class Format(object):
    __slots__ = 'impl', 'mod'

    def __init__(self, impl):
        self.impl = impl
        self.mod = impl.mod

    def get_formatted_row(self, row):
        fr = self.get_stat_data(row)
        rows = []
        rs = row.name.split('\n')
        subsequent_indent = len(fr)*' '
        rows.extend(self.mod.wrap(
            fr+rs[0],
            width=self.mod.line_length,
            subsequent_indent=subsequent_indent))

        for r in rs[1:]:
            rows.extend(self.mod.wrap(
                r,
                width=self.mod.line_length,
                initial_indent=subsequent_indent,
                subsequent_indent=subsequent_indent))
        return '\n'.join(rows)

    def load_statrow_csk(self, r):
        impl = self.impl
        count, size, kind = r.split(' ', 2)
        count = int(count)
        size = int(size)
        impl.cum_size += size
        return StatRow(count, size, kind, impl.cur_index, impl.cum_size)

    def load_statrow_sk(self, r):
        impl = self.impl
        size, kind = r.split(' ', 1)
        size = int(size)
        impl.cum_size += size
        return StatRow(1, size, kind, impl.cur_index, impl.cum_size)

    def _oh_get_num_lines(self):
        return self.impl.numrows

    def _oh_get_label(self):
        return self.get_label()

    def _oh_get_row_header(self):
        impl = self.impl
        if not (impl.count or impl.size):
            return ''
        sh = self.get_stat_header()
        return self.mod.fill(
            sh + self.impl.kindheader,
            width=self.mod.line_length,
            subsequent_indent=' '*len(sh))

    def _oh_get_more_msg(self, start_lineno, end_lineno):
        nummore = self.impl.numrows-(end_lineno+1)
        return "<%d more rows. Type e.g. '_.more' to view.>" % nummore

    def _oh_get_line_iter(self):
        for numrows, row in enumerate(self.impl.get_rows()):
            yield self.get_formatted_row(row)


class SetFormat(Format):
    __slots__ = ()

    def get_label(self):
        impl = self.impl
        if impl.count != 1:
            s = 's'
        else:
            s = ''
        return 'Partition of a set of %d object%s. Total size = %d bytes.' % (
            impl.count, s, impl.size)

    def get_rowdata(self, row):
        return '%d %d %s' % (row.count, row.size, row.name)

    def get_stat_header(self):
        return (
            ' Index  Count   %     Size   % Cumulative  % ')

    def get_stat_data(self, row):
        format = '%6d %6d %3d %8d %3d %9d %3d '
        impl = self.impl
        fr = format % (
            row.index,
            row.count, int('%.0f' % (row.count * 100.0/impl.count)),
            row.size, int('%.0f' % (row.size * 100.0/impl.size)),
            row.cumulsize, int('%.0f' % (row.cumulsize * 100.0/impl.size)),
        )
        return fr

    def load_statrow(self, r):
        return self.load_statrow_csk(r)


class IdFormat(Format):
    __slots__ = ()

    def get_label(self):
        impl = self.impl
        if impl.count != 1:
            s = 's'
        else:
            s = ''
        return (
            'Set of %d %s object%s. Total size = %d bytes.' % (
                impl.count, impl.kindname, s, impl.size))
        return part

    def get_rowdata(self, row):
        return '%d %s' % (row.size, row.name)

    def get_stat_header(self):
        return (
            ' Index     Size   %   Cumulative  %   ')

    def get_stat_data(self, row):
        impl = self.impl
        format = '%6d %8d %5.1f %9d %5.1f '
        fr = format % (
            row.index,
            row.size, (row.size * 100.0/impl.size),
            row.cumulsize, row.cumulsize * 100.0/impl.size,
        )
        return fr

    def load_statrow(self, r):
        return self.load_statrow_sk(r)


class DiffFormat(Format):
    __slots__ = ()

    def _percent_of_b(self, size):
        if self.impl.b_size != 0:
            return '%9.3g' % (size*100.0/self.impl.b_size,)
        else:
            return '   (n.a.)'

    def get_label(self):
        impl = self.impl
        x = (
            'Summary of difference operation (A-B).\n' +
            '        Count     Size\n' +
            '  A    %6d %8d\n' % (impl.count+impl.b_count, impl.size+impl.b_size) +
            '  B    %6d %8d\n' % (impl.b_count, impl.b_size) +
            '  A-B  %6d %8d  = %s %% of B\n' % (impl.count, impl.size, self._percent_of_b(impl.size)))

        if impl.count or impl.size:
            x += '\nDifferences by kind, largest absolute size diffs first.'
        return x

    def get_rowdata(self, row):
        return '%d %d %s' % (row.count, row.size, row.name)

    def get_stat_header(self):
        return (
            ' Index  Count     Size  Cumulative  % of B ')

    def get_stat_data(self, row):
        impl = self.impl
        format = '%6d %6d %8d %9d %s '
        fr = format % (
            row.index,
            row.count,
            row.size,
            row.cumulsize,
            self._percent_of_b(row.cumulsize),
        )
        return fr

    def load_statrow(self, r):
        return self.load_statrow_csk(r)


class StatRow(object):
    __slots__ = 'count', 'size', 'name', 'index', 'cumulsize'

    def __init__(self, count, size, name, index=None, cumulsize=None):
        self.count = count
        self.size = size
        self.name = name
        self.index = index
        self.cumulsize = cumulsize


class PartRow(StatRow):
    __slots__ = 'set', 'kind'

    def __init__(self, count, size, name, index, cumulsize, set, kind):
        self.count = count
        self.size = size
        self.name = name
        self.index = index
        self.cumulsize = cumulsize
        self.set = set
        self.kind = kind


class Stat:
    def __init__(self, mod, get_trows, firstheader=''):
        self.mod = mod
        self._hiding_tag_ = mod._hiding_tag_
        self.get_trows = get_trows
        self.firstheader = firstheader

        self.it = iter(get_trows())

        self.cur_index = 0
        self.cum_size = 0
        self.rows = []

        r = self.get_next()
        while r and not r.startswith('.r:'):
            name = r[1:r.index(':')]
            value = r[r.index(':')+1:].strip()
            try:
                value = int(value)
            except ValueError:
                pass
            setattr(self, name, value)
            r = self.get_next()

        self.format_name = self.format
        self.format_class = getattr(self.mod, self.format)
        self.format = self.format_class(self)

        self.timemade = float(self.timemade)

        self.mod.OutputHandling.setup_printing(self, self.format)

    def __getitem__(self, idx):
        if isinstance(idx, int):
            if idx < 0:
                idx = self.numrows + idx
            if not (0 <= idx < self.numrows):
                raise IndexError('Stat index out of range.')
            rows = [self.get_row(idx)]
        elif isinstance(idx, slice):
            start, stop, step = idx.indices(self.numrows)
            rows = [self.get_row(idx) for idx in range(start, stop, step)]
        else:
            raise IndexError('Stat indices must be integers or slices.')

        count = 0
        size = 0
        for r in rows:
            count += r.count
            size += r.size

        trows = [
            '.loader: _load_stat',
            '.format: %s' % self.format_name,
            '.timemade: %f' % self.timemade,
            '.count: %d' % count,
            '.size: %d' % size,
            '.kindheader: %s' % self.kindheader,
            '.kindname: %s' % self.kindname,
            '.numrows: %d' % len(rows),
        ]
        if getattr(self, 'b_count', None) is not None:
            trows.append('.b_count: %d' % self.b_count)
            trows.append('.b_size: %d' % self.b_size)
        for r in rows:
            trows.append('.r: %s' % self.format.get_rowdata(r))
        return self.mod.load(trows)

    def __len__(self):
        return self.numrows

    def __sub__(self, other):
        if not isinstance(other, Stat):
            raise TypeError(
                'Can only take difference with other Stat instance.')
        if self.kindheader != other.kindheader:
            raise ValueError('Mismatching table kind header, %r vs %r.' % (
                self.kindheader, other.kindheader))
        rows = []
        otab = {}
        stab = {}
        for r in other.get_rows():
            o = otab.get(r.name)
            if o:
                otab[r.name] = StatRow(
                    r.count+o.count, r.size+o.size, r.name, o.index,  None)
            else:
                otab[r.name] = r
        for r in self.get_rows():
            o = otab.get(r.name)
            if o:
                del otab[r.name]
                count = r.count - o.count
                size = r.size - o.size
            else:
                count = r.count
                size = r.size
            if count == 0 and size == 0:
                continue
            sr = stab.get(r.name)
            if sr:
                sr.count += count
                sr.size += size
            else:
                sr = StatRow(count, size, r.name)
                stab[sr.name] = sr
                rows.append(sr)
        rs = list(otab.values())
        rs.sort(key=lambda x: x.index)  # Preserve orig. order
        for r in rs:
            sr = StatRow(-r.count, -r.size, r.name)
            assert sr.name not in stab
            rows.append(sr)
        rows.sort(key=lambda x: abs(x.size), reverse=True)
        cumulcount = 0
        cumulsize = 0
        for r in rows:
            cumulcount += r.count
            cumulsize += r.size
            r.cumulsize = cumulsize
        trows = [
            '.loader: _load_stat',
            '.format: DiffFormat',
            '.timemade: %f' % self.mod.time.time(),
            '.b_count: %d' % other.count,
            '.b_size: %d' % other.size,
            '.count: %d' % cumulcount,
            '.size: %d' % cumulsize,
            '.kindheader: %s' % self.kindheader,
            '.kindname: %s' % self.kindname,
            '.numrows: %d' % len(rows),
        ]
        for r in rows:
            trows.append('.r: %d %d %s' % (r.count, r.size, r.name))
        return self.mod.load(trows)

    def dump(self, fn, mode='a'):
        if not hasattr(fn, 'write'):
            f = open(fn, mode)
        else:
            f = fn
        try:
            for r in self.get_trows():
                if not r[-1:] == '\n':
                    r += '\n'
                f.write(r)
            end = '.end: .loader: %s\n' % self.loader
            if r != end:
                f.write(end)
        finally:
            if f is not fn:
                f.close()

    def get_next(self):
        try:
            r = next(self.it)
        except StopIteration:
            r = None
        else:
            r = r.rstrip('\n')
        self.last = r
        return r

    def get_row(self, idx):
        while idx >= len(self.rows):
            self.parse_next_row()
        return self.rows[idx]

    def get_rows(self, idx=0):
        while idx < self.numrows:
            try:
                row = self.get_row(idx)
            except IndexError:
                return
            else:
                yield row
            idx += 1

    def get_rows_of_kinds(self, kinds):
        # Return the rows with names in sequence kinds of unique names
        # in that order. None if no such kind.

        kindtab = {}
        N = len(kinds)
        res = [None] * len(kinds)
        for i, kind in enumerate(kinds):
            kindtab[kind] = i
        assert len(kindtab) == N

        n = 0
        for row in self.get_rows():
            idx = kindtab.get(row.name)
            if idx is not None:
                res[idx] = row
            n += 1
            if n >= N:
                break
        return res

    def get_rows_n_and_other(self, N, sortby='Size'):
        # Get N rows, the largest first
        # mix in an '<Other>' row at a sorted position
        # Size is either size if sortby = 'Size',
        #    or count if sortby = 'Count'.
        # Returns a NEW LIST (caller may modify/sort it)

        if sortby not in ('Size', 'Count'):
            raise ValueError("Argument 'sortby' must be 'Size' or 'Count'.")

        # Rows are already sorted by Size, largest first.
        # If they want by Count, we need to resort them.

        rows = self.get_rows()

        if sortby == 'Count':
            rows = list(rows)
            rows.sort(key=lambda x: x.count, reverse=True)

        retrows = []
        cumulcount = 0
        cumulsize = 0

        for (i, r) in enumerate(rows):
            if i >= N:
                othercount = self.count - cumulcount
                othersize = self.size - cumulsize
                other = StatRow(othercount,
                                othersize,
                                '<Other>')

                if sortby == 'Size':
                    for (i, r) in enumerate(retrows):
                        if r.size < othersize:
                            retrows[i:i] = [other]
                            break
                    else:
                        retrows.append(other)
                elif sortby == 'Count':
                    for (i, r) in enumerate(retrows):
                        if r.count < othercount:
                            retrows[i:i] = [other]
                            break
                    else:
                        retrows.append(other)
                else:
                    assert 0
                break

            cumulcount += r.count
            cumulsize += r.size
            retrows.append(r)
        else:
            assert cumulcount == self.count
            assert cumulsize == self.size

        return retrows

    def parse_next_row(self):
        r = self.last
        if not r:
            raise IndexError('Row index out of range.')
        if r.startswith('.r: '):
            r = r[4:]
            sr = self.format.load_statrow(r)

            self.cur_index += 1
            self.rows.append(sr)
            self.get_next()
            return

        elif r.startswith('.end'):
            raise IndexError('Row index out of range.')
        else:
            raise SyntaxError


class Partition:
    def __init__(self, mod, set, er):
        self.mod = mod
        self.set = set
        self.er = er
        self._hiding_tag_ = mod._hiding_tag_
        self.timemade = mod.time.time()

    def __iter__(self):
        # The default iteration is over the sets
        # To iterate over rows (if more info is needed), get_rows() is available.
        return self.get_sets()

    def get_rows(self, rowindex=None):
        # Iterator over rows
        if rowindex is None:
            rowindex = 0
        while 1:
            try:
                row = self.get_row(rowindex)
            except IndexError:
                return
            else:
                yield row
            rowindex += 1

    def get_set(self, index):
        if isinstance(index, slice):
            start, stop, step = index.indices(self.numrows)
            ns = self.get_nodeset(start, stop, step)
            return self.mod.idset(ns, er=self.er)
        else:
            if index < 0:
                index += self.numrows
            return self.get_rowset(index)

    def get_sets(self, index=None):
        for idx in range(self.numrows):
            yield self.get_rowset(idx)

    def get_stat(self):
        # Avoid any references into the set!
        trows = list(self.get_trows())

        def get_trows():
            return trows
        return self.mod._load_stat(get_trows)

    def get_trows(self):
        yield '.loader: _load_stat'
        yield '.format: %s' % self.format.__class__.__name__
        yield '.timemade: %f' % self.timemade
        yield '.count: %d' % self.count
        yield '.size: %d' % self.size
        yield '.kindname: %s' % self.kindname
        yield '.kindheader: %s' % self.kindheader
        yield '.numrows: %d' % self.numrows
        for row in self.get_rows():
            yield '.r: %s' % self.format.get_rowdata(row)

    def init_format(self, FormatClass):
        self.format = FormatClass(self)

        self.mod.OutputHandling.setup_printing(self, self.format)


class IdentityPartitionCluster(object):
    # Contains objects of same size.
    # to speed up management of identity partition
    # - since otherwise we'd have to sort all the objects,
    #   on their string representation in worst case.
    __slots__ = 'objects', 'locount', 'hicount', 'losize', 'obsize', 'issorted'

    def __init__(self, objects, locount, count, losize, obsize):
        self.objects = objects          # tuple of objects in this segment
        self.locount = locount          # count BEFORE objects in this cluster
        self.hicount = locount+count    # count AFTER these objects
        self.losize = losize            # size BEFORE  objects in this cluster
        self.obsize = obsize            # size of EACH object in this segment
        self.issorted = False           # indicates if .objects is sorted


class IdentityPartition(Partition):
    def __init__(self, mod, set, er):
        Partition.__init__(self, mod, set, er)

        clusters = []
        sizeclasses = mod.Size.classifier.partition_cli(set.nodes)
        sizeclasses.sort()
        sizeclasses.reverse()
        totcount = 0
        totsize = 0
        for size, v in sizeclasses:
            count = len(v)
            clusters.append(IdentityPartitionCluster(
                self.mod.observation_list(v), totcount, count, totsize, size))
            totsize += size * count
            totcount += count
        assert totcount == set.count

        self.cluidx = 0
        self.clusters = clusters
        self.count = totcount
        self.kind = kind = set.byclodo.kind
        self.kindheader = kind.fam.c_get_idpart_header(kind)
        self.kindname = kind.fam.c_get_idpart_label(kind)
        self.numrows = totcount
        self.render = kind.fam.c_get_idpart_render(kind)
        self.size = totsize
        self.sortrender = kind.fam.c_get_idpart_sortrender(kind)

        self.init_format(IdFormat)

    def get_nodeset(self, start, stop, step):
        return self.get_nodeset_cluster(start, stop, step)[0]

    def get_nodeset_cluster(self, start, stop, step):
        if step <= 0:
            raise ValueError('Step must be positive.')
        ns = self.mod.mutnodeset()
        if start >= stop:
            return (ns, None)
        clusters = self.clusters
        lo = 0
        hi = len(clusters)
        cluidx = self.cluidx

        while lo < hi:
            clu = clusters[cluidx]
            if clu.locount <= start:
                if start < clu.hicount:
                    break
                else:
                    lo = cluidx + 1
            else:
                hi = cluidx
            cluidx = (lo + hi) // 2
        else:
            return (ns, None)
        clu_to_return = clu

        while 1:
            objects = clu.objects
            if start != clu.locount or stop < clu.hicount or step != 1:
                if not clu.issorted:
                    sortrender = self.sortrender
                    if sortrender == 'IDENTITY':
                        ks = objects
                    else:
                        ks = [sortrender(x) for x in objects]
                    ks = [(kind, i) for i, kind in enumerate(ks)]
                    ks.sort()
                    clu.objects = objects = self.mod.observation_list(
                        [objects[i] for (kind, i) in ks])
                    clu.issorted = True
                objects = objects[start-clu.locount:stop-clu.locount:step]

            ns |= objects
            self.cluidx = cluidx  # memo till next call
            start += len(objects)*step
            if start >= stop:
                break
            for cluidx in range(cluidx + 1, len(clusters)):
                clu = clusters[cluidx]
                if clu.locount <= start < clu.hicount:
                    break
            else:
                break
        return (ns, clu_to_return)

    def get_row(self, rowidx):
        ns, clu = self.get_nodeset_cluster(rowidx, rowidx+1, 1)
        if not ns:
            raise IndexError('Partition index out of range.')
        vi = self.mod.idset(ns, er=self.er)
        row = PartRow(1, clu.obsize, self.render(vi.theone),
                      rowidx, (rowidx+1-clu.locount)*clu.obsize + clu.losize,
                      vi, vi.kind)
        return row

    def get_rowset(self, rowidx):
        ns = self.get_nodeset(rowidx, rowidx+1, 1)
        if not ns:
            raise IndexError('Partition index out of range.')
        return self.mod.idset(ns, er=self.er)


class SetPartition(Partition):
    def __init__(self, mod, set, er):
        Partition.__init__(self, mod, set, er)

        classifier = er.classifier
        tosort = [(-part.size, classifier.get_tabrendering(kind, ''), kind, part)
                  for (kind, part) in classifier.partition(set.nodes)]
        tosort.sort()
        cumulsize = 0
        rows = []
        for (minusize, name, kind, part) in tosort:
            size = -minusize
            cumulsize += size
            rows.append(PartRow(
                part.count, size, name,
                len(rows), cumulsize,
                part, kind))

        self.count = set.count
        self.kindheader = classifier.get_tabheader('')
        self.kindname = ''
        self.numrows = len(rows)
        self.rows = rows
        self.size = cumulsize

        self.init_format(SetFormat)

    def get_nodeset(self, start, stop, step):
        if step <= 0:
            raise ValueError('Step must be positive.')
        ns = self.mod.mutnodeset()
        while start < stop:
            ns |= self.rows[start].set.nodes
            start += step
        return ns

    def get_row(self, idx):
        try:
            return self.rows[idx]
        except IndexError:
            raise IndexError('Partition index out of range.')

    def get_rowset(self, idx):
        row = self.get_row(idx)
        set = row.set
        try:
            set._partition
        except AttributeError:
            set._partition = self.mod._fast_partition(self, row)
        return set


class FastPartition(SetPartition):
    def __init__(self, mod, parent, row):
        Partition.__init__(self, mod, parent.set, parent.er)

        self.count = row.count
        self.kindheader = parent.kindheader
        self.kindname = ''
        self.numrows = 1
        self.rows = [row]
        self.size = row.size

        self.init_format(SetFormat)


class _GLUECLAMP_:
    _preload_ = ('_hiding_tag_',)
    _chgable_ = ('line_length', 'backup_suffix')
    _imports_ = (
        '_parent.ImpSet:mutnodeset',
        '_parent:OutputHandling',
        '_parent.Use:Id',
        '_parent.Use:Size',
        '_parent.Use:idset',
        '_parent.Use:load',
        '_parent.View:_hiding_tag_',
        '_parent.View:observation_list',
        '_root.os:rename',
        '_root.textwrap:fill',
        '_root.textwrap:wrap',
        '_root.textwrap:wrap',
        '_root:time',
    )

    # 'Config'

    line_length = 100
    backup_suffix = '.old'

    # Factory method

    def partition(self, set, er):
        if er.classifier is self.Id.classifier:
            return IdentityPartition(self, set, er)
        else:
            return SetPartition(self, set, er)

    # Private - Use.load is intended to be used directly.
    def _load_stat(self, get_trows):
        return Stat(self, get_trows)

    def _fast_partition(self, parent, row):
        return FastPartition(self, parent, row)