronom_pat, 'Offset')
                max_offset = get_text_tna(pronom_pat, 'MaxOffset')
                if not max_offset:
                    pass
                # print "working on puid:", puid, ", position: ", pos, "with offset, maxoffset: ", offset, ",", max_offset
                try:
                    regex = convert_to_regex(byte_seq, 'Little', pos, offset, max_offset)
                except ValueError as ve:
                    print('ValueError converting PUID {} signature to regex: {}'.format(puid, ve), file=sys.stderr)
                    regex = FLG_INCOMPATIBLE

                # print "done puid", puid
                if regex == FLG_INCOMPATIBLE:
                    print("Error: incompatible PRONOM signature found for puid {} skipping...".format(puid), file=sys.stderr)
                    # remove the empty 'signature' nodes
                    # now that the signature is not compatible and thus "regex" is empty
                    remove = fido_format.findall('signature')
                    for r in remove:
                        fido_format.remove(r)
                    continue
                ET.SubElement(fido_pat, 'position').text = pos
                ET.SubElement(fido_pat, 'pronom_pattern').text = byte_seq
                ET.SubElement(fido_pat, 'regex').text = regex
        # Get the format details
        fido_details = ET.SubElement(fido_format, 'details')
        ET.SubElement(fido_details, 'dc:description').text = get_text_tna(pronom_format, 'FormatDescription')
        ET.SubElement(fido_details, 'dcterms:available').text = get_text_tna(pronom_format, 'ReleaseDate')
        ET.SubElement(fido_details, 'dc:creator').text = get_text_tna(pronom_format, 'Developers/DeveloperCompoundName')
        ET.SubElement(fido_details, 'dcterms:publisher').text = get_text_tna(pronom_format, 'Developers/OrganisationName')
        for x in pronom_format.findall(TNA('RelatedFormat')):
            rel = get_text_tna(x, 'RelationshipType')
            if rel == 'Is supertype of':
                ET.SubElement(fido_details, 'is_supertype_of').text = get_text_tna(x, 'RelatedFormatID')
        for x in pronom_format.findall(TNA('RelatedFormat')):
            rel = get_text_tna(x, 'RelationshipType')
            if rel == 'Is subtype of':
                ET.SubElement(fido_details, 'is_subtype_of').text = get_text_tna(x, 'RelatedFormatID')
        ET.SubElement(fido_details, 'content_type').text = get_text_tna(pronom_format, 'FormatTypes')
        # References
        for x in pronom_format.findall(TNA("Document")):
            r = ET.SubElement(fido_details, 'reference')
            ET.SubElement(r, 'dc:title').text = get_text_tna(x, 'TitleText')
            ET.SubElement(r, 'dc:creator').text = get_text_tna(x, 'Author/AuthorCompoundName')
            ET.SubElement(r, 'dc:publisher').text = get_text_tna(x, 'Publisher/PublisherCompoundName')
            ET.SubElement(r, 'dcterms:available').text = get_text_tna(x, 'PublicationDate')
            for id in x.findall(TNA('DocumentIdentifier')):
                type = get_text_tna(id, 'IdentifierType')
                if type == 'URL':
                    ET.SubElement(r, 'dc:identifier').text = "http://" + get_text_tna(id, 'Identifier')
                else:
                    ET.SubElement(r, 'dc:identifier').text = get_text_tna(id, 'IdentifierType') + ":" + get_text_tna(id, 'Identifier')
            ET.SubElement(r, 'dc:description').text = get_text_tna(x, 'DocumentNote')
            ET.SubElement(r, 'dc:type').text = get_text_tna(x, 'DocumentType')
            ET.SubElement(r, 'dcterms:license').text = get_text_tna(x, 'AvailabilityDescription') + " " + get_text_tna(x, 'AvailabilityNote')
            ET.SubElement(r, 'dc:rights').text = get_text_tna(x, 'DocumentIPR')
        # Examples
        for x in pronom_format.findall(TNA("ReferenceFile")):
            rf = ET.SubElement(fido_details, 'example_file')
            ET.SubElement(rf, 'dc:title').text = get_text_tna(x, 'ReferenceFileName')
            ET.SubElement(rf, 'dc:description').text = get_text_tna(x, 'ReferenceFileDescription')
            checksum = ""
            for id in x.findall(TNA('ReferenceFileIdentifier')):
                type = get_text_tna(id, 'IdentifierType')
                if type == 'URL':
                    # Starting with PRONOM 89, some URLs contain http://
                    # and others do not.
                    url = get_text_tna(id, 'Identifier')
                    if not urlparse(url).scheme:
                        url = "http://" + url
                    ET.SubElement(rf, 'dc:identifier').text = url
                    # And calculate the checksum of this resource:
                    m = hashlib.md5()
                    try:
                        sock = urlopen(url)
                        m.update(sock.read())
                        sock.close()
                    except HTTPError as http_excep:
                        sys.stderr.write('HTTP {} error loading resource {}\n'.format(http_excep.code, url))
                        if http_excep.code == 404:
                            continue

                    checksum = m.hexdigest()
                else:
                    ET.SubElement(rf, 'dc:identifier').text = get_text_tna(id, 'IdentifierType') + ":" + get_text_tna(id, 'Identifier')
            ET.SubElement(rf, 'dcterms:license').text = ""
            ET.SubElement(rf, 'dc:rights').text = get_text_tna(x, 'ReferenceFileIPR')
            checksumElement = ET.SubElement(rf, 'checksum')
            checksumElement.text = checksum
            checksumElement.attrib['type'] = "md5"
        # Record Metadata
        md = ET.SubElement(fido_details, 'record_metadata')
        ET.SubElement(md, 'status').text = 'unknown'
        ET.SubElement(md, 'dc:creator').text = get_text_tna(pronom_format, 'ProvenanceName')
        ET.SubElement(md, 'dcterms:created').text = get_text_tna(pronom_format, 'ProvenanceSourceDate')
        ET.SubElement(md, 'dcterms:modified').text = get_text_tna(pronom_format, 'LastUpdatedDate')
        ET.SubElement(md, 'dc:description').text = get_text_tna(pronom_format, 'ProvenanceDescription')
        return fido_format

    # FIXME: I don't think that this quite works yet!
    def _sort_formats(self, formatlist):
        """Sort the format list based on their priority relationships so higher priority formats appear earlier in the list."""
        def compare_formats(f1, f2):
            f1ID = f1.find('puid').text
            f2ID = f2.find('puid').text
            for worse in f1.findall('has_priority_over'):
                if worse.text == f2ID:
                    return - 1
            for worse in f2.findall('has_priority_over'):
                if worse.text == f1ID:
                    return 1
            if f1ID < f2ID:
                return - 1
            if f1ID == f2ID:
                return 0
            return 1
        return sorted(formatlist, key=_cmp_to_key(compare_formats))


def _cmp_to_key(mycmp):
    """Convert a cmp= function into a key= function."""
    # From https://docs.python.org/3/howto/sorting.html#sortinghowto
    class K:
        """Wrapper class for comparator function."""

        def __init__(self, obj, *_):
            self.obj = obj

        def __lt__(self, other):
            return mycmp(self.obj, other.obj) < 0

        def __gt__(self, other):
            return mycmp(self.obj, other.obj) > 0

        def __eq__(self, other):
            return mycmp(self.obj, other.obj) == 0

        def __le__(self, other):
            return mycmp(self.obj, other.obj) <= 0

        def __ge__(self, other):
            return mycmp(self.obj, other.obj) >= 0

        def __ne__(self, other):
            return mycmp(self.obj, other.obj) != 0

    return K


def fido_position(pronom_position):
    """Return BOF/EOF/VAR instead of the more verbose pronom position names."""
    if pronom_position == 'Absolute from BOF':
        return 'BOF'
    if pronom_position == 'Absolute from EOF':
        return 'EOF'
    if pronom_position == 'Variable':
        return 'VAR'
    if pronom_position == 'Indirect From BOF':
        return 'IFB'
    # to make sure FIDO does not crash (IFB aftermath)
    sys.stderr.write("Unknown pronom PositionType:" + pronom_position)
    return 'VAR'


def _convert_err_msg(msg, c, i, chars, buf):
    return "Conversion: {0}: char='{1}', at pos {2} in \n  {3}\n  {4}^\nBuffer = {5}".format(msg, c, i, chars, i * ' ', buf.getvalue())


def do_byte(chars, i, littleendian, esc=True):
    """
    Convert two chars[i] and chars[i+1] into a byte.

    @return a tuple (byte, 2)
    """
    c1 = '0123456789ABCDEF'.find(chars[i].upper())
    c2 = '0123456789ABCDEF'.find(chars[i + 1].upper())
    buf = cStringIO()
    if (c1 < 0 or c2 < 0):
        raise Exception(_convert_err_msg('bad byte sequence', chars[i:i + 2], i, chars, buf))
    if littleendian:
        val = chr(16 * c1 + c2)
    else:
        val = chr(c1 + 16 * c2)
    if esc:
        return (escape(val), 2)
    return (val, 2)


# Python now allows regex repetitions to be max out somewhere between 4 and 4.3
# billion, cf. https://bugs.python.org/issue13169#msg180499
MAX_REGEX_REPS = 4e9


def calculate_repetition(char, pos, offset, maxoffset):
    """Recursively calculates offset/maxoffset repetition.

    This function only has an effect when one or both offsets is greater than
    MAX_REGEX_REPS bytes (4GB). See: https://bugs.python.org/issue13169.
    """
    calcbuf = cStringIO()

    calcremain = False
    offsetremain = 0
    maxoffsetremain = 0

    if offset is not None and int(offset) > MAX_REGEX_REPS:
        offsetremain = str(int(offset) - MAX_REGEX_REPS)
        offset = str(int(MAX_REGEX_REPS))
        calcremain = True
    if maxoffset is not None and int(maxoffset) > MAX_REGEX_REPS:
        maxoffsetremain = str(int(maxoffset) - MAX_REGEX_REPS)
        maxoffset = str(int(MAX_REGEX_REPS))
        calcremain = True

    if pos == "BOF" or pos == "EOF":
        if offset != '0':
            calcbuf.write(char + '{' + str(offset))
            if maxoffset is not None:
                calcbuf.write(',' + maxoffset)
            calcbuf.write('}')
        elif maxoffset is not None:
            calcbuf.write(char + '{0,' + maxoffset + '}')

    if pos == "IFB":
        if offset != '0':
            calcbuf.write(char + '{' + str(offset))
            if maxoffset is not None:
                calcbuf.write(',' + maxoffset)
            calcbuf.write('}')
            if maxoffset is not None:
                calcbuf.write(',}')
        elif maxoffset is not None:
            calcbuf.write(char + '{0,' + maxoffset + '}')

    if calcremain:  # recursion happens here
        calcbuf.write(calculate_repetition(char, pos, offsetremain, maxoffsetremain))

    val = calcbuf.getvalue()
    calcbuf.close()
    return val


def do_all_bitmasks(chars, i, littleendian):
    """(byte & bitmask) == bitmask."""
    return do_any_all_bitmasks(
        chars, i, lambda byt, bitmask: ((byt & bitmask) == bitmask),
        littleendian)


def do_any_bitmasks(chars, i, littleendian):
    """(byte & bitmask) != 0."""
    return do_any_all_bitmasks(
        chars, i, lambda byt, bitmask: ((byt & bitmask) != 0),
        littleendian)


def do_any_all_bitmasks(chars, i, predicate, littleendian):
    """Convert an all/any bitmask string (e.g., &07) to a Python regex.

    Whether the bitmask is "all" (&) or "any" (~) is determined by the supplied
    function ``predicate``. As an example, '&07' means 'match bytes with all first
    three bits set' or 'match all bytes where (byte & 0x07) == 0x07' or, in
    Python, match this disjunctive regex::

        >>> '0x7|0xf|0x17|0x1f|0x27|0x2f|0x37|0x3f|0x47|0x4f|0x57|0x5f|0x67|0x6f|0x77|0x7f|0x87|0x8f|0x97|0x9f|0xa7|0xaf|0xb7|0xbf|0xc7|0xcf|0xd7|0xdf|0xe7|0xef|0xf7|0xff'

    See https://github.com/nishihatapalmer/byteseek/wiki/Regular-Expression-Syntax#all-bitmasks
    and https://github.com/nishihatapalmer/byteseek/wiki/Regular-Expression-Syntax#any-bitmasks
    """
    byt, inc = do_byte(chars, i + 1, littleendian, esc=False)
    bitmask = ord(byt)
    regex = '({})'.format(
        '|'.join(['\\x' + hex(byte)[2:].zfill(2) for byte in range(0x100)
                  if predicate(byte, bitmask)]))
    return regex, inc + 1


def convert_to_regex(chars, endianness='', pos='BOF', offset='0', maxoffset=''):
    """
    Convert to regular expression.

    Endianness is not used.

    @param chars, a pronom bytesequence, into a
    @return regular expression.
    """
    if 'Big' in endianness:
        littleendian = False
    else:
        littleendian = True
    if len(offset) == 0:
        offset = '0'
    if len(maxoffset) == 0:
        maxoffset = None
    if maxoffset == '0':
        maxoffset = None
    buf = cStringIO()
    buf.write("(?s)")  # If a regex starts with (?s), it is equivalent to DOTALL.
    i = 0
    state = 'start'
    if 'BOF' in pos:
        buf.write('\\A')  # start of regex
        buf.write(calculate_repetition('.', pos, offset, maxoffset))

    if 'IFB' in pos:
        buf.write('\\A')
        buf.write(calculate_repetition('.', pos, offset, maxoffset))

    while True:
        if i == len(chars):
            break
        # print _convert_err_msg(state,chars[i],i,chars)
        if state == 'start':
            if chars[i].isalnum():
                state = 'bytes'
            elif chars[i] == '&':
                state = 'all-bitmask'
            elif chars[i] == '~':
                state = 'any-bitmask'
            elif chars[i] == '[' and chars[i + 1] == '!':
                state = 'non-match'
            elif chars[i] == '[':
                state = 'bracket'
            elif chars[i] == '{':
                state = 'curly'
            elif chars[i] == '(':
                state = 'paren'
            elif chars[i] in '*+?':
                state = 'specials'
            else:
                raise ValueError(_convert_err_msg('Illegal character in start', chars[i], i, chars, buf))
        elif state == 'bytes':
            (byt, inc) = do_byte(chars, i, littleendian)
            buf.write(byt)
            i += inc
            state = 'start'
        elif state == 'all-bitmask':
            (byt, inc) = do_all_bitmasks(chars, i, littleendian)
            buf.write(byt)
            i += inc
            state = 'start'
        elif state == 'any-bitmask':
            (byt, inc) = do_any_bitmasks(chars, i, littleendian)
            buf.write(byt)
            i += inc
            state = 'start'
        elif state == 'non-match':
            buf.write('(?!')
            i += 2
            while True:
                if chars[i].isalnum():
                    (byt, inc) = do_byte(chars, i, littleendian)
                    buf.write(byt)
                    i += inc
                elif chars[i] == '&':
                    (byt, inc) = do_all_bitmasks(chars, i, littleendian)
                    buf.write(byt)
                    i += inc
                elif chars[i] == '~':
                    (byt, inc) = do_any_bitmasks(chars, i, littleendian)
                    buf.write(byt)
                    i += inc
                elif chars[i] == ']':
                    break
                else:
                    raise Exception(_convert_err_msg('Illegal character in non-match', chars[i], i, chars, buf))
            buf.write(')')
            i += 1
            state = 'start'

        elif state == 'bracket':
            try:
                buf.write('[')
                i += 1
                (byt, inc) = do_byte(chars, i, littleendian)
                buf.write(byt)
                i += inc
                # assert(chars[i] == ':')
                if chars[i] != ':':
                    return "__INCOMPATIBLE_SIG__"
                buf.write('-')
                i += 1
                (byt, inc) = do_byte(chars, i, littleendian)
                buf.write(byt)
                i += inc
                # assert(chars[i] == ']')
                if chars[i] != ']':
                    return "__INCOMPATIBLE_SIG__"
                buf.write(']')
                i += 1
            except Exception:
                print(_convert_err_msg('Illegal character in bracket', chars[i], i, chars, buf))
                raise
            if i < len(chars) and chars[i] == '{':
                state = 'curly-after-bracket'
            else:
                state = 'start'
        elif state == 'paren':
            buf.write('(?:')
            i += 1
            while True:
                if chars[i].isalnum():
                    (byt, inc) = do_byte(chars, i, littleendian)
                    buf.write(byt)
                    i += inc
                elif chars[i] == '|':
                    buf.write('|')
                    i += 1
                elif chars[i] == ')':
                    break
                # START fix FIDO-20
                elif chars[i] == '[':
                    buf.write('[')
                    i += 1
                    (byt, inc) = do_byte(chars, i, littleendian)
                    buf.write(byt)
                    i += inc
                    # assert(chars[i] == ':')
                    if chars[i] != ':':
                        return "__INCOMPATIBLE_SIG__"
                    buf.write('-')
                    i += 1
                    (byt, inc) = do_byte(chars, i, littleendian)
                    buf.write(byt)
                    i += inc

                    # assert(chars[i] == ']')
                    if chars[i] != ']':
                        return "__INCOMPATIBLE_SIG__"
                    buf.write(']')
                    i += 1
                else:
                    raise Exception(_convert_err_msg(('Current state = \'{0}\' : Illegal character in paren').format(state), chars[i], i, chars, buf))
            buf.write(')')
            i += 1
            state = 'start'
            # END fix FIDO-20
        elif state in ['curly', 'curly-after-bracket']:
            # {nnnn} or {nnn-nnn} or {nnn-*}
            # {nnn} or {nnn,nnn} or {nnn,}
            # when there is a curly-after-bracket, then the {m,n} applies to the bracketed item
            # The above, while sensible, appears to be incorrect.  A '.' is always needed.
            # for droid equiv behavior
            # if state == 'curly':
            buf.write('.')
            buf.write('{')
            i += 1                # skip the (
            while True:
                if chars[i].isalnum():
                    buf.write(chars[i])
                    i += 1
                elif chars[i] == '-':
                    buf.write(',')
                    i += 1
                elif chars[i] == '*':  # skip the *
                    i += 1
                elif chars[i] == '}':
                    break
                else:
                    raise Exception(_convert_err_msg('Illegal character in curly', chars[i], i, chars, buf))
            buf.write('}')
            i += 1                # skip the )
            state = 'start'
        elif state == 'specials':
            if chars[i] == '*':
                buf.write('.*')
                i += 1
            elif chars[i] == '+':
                buf.write('.+')
                i += 1
            elif chars[i] == '?':
                if chars[i + 1] != '?':
                    raise Exception(_convert_err_msg('Illegal character after ?', chars[i + 1], i + 1, chars, buf))
                buf.write('.?')
                i += 2
            state = 'start'
        else:
            raise Exception('Illegal state {0}'.format(state))

    if 'EOF' in pos:
        buf.write(calculate_repetition('.', pos, offset, maxoffset))
        buf.write('\\Z')

    val = buf.getvalue()
    buf.close()
    return val


def run(input=None, output=None, puid=None):
    """Convert PRONOM formats into FIDO signatures."""
    versions = get_local_versions()

    if input is None:
        input = versions.get_zip_file()
    if output is None:
        output = versions.get_signature_file()

    info = FormatInfo(input)
    info.load_pronom_xml(puid)
    info.save(output)
    print('Converted {0} PRONOM formats to FIDO signatures'.format(len(info.formats)), file=sys.stderr)


def main(args=None):
    """Main CLI entrypoint."""
    if args is None:
        args = sys.argv[1:]

    parser = ArgumentParser(description='Produce the FIDO format XML that is loaded at run-time')
    parser.add_argument('-input', default=None, help='Input file, a Zip containing PRONOM XML files')
    parser.add_argument('-output', default=None, help='Output file')
    parser.add_argument('-puid', default=None, help='A particular PUID record to extract')
    args = parser.parse_args(args)

    run(input=args.input, output=args.output, puid=args.puid)


if __name__ == '__main__':
    main()