Tim Peters' program for checking whether a program's indentation is

sensitive to tab size.  Uses a different strategy than tabpolice.py,
but has compatible usage.

Lib/tabnanny.py

+#! /home/guido/python/src/sparc/python
+#! /usr/bin/env python
+
+"""The Tab Nanny despises ambiguous indentation.  She knows no mercy.
+
+CAUTION:  this version requires Guido's "NL" patch to lib/tokenize.py,
+posted 30-Mar-98.  This version will not run at all with an unpatched
+tokenize (it will raise AttributeError while loading), while previous
+versions will run incorrectly with the patched tokenize.
+"""
+
+# Released to the public domain, by Tim Peters, 30 March 1998.
+
+__version__ = "2"
+
+import os
+import sys
+import getopt
+import tokenize
+
+try:
+    tokenize.NL
+except AttributeError:
+    raise AttributeError, "Sorry, I need a version of tokenize.py " \
+                          "that supports the NL pseudo-token."
+
+verbose = 0
+
+def main():
+    global verbose
+    try:
+        opts, args = getopt.getopt(sys.argv[1:], "v")
+    except getopt.error, msg:
+        print msg
+    for o, a in opts:
+        if o == '-v':
+            verbose = verbose + 1
+    for arg in args:
+        check(arg)
+
+class NannyNag:
+    def __init__(self, lineno, msg, line):
+        self.lineno, self.msg, self.line = lineno, msg, line
+    def get_lineno(self):
+        return self.lineno
+    def get_msg(self):
+        return self.msg
+    def get_line(self):
+        return self.line
+
+def check(file):
+    if os.path.isdir(file) and not os.path.islink(file):
+        if verbose:
+            print "%s: listing directory" % `file`
+        names = os.listdir(file)
+        for name in names:
+            fullname = os.path.join(file, name)
+            if (os.path.isdir(fullname) and
+                not os.path.islink(fullname) or
+                os.path.normcase(name[-3:]) == ".py"):
+                check(fullname)
+        return
+
+    try:
+        f = open(file)
+    except IOError, msg:
+        print "%s: I/O Error: %s" % (`file`, str(msg))
+        return
+
+    if verbose > 1:
+        print "checking", `file`, "..."
+
+    reset_globals()
+    try:
+        tokenize.tokenize(f.readline, tokeneater)
+
+    except tokenize.TokenError, msg:
+        print "%s: Token Error: %s" % (`fname`, str(msg))
+        return
+
+    except NannyNag, nag:
+        badline = nag.get_lineno()
+        line = nag.get_line()
+        if verbose:
+            print "%s: *** Line %d: trouble in tab city! ***" % (
+                `file`, badline)
+            print "offending line:", `line`
+            print nag.get_msg()
+        else:
+            print file, badline, `line`
+        return
+
+    if verbose:
+        print "%s: Clean bill of health." % `file`
+
+class Whitespace:
+    # the characters used for space and tab
+    S, T = ' \t'
+
+    # members:
+    #   raw
+    #       the original string
+    #   n
+    #       the number of leading whitespace characters in raw
+    #   nt
+    #       the number of tabs in raw[:n]
+    #   norm
+    #       the normal form as a pair (count, trailing), where:
+    #       count
+    #           a tuple such that raw[:n] contains count[i]
+    #           instances of S * i + T
+    #       trailing
+    #           the number of trailing spaces in raw[:n]
+    #       It's A Theorem that m.indent_level(t) ==
+    #       n.indent_level(t) for all t >= 1 iff m.norm == n.norm.
+    #   is_simple
+    #       true iff raw[:n] is of the form (T*)(S*)
+
+    def __init__(self, ws):
+        self.raw  = ws
+        S, T = Whitespace.S, Whitespace.T
+        count = []
+        b = n = nt = 0
+        for ch in self.raw:
+            if ch == S:
+                n = n + 1
+                b = b + 1
+            elif ch == T:
+                n = n + 1
+                nt = nt + 1
+                if b >= len(count):
+                    count = count + [0] * (b - len(count) + 1)
+                count[b] = count[b] + 1
+                b = 0
+            else:
+                break
+        self.n    = n
+        self.nt   = nt
+        self.norm = tuple(count), b
+        self.is_simple = len(count) <= 1
+
+    # return length of longest contiguous run of spaces (whether or not
+    # preceding a tab)
+    def longest_run_of_spaces(self):
+        count, trailing = self.norm
+        return max(len(count)-1, trailing)
+
+    def indent_level(self, tabsize):
+        # count, il = self.norm
+        # for i in range(len(count)):
+        #    if count[i]:
+        #        il = il + (i/tabsize + 1)*tabsize * count[i]
+        # return il
+
+        # quicker:
+        # il = trailing + sum (i/ts + 1)*ts*count[i] =
+        # trailing + ts * sum (i/ts + 1)*count[i] =
+        # trailing + ts * sum i/ts*count[i] + count[i] =
+        # trailing + ts * [(sum i/ts*count[i]) + (sum count[i])] =
+        # trailing + ts * [(sum i/ts*count[i]) + num_tabs]
+        # and note that i/ts*count[i] is 0 when i < ts
+
+        count, trailing = self.norm
+        il = 0
+        for i in range(tabsize, len(count)):
+            il = il + i/tabsize * count[i]
+        return trailing + tabsize * (il + self.nt)
+
+    # return true iff self.indent_level(t) == other.indent_level(t)
+    # for all t >= 1
+    def equal(self, other):
+        return self.norm == other.norm
+
+    # return a list of tuples (ts, i1, i2) such that
+    # i1 == self.indent_level(ts) != other.indent_level(ts) == i2.
+    # Intended to be used after not self.equal(other) is known, in which
+    # case it will return at least one witnessing tab size.
+    def not_equal_witness(self, other):
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        a = []
+        for ts in range(1, n+1):
+            if self.indent_level(ts) != other.indent_level(ts):
+                a.append( (ts,
+                           self.indent_level(ts),
+                           other.indent_level(ts)) )
+        return a
+
+    # Return true iff self.indent_level(t) < other.indent_level(t)
+    # for all t >= 1.
+    # The algorithm is due to Vincent Broman.
+    # Easy to prove it's correct.
+    # XXXpost that.
+    # Trivial to prove n is sharp (consider T vs ST).
+    # Unknown whether there's a faster general way.  I suspected so at
+    # first, but no longer.
+    # For the special (but common!) case where M and N are both of the
+    # form (T*)(S*), M.less(N) iff M.len() < N.len() and
+    # M.num_tabs() <= N.num_tabs(). Proof is easy but kinda long-winded.
+    # XXXwrite that up.
+    # Note that M is of the form (T*)(S*) iff len(M.norm[0]) <= 1.
+    def less(self, other):
+        if self.n >= other.n:
+            return 0
+        if self.is_simple and other.is_simple:
+            return self.nt <= other.nt
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        # the self.n >= other.n test already did it for ts=1
+        for ts in range(2, n+1):
+            if self.indent_level(ts) >= other.indent_level(ts):
+                return 0
+        return 1
+
+    # return a list of tuples (ts, i1, i2) such that
+    # i1 == self.indent_level(ts) >= other.indent_level(ts) == i2.
+    # Intended to be used after not self.less(other) is known, in which
+    # case it will return at least one witnessing tab size.
+    def not_less_witness(self, other):
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        a = []
+        for ts in range(1, n+1):
+            if self.indent_level(ts) >= other.indent_level(ts):
+                a.append( (ts,
+                           self.indent_level(ts),
+                           other.indent_level(ts)) )
+        return a
+
+def format_witnesses(w):
+    import string
+    firsts = map(lambda tup: str(tup[0]), w)
+    prefix = "at tab size"
+    if len(w) > 1:
+        prefix = prefix + "s"
+    return prefix + " " + string.join(firsts, ', ')
+
+indents = []
+check_equal = 0
+
+def reset_globals():
+    global indents, check_equal
+    check_equal = 0
+    indents = [Whitespace("")]
+
+def tokeneater(type, token, start, end, line,
+               INDENT=tokenize.INDENT,
+               DEDENT=tokenize.DEDENT,
+               NEWLINE=tokenize.NEWLINE,
+               COMMENT=tokenize.COMMENT,
+               NL=tokenize.NL):
+    global indents, check_equal
+
+    # test in decreasing order of frequency, although the check_equal
+    # test *must* be last; INDENT and DEDENT appear equally often
+
+    if type in (COMMENT, NL):
+        # the indentation of these guys is meaningless
+        pass
+
+    elif type == NEWLINE:
+        # a program statement, or ENDMARKER, will eventually follow,
+        # after some (possibly empty) run of tokens of the form
+        #     (NL | COMMENT)* (INDENT | DEDENT+)?
+        # If an INDENT appears, setting check_equal is wrong, and will
+        # be undone when we see the INDENT.
+        check_equal = 1
+
+    elif type == INDENT:
+        check_equal = 0
+        thisguy = Whitespace(token)
+        if not indents[-1].less(thisguy):
+            witness = indents[-1].not_less_witness(thisguy)
+            msg = "indent not greater e.g. " + format_witnesses(witness)
+            raise NannyNag(start[0], msg, line)
+        indents.append(thisguy)
+
+    elif type == DEDENT:
+        # there's nothing we need to check here!  what's important is
+        # that when the run of DEDENTs ends, the indentation of the
+        # program statement (or ENDMARKER) that triggered the run is
+        # equal to what's left at the top of the indents stack
+        assert check_equal  # else no earlier NEWLINE, or an earlier INDENT
+        del indents[-1]
+
+    elif check_equal:
+        # this is the first "real token" following a NEWLINE, so it
+        # must be the first token of the next program statment, or an
+        # ENDMARKER; the "line" argument exposes the leading whitespace
+        # for this statement; in the case of ENDMARKER, line is an empty
+        # string, so will properly match the empty string with which the
+        # "indents" stack was seeded
+        check_equal = 0
+        thisguy = Whitespace(line)
+        if not indents[-1].equal(thisguy):
+            witness = indents[-1].not_equal_witness(thisguy)
+            msg = "indent not equal e.g. " + format_witnesses(witness)
+            raise NannyNag(start[0], msg, line)
+
+if __name__ == '__main__':
+    main()
+

Tools/idle/tabnanny.py

+#! /home/guido/python/src/sparc/python
+#! /usr/bin/env python
+
+"""The Tab Nanny despises ambiguous indentation.  She knows no mercy.
+
+CAUTION:  this version requires Guido's "NL" patch to lib/tokenize.py,
+posted 30-Mar-98.  This version will not run at all with an unpatched
+tokenize (it will raise AttributeError while loading), while previous
+versions will run incorrectly with the patched tokenize.
+"""
+
+# Released to the public domain, by Tim Peters, 30 March 1998.
+
+__version__ = "2"
+
+import os
+import sys
+import getopt
+import tokenize
+
+try:
+    tokenize.NL
+except AttributeError:
+    raise AttributeError, "Sorry, I need a version of tokenize.py " \
+                          "that supports the NL pseudo-token."
+
+verbose = 0
+
+def main():
+    global verbose
+    try:
+        opts, args = getopt.getopt(sys.argv[1:], "v")
+    except getopt.error, msg:
+        print msg
+    for o, a in opts:
+        if o == '-v':
+            verbose = verbose + 1
+    for arg in args:
+        check(arg)
+
+class NannyNag:
+    def __init__(self, lineno, msg, line):
+        self.lineno, self.msg, self.line = lineno, msg, line
+    def get_lineno(self):
+        return self.lineno
+    def get_msg(self):
+        return self.msg
+    def get_line(self):
+        return self.line
+
+def check(file):
+    if os.path.isdir(file) and not os.path.islink(file):
+        if verbose:
+            print "%s: listing directory" % `file`
+        names = os.listdir(file)
+        for name in names:
+            fullname = os.path.join(file, name)
+            if (os.path.isdir(fullname) and
+                not os.path.islink(fullname) or
+                os.path.normcase(name[-3:]) == ".py"):
+                check(fullname)
+        return
+
+    try:
+        f = open(file)
+    except IOError, msg:
+        print "%s: I/O Error: %s" % (`file`, str(msg))
+        return
+
+    if verbose > 1:
+        print "checking", `file`, "..."
+
+    reset_globals()
+    try:
+        tokenize.tokenize(f.readline, tokeneater)
+
+    except tokenize.TokenError, msg:
+        print "%s: Token Error: %s" % (`fname`, str(msg))
+        return
+
+    except NannyNag, nag:
+        badline = nag.get_lineno()
+        line = nag.get_line()
+        if verbose:
+            print "%s: *** Line %d: trouble in tab city! ***" % (
+                `file`, badline)
+            print "offending line:", `line`
+            print nag.get_msg()
+        else:
+            print file, badline, `line`
+        return
+
+    if verbose:
+        print "%s: Clean bill of health." % `file`
+
+class Whitespace:
+    # the characters used for space and tab
+    S, T = ' \t'
+
+    # members:
+    #   raw
+    #       the original string
+    #   n
+    #       the number of leading whitespace characters in raw
+    #   nt
+    #       the number of tabs in raw[:n]
+    #   norm
+    #       the normal form as a pair (count, trailing), where:
+    #       count
+    #           a tuple such that raw[:n] contains count[i]
+    #           instances of S * i + T
+    #       trailing
+    #           the number of trailing spaces in raw[:n]
+    #       It's A Theorem that m.indent_level(t) ==
+    #       n.indent_level(t) for all t >= 1 iff m.norm == n.norm.
+    #   is_simple
+    #       true iff raw[:n] is of the form (T*)(S*)
+
+    def __init__(self, ws):
+        self.raw  = ws
+        S, T = Whitespace.S, Whitespace.T
+        count = []
+        b = n = nt = 0
+        for ch in self.raw:
+            if ch == S:
+                n = n + 1
+                b = b + 1
+            elif ch == T:
+                n = n + 1
+                nt = nt + 1
+                if b >= len(count):
+                    count = count + [0] * (b - len(count) + 1)
+                count[b] = count[b] + 1
+                b = 0
+            else:
+                break
+        self.n    = n
+        self.nt   = nt
+        self.norm = tuple(count), b
+        self.is_simple = len(count) <= 1
+
+    # return length of longest contiguous run of spaces (whether or not
+    # preceding a tab)
+    def longest_run_of_spaces(self):
+        count, trailing = self.norm
+        return max(len(count)-1, trailing)
+
+    def indent_level(self, tabsize):
+        # count, il = self.norm
+        # for i in range(len(count)):
+        #    if count[i]:
+        #        il = il + (i/tabsize + 1)*tabsize * count[i]
+        # return il
+
+        # quicker:
+        # il = trailing + sum (i/ts + 1)*ts*count[i] =
+        # trailing + ts * sum (i/ts + 1)*count[i] =
+        # trailing + ts * sum i/ts*count[i] + count[i] =
+        # trailing + ts * [(sum i/ts*count[i]) + (sum count[i])] =
+        # trailing + ts * [(sum i/ts*count[i]) + num_tabs]
+        # and note that i/ts*count[i] is 0 when i < ts
+
+        count, trailing = self.norm
+        il = 0
+        for i in range(tabsize, len(count)):
+            il = il + i/tabsize * count[i]
+        return trailing + tabsize * (il + self.nt)
+
+    # return true iff self.indent_level(t) == other.indent_level(t)
+    # for all t >= 1
+    def equal(self, other):
+        return self.norm == other.norm
+
+    # return a list of tuples (ts, i1, i2) such that
+    # i1 == self.indent_level(ts) != other.indent_level(ts) == i2.
+    # Intended to be used after not self.equal(other) is known, in which
+    # case it will return at least one witnessing tab size.
+    def not_equal_witness(self, other):
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        a = []
+        for ts in range(1, n+1):
+            if self.indent_level(ts) != other.indent_level(ts):
+                a.append( (ts,
+                           self.indent_level(ts),
+                           other.indent_level(ts)) )
+        return a
+
+    # Return true iff self.indent_level(t) < other.indent_level(t)
+    # for all t >= 1.
+    # The algorithm is due to Vincent Broman.
+    # Easy to prove it's correct.
+    # XXXpost that.
+    # Trivial to prove n is sharp (consider T vs ST).
+    # Unknown whether there's a faster general way.  I suspected so at
+    # first, but no longer.
+    # For the special (but common!) case where M and N are both of the
+    # form (T*)(S*), M.less(N) iff M.len() < N.len() and
+    # M.num_tabs() <= N.num_tabs(). Proof is easy but kinda long-winded.
+    # XXXwrite that up.
+    # Note that M is of the form (T*)(S*) iff len(M.norm[0]) <= 1.
+    def less(self, other):
+        if self.n >= other.n:
+            return 0
+        if self.is_simple and other.is_simple:
+            return self.nt <= other.nt
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        # the self.n >= other.n test already did it for ts=1
+        for ts in range(2, n+1):
+            if self.indent_level(ts) >= other.indent_level(ts):
+                return 0
+        return 1
+
+    # return a list of tuples (ts, i1, i2) such that
+    # i1 == self.indent_level(ts) >= other.indent_level(ts) == i2.
+    # Intended to be used after not self.less(other) is known, in which
+    # case it will return at least one witnessing tab size.
+    def not_less_witness(self, other):
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        a = []
+        for ts in range(1, n+1):
+            if self.indent_level(ts) >= other.indent_level(ts):
+                a.append( (ts,
+                           self.indent_level(ts),
+                           other.indent_level(ts)) )
+        return a
+
+def format_witnesses(w):
+    import string
+    firsts = map(lambda tup: str(tup[0]), w)
+    prefix = "at tab size"
+    if len(w) > 1:
+        prefix = prefix + "s"
+    return prefix + " " + string.join(firsts, ', ')
+
+indents = []
+check_equal = 0
+
+def reset_globals():
+    global indents, check_equal
+    check_equal = 0
+    indents = [Whitespace("")]
+
+def tokeneater(type, token, start, end, line,
+               INDENT=tokenize.INDENT,
+               DEDENT=tokenize.DEDENT,
+               NEWLINE=tokenize.NEWLINE,
+               COMMENT=tokenize.COMMENT,
+               NL=tokenize.NL):
+    global indents, check_equal
+
+    # test in decreasing order of frequency, although the check_equal
+    # test *must* be last; INDENT and DEDENT appear equally often
+
+    if type in (COMMENT, NL):
+        # the indentation of these guys is meaningless
+        pass
+
+    elif type == NEWLINE:
+        # a program statement, or ENDMARKER, will eventually follow,
+        # after some (possibly empty) run of tokens of the form
+        #     (NL | COMMENT)* (INDENT | DEDENT+)?
+        # If an INDENT appears, setting check_equal is wrong, and will
+        # be undone when we see the INDENT.
+        check_equal = 1
+
+    elif type == INDENT:
+        check_equal = 0
+        thisguy = Whitespace(token)
+        if not indents[-1].less(thisguy):
+            witness = indents[-1].not_less_witness(thisguy)
+            msg = "indent not greater e.g. " + format_witnesses(witness)
+            raise NannyNag(start[0], msg, line)
+        indents.append(thisguy)
+
+    elif type == DEDENT:
+        # there's nothing we need to check here!  what's important is
+        # that when the run of DEDENTs ends, the indentation of the
+        # program statement (or ENDMARKER) that triggered the run is
+        # equal to what's left at the top of the indents stack
+        assert check_equal  # else no earlier NEWLINE, or an earlier INDENT
+        del indents[-1]
+
+    elif check_equal:
+        # this is the first "real token" following a NEWLINE, so it
+        # must be the first token of the next program statment, or an
+        # ENDMARKER; the "line" argument exposes the leading whitespace
+        # for this statement; in the case of ENDMARKER, line is an empty
+        # string, so will properly match the empty string with which the
+        # "indents" stack was seeded
+        check_equal = 0
+        thisguy = Whitespace(line)
+        if not indents[-1].equal(thisguy):
+            witness = indents[-1].not_equal_witness(thisguy)
+            msg = "indent not equal e.g. " + format_witnesses(witness)
+            raise NannyNag(start[0], msg, line)
+
+if __name__ == '__main__':
+    main()
+

Tools/scripts/tabnanny.py

+#! /home/guido/python/src/sparc/python
+#! /usr/bin/env python
+
+"""The Tab Nanny despises ambiguous indentation.  She knows no mercy.
+
+CAUTION:  this version requires Guido's "NL" patch to lib/tokenize.py,
+posted 30-Mar-98.  This version will not run at all with an unpatched
+tokenize (it will raise AttributeError while loading), while previous
+versions will run incorrectly with the patched tokenize.
+"""
+
+# Released to the public domain, by Tim Peters, 30 March 1998.
+
+__version__ = "2"
+
+import os
+import sys
+import getopt
+import tokenize
+
+try:
+    tokenize.NL
+except AttributeError:
+    raise AttributeError, "Sorry, I need a version of tokenize.py " \
+                          "that supports the NL pseudo-token."
+
+verbose = 0
+
+def main():
+    global verbose
+    try:
+        opts, args = getopt.getopt(sys.argv[1:], "v")
+    except getopt.error, msg:
+        print msg
+    for o, a in opts:
+        if o == '-v':
+            verbose = verbose + 1
+    for arg in args:
+        check(arg)
+
+class NannyNag:
+    def __init__(self, lineno, msg, line):
+        self.lineno, self.msg, self.line = lineno, msg, line
+    def get_lineno(self):
+        return self.lineno
+    def get_msg(self):
+        return self.msg
+    def get_line(self):
+        return self.line
+
+def check(file):
+    if os.path.isdir(file) and not os.path.islink(file):
+        if verbose:
+            print "%s: listing directory" % `file`
+        names = os.listdir(file)
+        for name in names:
+            fullname = os.path.join(file, name)
+            if (os.path.isdir(fullname) and
+                not os.path.islink(fullname) or
+                os.path.normcase(name[-3:]) == ".py"):
+                check(fullname)
+        return
+
+    try:
+        f = open(file)
+    except IOError, msg:
+        print "%s: I/O Error: %s" % (`file`, str(msg))
+        return
+
+    if verbose > 1:
+        print "checking", `file`, "..."
+
+    reset_globals()
+    try:
+        tokenize.tokenize(f.readline, tokeneater)
+
+    except tokenize.TokenError, msg:
+        print "%s: Token Error: %s" % (`fname`, str(msg))
+        return
+
+    except NannyNag, nag:
+        badline = nag.get_lineno()
+        line = nag.get_line()
+        if verbose:
+            print "%s: *** Line %d: trouble in tab city! ***" % (
+                `file`, badline)
+            print "offending line:", `line`
+            print nag.get_msg()
+        else:
+            print file, badline, `line`
+        return
+
+    if verbose:
+        print "%s: Clean bill of health." % `file`
+
+class Whitespace:
+    # the characters used for space and tab
+    S, T = ' \t'
+
+    # members:
+    #   raw
+    #       the original string
+    #   n
+    #       the number of leading whitespace characters in raw
+    #   nt
+    #       the number of tabs in raw[:n]
+    #   norm
+    #       the normal form as a pair (count, trailing), where:
+    #       count
+    #           a tuple such that raw[:n] contains count[i]
+    #           instances of S * i + T
+    #       trailing
+    #           the number of trailing spaces in raw[:n]
+    #       It's A Theorem that m.indent_level(t) ==
+    #       n.indent_level(t) for all t >= 1 iff m.norm == n.norm.
+    #   is_simple
+    #       true iff raw[:n] is of the form (T*)(S*)
+
+    def __init__(self, ws):
+        self.raw  = ws
+        S, T = Whitespace.S, Whitespace.T
+        count = []
+        b = n = nt = 0
+        for ch in self.raw:
+            if ch == S:
+                n = n + 1
+                b = b + 1
+            elif ch == T:
+                n = n + 1
+                nt = nt + 1
+                if b >= len(count):
+                    count = count + [0] * (b - len(count) + 1)
+                count[b] = count[b] + 1
+                b = 0
+            else:
+                break
+        self.n    = n
+        self.nt   = nt
+        self.norm = tuple(count), b
+        self.is_simple = len(count) <= 1
+
+    # return length of longest contiguous run of spaces (whether or not
+    # preceding a tab)
+    def longest_run_of_spaces(self):
+        count, trailing = self.norm
+        return max(len(count)-1, trailing)
+
+    def indent_level(self, tabsize):
+        # count, il = self.norm
+        # for i in range(len(count)):
+        #    if count[i]:
+        #        il = il + (i/tabsize + 1)*tabsize * count[i]
+        # return il
+
+        # quicker:
+        # il = trailing + sum (i/ts + 1)*ts*count[i] =
+        # trailing + ts * sum (i/ts + 1)*count[i] =
+        # trailing + ts * sum i/ts*count[i] + count[i] =
+        # trailing + ts * [(sum i/ts*count[i]) + (sum count[i])] =
+        # trailing + ts * [(sum i/ts*count[i]) + num_tabs]
+        # and note that i/ts*count[i] is 0 when i < ts
+
+        count, trailing = self.norm
+        il = 0
+        for i in range(tabsize, len(count)):
+            il = il + i/tabsize * count[i]
+        return trailing + tabsize * (il + self.nt)
+
+    # return true iff self.indent_level(t) == other.indent_level(t)
+    # for all t >= 1
+    def equal(self, other):
+        return self.norm == other.norm
+
+    # return a list of tuples (ts, i1, i2) such that
+    # i1 == self.indent_level(ts) != other.indent_level(ts) == i2.
+    # Intended to be used after not self.equal(other) is known, in which
+    # case it will return at least one witnessing tab size.
+    def not_equal_witness(self, other):
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        a = []
+        for ts in range(1, n+1):
+            if self.indent_level(ts) != other.indent_level(ts):
+                a.append( (ts,
+                           self.indent_level(ts),
+                           other.indent_level(ts)) )
+        return a
+
+    # Return true iff self.indent_level(t) < other.indent_level(t)
+    # for all t >= 1.
+    # The algorithm is due to Vincent Broman.
+    # Easy to prove it's correct.
+    # XXXpost that.
+    # Trivial to prove n is sharp (consider T vs ST).
+    # Unknown whether there's a faster general way.  I suspected so at
+    # first, but no longer.
+    # For the special (but common!) case where M and N are both of the
+    # form (T*)(S*), M.less(N) iff M.len() < N.len() and
+    # M.num_tabs() <= N.num_tabs(). Proof is easy but kinda long-winded.
+    # XXXwrite that up.
+    # Note that M is of the form (T*)(S*) iff len(M.norm[0]) <= 1.
+    def less(self, other):
+        if self.n >= other.n:
+            return 0
+        if self.is_simple and other.is_simple:
+            return self.nt <= other.nt
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        # the self.n >= other.n test already did it for ts=1
+        for ts in range(2, n+1):
+            if self.indent_level(ts) >= other.indent_level(ts):
+                return 0
+        return 1
+
+    # return a list of tuples (ts, i1, i2) such that
+    # i1 == self.indent_level(ts) >= other.indent_level(ts) == i2.
+    # Intended to be used after not self.less(other) is known, in which
+    # case it will return at least one witnessing tab size.
+    def not_less_witness(self, other):
+        n = max(self.longest_run_of_spaces(),
+                other.longest_run_of_spaces()) + 1
+        a = []
+        for ts in range(1, n+1):
+            if self.indent_level(ts) >= other.indent_level(ts):
+                a.append( (ts,
+                           self.indent_level(ts),
+                           other.indent_level(ts)) )
+        return a
+
+def format_witnesses(w):
+    import string
+    firsts = map(lambda tup: str(tup[0]), w)
+    prefix = "at tab size"
+    if len(w) > 1:
+        prefix = prefix + "s"
+    return prefix + " " + string.join(firsts, ', ')
+
+indents = []
+check_equal = 0
+
+def reset_globals():
+    global indents, check_equal
+    check_equal = 0
+    indents = [Whitespace("")]
+
+def tokeneater(type, token, start, end, line,
+               INDENT=tokenize.INDENT,
+               DEDENT=tokenize.DEDENT,
+               NEWLINE=tokenize.NEWLINE,
+               COMMENT=tokenize.COMMENT,
+               NL=tokenize.NL):
+    global indents, check_equal
+
+    # test in decreasing order of frequency, although the check_equal
+    # test *must* be last; INDENT and DEDENT appear equally often
+
+    if type in (COMMENT, NL):
+        # the indentation of these guys is meaningless
+        pass
+
+    elif type == NEWLINE:
+        # a program statement, or ENDMARKER, will eventually follow,
+        # after some (possibly empty) run of tokens of the form
+        #     (NL | COMMENT)* (INDENT | DEDENT+)?
+        # If an INDENT appears, setting check_equal is wrong, and will
+        # be undone when we see the INDENT.
+        check_equal = 1
+
+    elif type == INDENT:
+        check_equal = 0
+        thisguy = Whitespace(token)
+        if not indents[-1].less(thisguy):
+            witness = indents[-1].not_less_witness(thisguy)
+            msg = "indent not greater e.g. " + format_witnesses(witness)
+            raise NannyNag(start[0], msg, line)
+        indents.append(thisguy)
+
+    elif type == DEDENT:
+        # there's nothing we need to check here!  what's important is
+        # that when the run of DEDENTs ends, the indentation of the
+        # program statement (or ENDMARKER) that triggered the run is
+        # equal to what's left at the top of the indents stack
+        assert check_equal  # else no earlier NEWLINE, or an earlier INDENT
+        del indents[-1]
+
+    elif check_equal:
+        # this is the first "real token" following a NEWLINE, so it
+        # must be the first token of the next program statment, or an
+        # ENDMARKER; the "line" argument exposes the leading whitespace
+        # for this statement; in the case of ENDMARKER, line is an empty
+        # string, so will properly match the empty string with which the
+        # "indents" stack was seeded
+        check_equal = 0
+        thisguy = Whitespace(line)
+        if not indents[-1].equal(thisguy):
+            witness = indents[-1].not_equal_witness(thisguy)
+            msg = "indent not equal e.g. " + format_witnesses(witness)
+            raise NannyNag(start[0], msg, line)
+
+if __name__ == '__main__':
+    main()
+