Adiciona nova versao tgrep

Erickson Silva
1 parent 4fd9f50a
Showing 1 changed file with 144 additions and 122 deletions Show diff stats
src/new/tgrep.py
@@ -38,6 +38,7 @@ Tgrep2 source:
 http://tedlab.mit.edu/~dr/Tgrep2/
 '''
+from builtins import bytes, range, str
 import nltk.tree
 import pyparsing
 import re
@@ -48,11 +49,12 @@ def ancestors(node):
     This method will not work with leaf nodes, since there is no way
     to recover the parent.
     '''
-    # if node is a leaf, we cannot retrieve its parent
-    if not hasattr(node, 'parent'):
-        return []
     results = []
-    current = node.parent()
+    try:
+        current = node.parent()
+    except AttributeError:
+        # if node is a leaf, we cannot retrieve its parent
+        return results
     while current:
         results.append(current)
         current = current.parent()
@@ -63,11 +65,12 @@ def unique_ancestors(node):
     Returns the list of all nodes dominating the given node, where
     there is only a single path of descent.
     '''
-    # if node is a leaf, we cannot retrieve its parent
-    if not hasattr(node, 'parent'):
-        return []
     results = []
-    current = node.parent()
+    try:
+        current = node.parent()
+    except AttributeError:
+        # if node is a leaf, we cannot retrieve its parent
+        return results
     while current and len(current) == 1:
         results.append(current)
         current = current.parent()
@@ -78,29 +81,38 @@ def _descendants(node):
     Returns the list of all nodes which are descended from the given
     tree node in some way.
     '''
-    if not hasattr(node, 'treepositions'):
+    try:
+        treepos = node.treepositions()
+    except AttributeError:
         return []
-    return [node[x] for x in node.treepositions()[1:]]
+    return [node[x] for x in treepos[1:]]
 def _leftmost_descendants(node):
     '''
     Returns the set of all nodes descended in some way through
     left branches from this node.
     '''
-    if not hasattr(node, 'treepositions'):
+    try:
+        treepos = node.treepositions()
+    except AttributeError:
         return []
-    return [node[x] for x in node.treepositions()[1:] if all(y == 0 for y in x)]
+    return [node[x] for x in treepos[1:] if all(y == 0 for y in x)]
 def _rightmost_descendants(node):
     '''
     Returns the set of all nodes descended in some way through
     right branches from this node.
     '''
-    if not hasattr(node, 'treepositions'):
+    try:
+        rightmost_leaf = max(node.treepositions())
+    except AttributeError:
         return []
-    rightmost_leaf = max(node.treepositions())
     return [node[rightmost_leaf[:i]] for i in range(1, len(rightmost_leaf) + 1)]
+def _istree(obj):
+    '''Predicate to check whether `obj` is a nltk.tree.Tree.'''
+    return isinstance(obj, nltk.tree.Tree)
+
 def _unique_descendants(node):
     '''
     Returns the list of all nodes descended from the given node, where
@@ -108,7 +120,7 @@ def _unique_descendants(node):
     '''
     results = []
     current = node
-    while current and isinstance(current, nltk.tree.Tree) and len(current) == 1:
+    while current and _istree(current) and len(current) == 1:
         current = current[0]
         results.append(current)
     return results
@@ -117,10 +129,11 @@ def _before(node):
     '''
     Returns the set of all nodes that are before the given node.
     '''
-    if not hasattr(node, 'root') or not hasattr(node, 'treeposition'):
+    try:
+        pos = node.treeposition()
+        tree = node.root()
+    except AttributeError:
         return []
-    pos = node.treeposition()
-    tree = node.root()
     return [tree[x] for x in tree.treepositions()
             if x[:len(pos)] < pos[:len(x)]]
@@ -133,9 +146,11 @@ def _immediately_before(node):
     symbol (word) produced by A immediately precedes the first
     terminal symbol produced by B.
     '''
-    if not hasattr(node, 'root') or not hasattr(node, 'treeposition'):
+    try:
+        pos = node.treeposition()
+        tree = node.root()
+    except AttributeError:
         return []
-    pos = node.treeposition()
     # go "upwards" from pos until there is a place we can go to the left
     idx = len(pos) - 1
     while 0 <= idx and pos[idx] == 0:
@@ -144,17 +159,18 @@ def _immediately_before(node):
         return []
     pos = list(pos[:idx + 1])
     pos[-1] -= 1
-    before = node.root()[pos]
+    before = tree[pos]
     return [before] + _rightmost_descendants(before)
 def _after(node):
     '''
     Returns the set of all nodes that are after the given node.
     '''
-    if not hasattr(node, 'root') or not hasattr(node, 'treeposition'):
+    try:
+        pos = node.treeposition()
+        tree = node.root()
+    except AttributeError:
         return []
-    pos = node.treeposition()
-    tree = node.root()
     return [tree[x] for x in tree.treepositions()
             if x[:len(pos)] > pos[:len(x)]]
@@ -167,14 +183,15 @@ def _immediately_after(node):
     symbol (word) produced by A immediately follows the last
     terminal symbol produced by B.
     '''
-    if (not hasattr(node, 'root') or not hasattr(node, 'treeposition') or
-        not hasattr(node, 'parent')):
+    try:
+        pos = node.treeposition()
+        tree = node.root()
+        current = node.parent()
+    except AttributeError:
         return []
-    pos = node.treeposition()
     # go "upwards" from pos until there is a place we can go to the
     # right
     idx = len(pos) - 1
-    current = node.parent()
     while 0 <= idx and pos[idx] == len(current) - 1:
         idx -= 1
         current = current.parent()
@@ -182,7 +199,7 @@ def _immediately_after(node):
         return []
     pos = list(pos[:idx + 1])
     pos[-1] += 1
-    after = node.root()[pos]
+    after = tree[pos]
     return [after] + _leftmost_descendants(after)
 def _tgrep_node_literal_value(node):
@@ -190,7 +207,7 @@ def _tgrep_node_literal_value(node):
     Gets the string value of a given parse tree node, for comparison
     using the tgrep node literal predicates.
     '''
-    return (node.label() if isinstance(node, nltk.tree.Tree) else unicode(node))
+    return (node.label() if _istree(node) else str(node))
 def _tgrep_node_action(_s, _l, tokens):
     '''
@@ -198,30 +215,30 @@ def _tgrep_node_action(_s, _l, tokens):
     depending on the name of its node.
     '''
     # print 'node tokens: ', tokens
-    if tokens[0] == "'":
+    if tokens[0] == u"'":
         # strip initial apostrophe (tgrep2 print command)
         tokens = tokens[1:]
     if len(tokens) > 1:
         # disjunctive definition of a node name
-        assert list(set(tokens[1::2])) == ['|']
+        assert list(set(tokens[1::2])) == [u'|']
         # recursively call self to interpret each node name definition
         tokens = [_tgrep_node_action(None, None, [node])
                   for node in tokens[::2]]
         # capture tokens and return the disjunction
         return (lambda t: lambda n: any(f(n) for f in t))(tokens)
     else:
-        if hasattr(tokens[0], '__call__'):
+        if hasattr(tokens[0], u'__call__'):
             # this is a previously interpreted parenthetical node
             # definition (lambda function)
             return tokens[0]
-        elif tokens[0] == '*' or tokens[0] == '__':
+        elif tokens[0] == u'*' or tokens[0] == u'__':
             return lambda n: True
-        elif tokens[0].startswith('"'):
-            return (lambda s: lambda n: _tgrep_node_literal_value(n) == s)(tokens[0].strip('"'))
-        elif tokens[0].startswith('/'):
+        elif tokens[0].startswith(u'"'):
+            return (lambda s: lambda n: _tgrep_node_literal_value(n) == s)(tokens[0].strip(u'"'))
+        elif tokens[0].startswith(u'/'):
             return (lambda r: lambda n:
-                        r.match(_tgrep_node_literal_value(n)))(re.compile(tokens[0].strip('/')))
-        elif tokens[0].startswith('i@'):
+                        r.match(_tgrep_node_literal_value(n)))(re.compile(tokens[0].strip(u'/')))
+        elif tokens[0].startswith(u'i@'):
             return (lambda s: lambda n:
                         _tgrep_node_literal_value(n).lower() == s)(tokens[0][2:].lower())
         else:
@@ -234,8 +251,8 @@ def _tgrep_parens_action(_s, _l, tokens):
     '''
     # print 'parenthetical tokens: ', tokens
     assert len(tokens) == 3
-    assert tokens[0] == '('
-    assert tokens[2] == ')'
+    assert tokens[0] == u'('
+    assert tokens[2] == u')'
     return tokens[1]
 def _tgrep_nltk_tree_pos_action(_s, _l, tokens):
@@ -247,7 +264,7 @@ def _tgrep_nltk_tree_pos_action(_s, _l, tokens):
     # recover the tuple from the parsed sting
     node_tree_position = tuple(int(x) for x in tokens if x.isdigit())
     # capture the node's tree position
-    return (lambda i: lambda n: (hasattr(n, 'treeposition') and
+    return (lambda i: lambda n: (hasattr(n, u'treeposition') and
                                  n.treeposition() == i))(node_tree_position)
 def _tgrep_relation_action(_s, _l, tokens):
@@ -258,177 +275,177 @@ def _tgrep_relation_action(_s, _l, tokens):
     # print 'relation tokens: ', tokens
     # process negation first if needed
     negated = False
-    if tokens[0] == '!':
+    if tokens[0] == u'!':
         negated = True
         tokens = tokens[1:]
-    if tokens[0] == '[':
+    if tokens[0] == u'[':
         # process square-bracketed relation expressions
         assert len(tokens) == 3
-        assert tokens[2] == ']'
+        assert tokens[2] == u']'
         retval = tokens[1]
     else:
         # process operator-node relation expressions
         assert len(tokens) == 2
         operator, predicate = tokens
         # A < B       A is the parent of (immediately dominates) B.
-        if operator == '<':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        if operator == u'<':
+            retval = lambda n: (_istree(n) and
                                 any(predicate(x) for x in n))
         # A > B       A is the child of B.
-        elif operator == '>':
-            retval = lambda n: (hasattr(n, 'parent') and
+        elif operator == u'>':
+            retval = lambda n: (hasattr(n, u'parent') and
                                 bool(n.parent()) and
                                 predicate(n.parent()))
         # A <, B      Synonymous with A <1 B.
-        elif operator == '<,' or operator == '<1':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        elif operator == u'<,' or operator == u'<1':
+            retval = lambda n: (_istree(n) and
                                 bool(list(n)) and
                                 predicate(n[0]))
         # A >, B      Synonymous with A >1 B.
-        elif operator == '>,' or operator == '>1':
-            retval = lambda n: (hasattr(n, 'parent') and
+        elif operator == u'>,' or operator == u'>1':
+            retval = lambda n: (hasattr(n, u'parent') and
                                 bool(n.parent()) and
                                 (n is n.parent()[0]) and
                                 predicate(n.parent()))
         # A <N B      B is the Nth child of A (the first child is <1).
-        elif operator[0] == '<' and operator[1:].isdigit():
+        elif operator[0] == u'<' and operator[1:].isdigit():
             idx = int(operator[1:])
             # capture the index parameter
-            retval = (lambda i: lambda n: (isinstance(n, nltk.tree.Tree) and
+            retval = (lambda i: lambda n: (_istree(n) and
                                            bool(list(n)) and
                                            0 <= i < len(n) and
                                            predicate(n[i])))(idx - 1)
         # A >N B      A is the Nth child of B (the first child is >1).
-        elif operator[0] == '>' and operator[1:].isdigit():
+        elif operator[0] == u'>' and operator[1:].isdigit():
             idx = int(operator[1:])
             # capture the index parameter
-            retval = (lambda i: lambda n: (hasattr(n, 'parent') and
+            retval = (lambda i: lambda n: (hasattr(n, u'parent') and
                                            bool(n.parent()) and
                                            0 <= i < len(n.parent()) and
                                            (n is n.parent()[i]) and
                                            predicate(n.parent())))(idx - 1)
         # A <' B      B is the last child of A (also synonymous with A <-1 B).
         # A <- B      B is the last child of A (synonymous with A <-1 B).
-        elif operator == '<\'' or operator == '<-' or operator == '<-1':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and bool(list(n))
+        elif operator == u'<\'' or operator == u'<-' or operator == u'<-1':
+            retval = lambda n: (_istree(n) and bool(list(n))
                                 and predicate(n[-1]))
         # A >' B      A is the last child of B (also synonymous with A >-1 B).
         # A >- B      A is the last child of B (synonymous with A >-1 B).
-        elif operator == '>\'' or operator == '>-' or operator == '>-1':
-            retval = lambda n: (hasattr(n, 'parent') and
+        elif operator == u'>\'' or operator == u'>-' or operator == u'>-1':
+            retval = lambda n: (hasattr(n, u'parent') and
                                 bool(n.parent()) and
                                 (n is n.parent()[-1]) and
                                 predicate(n.parent()))
         # A <-N B 	  B is the N th-to-last child of A (the last child is <-1).
-        elif operator[:2] == '<-' and operator[2:].isdigit():
+        elif operator[:2] == u'<-' and operator[2:].isdigit():
             idx = -int(operator[2:])
             # capture the index parameter
-            retval = (lambda i: lambda n: (isinstance(n, nltk.tree.Tree) and
+            retval = (lambda i: lambda n: (_istree(n) and
                                            bool(list(n)) and
                                            0 <= (i + len(n)) < len(n) and
                                            predicate(n[i + len(n)])))(idx)
         # A >-N B 	  A is the N th-to-last child of B (the last child is >-1).
-        elif operator[:2] == '>-' and operator[2:].isdigit():
+        elif operator[:2] == u'>-' and operator[2:].isdigit():
             idx = -int(operator[2:])
             # capture the index parameter
             retval = (lambda i: lambda n:
-                          (hasattr(n, 'parent') and
+                          (hasattr(n, u'parent') and
                            bool(n.parent()) and
                            0 <= (i + len(n.parent())) < len(n.parent()) and
                            (n is n.parent()[i + len(n.parent())]) and
                            predicate(n.parent())))(idx)
         # A <: B      B is the only child of A
-        elif operator == '<:':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        elif operator == u'<:':
+            retval = lambda n: (_istree(n) and
                                 len(n) == 1 and
                                 predicate(n[0]))
         # A >: B      A is the only child of B.
-        elif operator == '>:':
-            retval = lambda n: (hasattr(n, 'parent') and
+        elif operator == u'>:':
+            retval = lambda n: (hasattr(n, u'parent') and
                                 bool(n.parent()) and
                                 len(n.parent()) == 1 and
                                 predicate(n.parent()))
         # A << B      A dominates B (A is an ancestor of B).
-        elif operator == '<<':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        elif operator == u'<<':
+            retval = lambda n: (_istree(n) and
                                 any(predicate(x) for x in _descendants(n)))
         # A >> B      A is dominated by B (A is a descendant of B).
-        elif operator == '>>':
+        elif operator == u'>>':
             retval = lambda n: any(predicate(x) for x in ancestors(n))
         # A <<, B     B is a left-most descendant of A.
-        elif operator == '<<,' or operator == '<<1':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        elif operator == u'<<,' or operator == u'<<1':
+            retval = lambda n: (_istree(n) and
                                 any(predicate(x)
                                     for x in _leftmost_descendants(n)))
         # A >>, B     A is a left-most descendant of B.
-        elif operator == '>>,':
+        elif operator == u'>>,':
             retval = lambda n: any((predicate(x) and
                                     n in _leftmost_descendants(x))
                                    for x in ancestors(n))
         # A <<' B     B is a right-most descendant of A.
-        elif operator == '<<\'':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        elif operator == u'<<\'':
+            retval = lambda n: (_istree(n) and
                                 any(predicate(x)
                                     for x in _rightmost_descendants(n)))
         # A >>' B     A is a right-most descendant of B.
-        elif operator == '>>':
+        elif operator == u'>>':
             retval = lambda n: any((predicate(x) and
                                     n in _rightmost_descendants(x))
                                    for x in ancestors(n))
         # A <<: B     There is a single path of descent from A and B is on it.
-        elif operator == '<<:':
-            retval = lambda n: (isinstance(n, nltk.tree.Tree) and
+        elif operator == u'<<:':
+            retval = lambda n: (_istree(n) and
                                 any(predicate(x)
                                     for x in _unique_descendants(n)))
         # A >>: B     There is a single path of descent from B and A is on it.
-        elif operator == '>>:':
+        elif operator == u'>>:':
             retval = lambda n: any(predicate(x) for x in unique_ancestors(n))
         # A . B       A immediately precedes B.
-        elif operator == '.':
+        elif operator == u'.':
             retval = lambda n: any(predicate(x)
                                    for x in _immediately_after(n))
         # A , B       A immediately follows B.
-        elif operator == ',':
+        elif operator == u',':
             retval = lambda n: any(predicate(x)
                                    for x in _immediately_before(n))
         # A .. B      A precedes B.
-        elif operator == '..':
+        elif operator == u'..':
             retval = lambda n: any(predicate(x) for x in _after(n))
         # A ,, B      A follows B.
-        elif operator == ',,':
+        elif operator == u',,':
             retval = lambda n: any(predicate(x) for x in _before(n))
         # A $ B       A is a sister of B (and A != B).
-        elif operator == '$' or operator == '%':
-            retval = lambda n: (hasattr(n, 'parent') and
+        elif operator == u'$' or operator == u'%':
+            retval = lambda n: (hasattr(n, u'parent') and
                                 bool(n.parent()) and
                                 any(predicate(x)
                                     for x in n.parent() if x is not n))
         # A $. B      A is a sister of and immediately precedes B.
-        elif operator == '$.' or operator == '%.':
-            retval = lambda n: (hasattr(n, 'right_sibling') and
+        elif operator == u'$.' or operator == u'%.':
+            retval = lambda n: (hasattr(n, u'right_sibling') and
                                 bool(n.right_sibling()) and
                                 predicate(n.right_sibling()))
         # A $, B      A is a sister of and immediately follows B.
-        elif operator == '$,' or operator == '%,':
-            retval = lambda n: (hasattr(n, 'left_sibling') and
+        elif operator == u'$,' or operator == u'%,':
+            retval = lambda n: (hasattr(n, u'left_sibling') and
                                 bool(n.left_sibling()) and
                                 predicate(n.left_sibling()))
         # A $.. B     A is a sister of and precedes B.
-        elif operator == '$..' or operator == '%..':
-            retval = lambda n: (hasattr(n, 'parent') and
-                                hasattr(n, 'parent_index') and
+        elif operator == u'$..' or operator == u'%..':
+            retval = lambda n: (hasattr(n, u'parent') and
+                                hasattr(n, u'parent_index') and
                                 bool(n.parent()) and
                                 any(predicate(x) for x in
                                     n.parent()[n.parent_index() + 1:]))
         # A $,, B     A is a sister of and follows B.
-        elif operator == '$,,' or operator == '%,,':
-            retval = lambda n: (hasattr(n, 'parent') and
-                                hasattr(n, 'parent_index') and
+        elif operator == u'$,,' or operator == u'%,,':
+            retval = lambda n: (hasattr(n, u'parent') and
+                                hasattr(n, u'parent_index') and
                                 bool(n.parent()) and
                                 any(predicate(x) for x in
                                     n.parent()[:n.parent_index()]))
         else:
-            assert False, 'cannot interpret tgrep operator "{0}"'.format(
+            assert False, u'cannot interpret tgrep operator "{0}"'.format(
                 operator)
     # now return the built function
     if negated:
@@ -442,7 +459,7 @@ def _tgrep_rel_conjunction_action(_s, _l, tokens):
     from the conjunction of several other such lambda functions.
     '''
     # filter out the ampersand
-    tokens = [x for x in tokens if x != '&']
+    tokens = [x for x in tokens if x != u'&']
     # print 'relation conjunction tokens: ', tokens
     if len(tokens) == 1:
         return tokens[0]
@@ -455,7 +472,7 @@ def _tgrep_rel_disjunction_action(_s, _l, tokens):
     from the disjunction of several other such lambda functions.
     '''
     # filter out the pipe
-    tokens = [x for x in tokens if x != '|']
+    tokens = [x for x in tokens if x != u'|']
     # print 'relation disjunction tokens: ', tokens
     if len(tokens) == 1:
         return tokens[0]
@@ -467,40 +484,40 @@ def _build_tgrep_parser(set_parse_actions = True):
     Builds a pyparsing-based parser object for tokenizing and
     interpreting tgrep search strings.
     '''
-    tgrep_op = (pyparsing.Optional('!') +
-                pyparsing.Regex('[$%,.<>][%,.<>0-9-\':]*'))
-    tgrep_qstring = pyparsing.QuotedString(quoteChar='"', escChar='\\',
+    tgrep_op = (pyparsing.Optional(u'!') +
+                pyparsing.Regex(u'[$%,.<>][%,.<>0-9-\':]*'))
+    tgrep_qstring = pyparsing.QuotedString(quoteChar=u'"', escChar=u'\\',
                                            unquoteResults=False)
-    tgrep_node_regex = pyparsing.QuotedString(quoteChar='/', escChar='\',
+    tgrep_node_regex = pyparsing.QuotedString(quoteChar=u'/', escChar=u'\',
                                               unquoteResults=False)
-    tgrep_node_literal = pyparsing.Regex('[^][ \r\t\n;:.,&|<>()$!@%^=]+')
+    tgrep_node_literal = pyparsing.Regex(u'[^][ \r\t\n;:.,&|<>()$!@%^=]+')
     tgrep_expr = pyparsing.Forward()
     tgrep_relations = pyparsing.Forward()
-    tgrep_parens = pyparsing.Literal('(') + tgrep_expr + ')'
+    tgrep_parens = pyparsing.Literal(u'(') + tgrep_expr + u')'
     tgrep_nltk_tree_pos = (
-        pyparsing.Literal('N(') +
-        pyparsing.Optional(pyparsing.Word(pyparsing.nums) + ',' +
+        pyparsing.Literal(u'N(') +
+        pyparsing.Optional(pyparsing.Word(pyparsing.nums) + u',' +
                            pyparsing.Optional(pyparsing.delimitedList(
-                    pyparsing.Word(pyparsing.nums), delim=',') +
-                                              pyparsing.Optional(','))) + ')')
+                    pyparsing.Word(pyparsing.nums), delim=u',') +
+                                              pyparsing.Optional(u','))) + u')')
     tgrep_node_expr = (tgrep_qstring |
                        tgrep_node_regex |
-                       '*' |
+                       u'*' |
                        tgrep_node_literal)
     tgrep_node = (tgrep_parens |
                   tgrep_nltk_tree_pos |
-                  (pyparsing.Optional("'") +
+                  (pyparsing.Optional(u"'") +
                    tgrep_node_expr +
-                   pyparsing.ZeroOrMore("|" + tgrep_node_expr)))
+                   pyparsing.ZeroOrMore(u"|" + tgrep_node_expr)))
     tgrep_relation = pyparsing.Forward()
-    tgrep_brackets = pyparsing.Optional('!') + '[' + tgrep_relations + ']'
+    tgrep_brackets = pyparsing.Optional(u'!') + u'[' + tgrep_relations + u']'
     tgrep_relation = tgrep_brackets | tgrep_op + tgrep_node
     tgrep_rel_conjunction = pyparsing.Forward()
     tgrep_rel_conjunction << (tgrep_relation +
-                              pyparsing.ZeroOrMore(pyparsing.Optional('&') +
+                              pyparsing.ZeroOrMore(pyparsing.Optional(u'&') +
                                                    tgrep_rel_conjunction))
     tgrep_relations << tgrep_rel_conjunction + pyparsing.ZeroOrMore(
-        "|" + tgrep_relations)
+        u"|" + tgrep_relations)
     tgrep_expr << tgrep_node + pyparsing.Optional(tgrep_relations)
     if set_parse_actions:
         tgrep_node.setParseAction(_tgrep_node_action)
@@ -520,6 +537,8 @@ def tgrep_tokenize(tgrep_string):
     Tokenizes a TGrep search string into separate tokens.
     '''
     parser = _build_tgrep_parser(False)
+    if isinstance(tgrep_string, bytes):
+        tgrep_string = tgrep_string.decode()
     return list(parser.parseString(tgrep_string))
 def tgrep_compile(tgrep_string):
@@ -528,6 +547,8 @@ def tgrep_compile(tgrep_string):
     lambda function.
     '''
     parser = _build_tgrep_parser(True)
+    if isinstance(tgrep_string, bytes):
+        tgrep_string = tgrep_string.decode()
     return list(parser.parseString(tgrep_string, parseAll=True))[0]
 def treepositions_no_leaves(tree):
@@ -552,14 +573,15 @@ def tgrep_positions(tree, tgrep_string, search_leaves = True):
     If `search_leaves` is False, the method will not return any
     results in leaf positions.
     '''
-    if not hasattr(tree, 'treepositions'):
+    try:
+        if search_leaves:
+            search_positions = tree.treepositions()
+        else:
+            search_positions = treepositions_no_leaves(tree)
+    except AttributeError:
         return []
-    if isinstance(tgrep_string, basestring):
+    if isinstance(tgrep_string, (bytes, str)):
         tgrep_string = tgrep_compile(tgrep_string)
-    if search_leaves:
-        search_positions = tree.treepositions()
-    else:
-        search_positions = treepositions_no_leaves(tree)
     return [position for position in search_positions
             if tgrep_string(tree[position])]