Upstream andialbrecht/sqlparse state of the touched files before 730ddba559

sqlparse/engine/grouping.py

@@ -309,34 +309,13 @@ def group_operator(tlist):
 
 def group_identifier_list(tlist):
     m_role = T.Keyword, ('null', 'role')
-    m_to = T.Keyword, 'TO'
     sqlcls = (sql.Function, sql.Case, sql.Identifier, sql.Comparison,
               sql.IdentifierList, sql.Operation)
     ttypes = (T_NUMERICAL + T_STRING + T_NAME
               + (T.Keyword, T.Comment, T.Wildcard))
 
-    def is_rename_separator(token):
-        # In ``RENAME TABLE a TO b, c TO d`` the comma separates independent
-        # rename clauses, so the target of one clause (``b``) and the source
-        # of the next (``c``) must not be merged into a single IdentifierList.
-        # Detect that layout by the ``TO`` keywords flanking both sides of the
-        # comma and leave such a comma ungrouped (issue779).
-        parent = token.parent
-        if parent is None:
-            return False
-        cidx = parent.token_index(token)
-        pidx, prev_ = parent.token_prev(cidx)
-        nidx, next_ = parent.token_next(cidx)
-        if prev_ is None or next_ is None:
-            return False
-        _, before_prev = parent.token_prev(pidx)
-        _, after_next = parent.token_next(nidx)
-        return (before_prev is not None and before_prev.match(*m_to)
-                and after_next is not None and after_next.match(*m_to))
-
     def match(token):
-        return token.match(T.Punctuation, ',') \
-            and not is_rename_separator(token)
+        return token.match(T.Punctuation, ',')
 
     def valid(token):
         return imt(token, i=sqlcls, m=m_role, t=ttypes)

sqlparse/filters/others.py

@@ -0,0 +1,174 @@
+#
+# Copyright (C) 2009-2020 the sqlparse authors and contributors
+# <see AUTHORS file>
+#
+# This module is part of python-sqlparse and is released under
+# the BSD License: https://opensource.org/licenses/BSD-3-Clause
+
+import re
+
+from sqlparse import sql
+from sqlparse import tokens as T
+from sqlparse.utils import split_unquoted_newlines
+
+
+class StripCommentsFilter:
+
+    @staticmethod
+    def _process(tlist):
+        def get_next_comment(idx=-1):
+            # TODO(andi) Comment types should be unified, see related issue38
+            return tlist.token_next_by(i=sql.Comment, t=T.Comment, idx=idx)
+
+        def _get_insert_token(token):
+            """Returns either a whitespace or the line breaks from token."""
+            # See issue484 why line breaks should be preserved.
+            # Note: The actual value for a line break is replaced by \n
+            # in SerializerUnicode which will be executed in the
+            # postprocessing state.
+            m = re.search(r'([\r\n]+) *$', token.value)
+            if m is not None:
+                return sql.Token(T.Whitespace.Newline, m.groups()[0])
+            else:
+                return sql.Token(T.Whitespace, ' ')
+
+        sql_hints = (T.Comment.Multiline.Hint, T.Comment.Single.Hint)
+        tidx, token = get_next_comment()
+        while token:
+            # skipping token remove if token is a SQL-Hint. issue262
+            is_sql_hint = False
+            if token.ttype in sql_hints:
+                is_sql_hint = True
+            elif isinstance(token, sql.Comment):
+                comment_tokens = token.tokens
+                if len(comment_tokens) > 0:
+                    if comment_tokens[0].ttype in sql_hints:
+                        is_sql_hint = True
+
+            if is_sql_hint:
+                # using current index as start index to search next token for
+                # preventing infinite loop in cases when token type is a
+                # "SQL-Hint" and has to be skipped
+                tidx, token = get_next_comment(idx=tidx)
+                continue
+
+            pidx, prev_ = tlist.token_prev(tidx, skip_ws=False)
+            nidx, next_ = tlist.token_next(tidx, skip_ws=False)
+            # Replace by whitespace if prev and next exist and if they're not
+            # whitespaces. This doesn't apply if prev or next is a parenthesis.
+            if (
+                prev_ is None or next_ is None
+                or prev_.is_whitespace or prev_.match(T.Punctuation, '(')
+                or next_.is_whitespace or next_.match(T.Punctuation, ')')
+            ):
+                # Insert a whitespace to ensure the following SQL produces
+                # a valid SQL (see #425).
+                if prev_ is not None and not prev_.match(T.Punctuation, '('):
+                    tlist.tokens.insert(tidx, _get_insert_token(token))
+                tlist.tokens.remove(token)
+                tidx -= 1
+            else:
+                tlist.tokens[tidx] = _get_insert_token(token)
+
+            # using current index as start index to search next token for
+            # preventing infinite loop in cases when token type is a
+            # "SQL-Hint" and has to be skipped
+            tidx, token = get_next_comment(idx=tidx)
+
+    def process(self, stmt):
+        [self.process(sgroup) for sgroup in stmt.get_sublists()]
+        StripCommentsFilter._process(stmt)
+        return stmt
+
+
+class StripWhitespaceFilter:
+    def _stripws(self, tlist):
+        func_name = f'_stripws_{type(tlist).__name__}'
+        func = getattr(self, func_name.lower(), self._stripws_default)
+        func(tlist)
+
+    @staticmethod
+    def _stripws_default(tlist):
+        last_was_ws = False
+        is_first_char = True
+        for token in tlist.tokens:
+            if token.is_whitespace:
+                token.value = '' if last_was_ws or is_first_char else ' '
+            last_was_ws = token.is_whitespace
+            is_first_char = False
+
+    def _stripws_identifierlist(self, tlist):
+        # Removes newlines before commas, see issue140
+        last_nl = None
+        for token in list(tlist.tokens):
+            if last_nl and token.ttype is T.Punctuation and token.value == ',':
+                tlist.tokens.remove(last_nl)
+            last_nl = token if token.is_whitespace else None
+
+            # next_ = tlist.token_next(token, skip_ws=False)
+            # if (next_ and not next_.is_whitespace and
+            #             token.ttype is T.Punctuation and token.value == ','):
+            #     tlist.insert_after(token, sql.Token(T.Whitespace, ' '))
+        return self._stripws_default(tlist)
+
+    def _stripws_parenthesis(self, tlist):
+        while tlist.tokens[1].is_whitespace:
+            tlist.tokens.pop(1)
+        while tlist.tokens[-2].is_whitespace:
+            tlist.tokens.pop(-2)
+        if tlist.tokens[-2].is_group:
+            # save to remove the last whitespace
+            while tlist.tokens[-2].tokens[-1].is_whitespace:
+                tlist.tokens[-2].tokens.pop(-1)
+        self._stripws_default(tlist)
+
+    def process(self, stmt, depth=0):
+        [self.process(sgroup, depth + 1) for sgroup in stmt.get_sublists()]
+        self._stripws(stmt)
+        if depth == 0 and stmt.tokens and stmt.tokens[-1].is_whitespace:
+            stmt.tokens.pop(-1)
+        return stmt
+
+
+class SpacesAroundOperatorsFilter:
+    @staticmethod
+    def _process(tlist):
+
+        ttypes = (T.Operator, T.Comparison)
+        tidx, token = tlist.token_next_by(t=ttypes)
+        while token:
+            nidx, next_ = tlist.token_next(tidx, skip_ws=False)
+            if next_ and next_.ttype != T.Whitespace:
+                tlist.insert_after(tidx, sql.Token(T.Whitespace, ' '))
+
+            pidx, prev_ = tlist.token_prev(tidx, skip_ws=False)
+            if prev_ and prev_.ttype != T.Whitespace:
+                tlist.insert_before(tidx, sql.Token(T.Whitespace, ' '))
+                tidx += 1  # has to shift since token inserted before it
+
+            # assert tlist.token_index(token) == tidx
+            tidx, token = tlist.token_next_by(t=ttypes, idx=tidx)
+
+    def process(self, stmt):
+        [self.process(sgroup) for sgroup in stmt.get_sublists()]
+        SpacesAroundOperatorsFilter._process(stmt)
+        return stmt
+
+
+class StripTrailingSemicolonFilter:
+
+    def process(self, stmt):
+        while stmt.tokens and (stmt.tokens[-1].is_whitespace
+                               or stmt.tokens[-1].value == ';'):
+            stmt.tokens.pop()
+        return stmt
+
+
+# ---------------------------
+# postprocess
+
+class SerializerUnicode:
+    @staticmethod
+    def process(stmt):
+        lines = split_unquoted_newlines(stmt)
+        return '\n'.join(line.rstrip() for line in lines)

sqlparse/sql.py

@@ -0,0 +1,664 @@
+#
+# Copyright (C) 2009-2020 the sqlparse authors and contributors
+# <see AUTHORS file>
+#
+# This module is part of python-sqlparse and is released under
+# the BSD License: https://opensource.org/licenses/BSD-3-Clause
+
+"""This module contains classes representing syntactical elements of SQL."""
+
+import re
+
+from sqlparse import tokens as T
+from sqlparse.utils import imt, remove_quotes
+
+
+class NameAliasMixin:
+    """Implements get_real_name and get_alias."""
+
+    def get_real_name(self):
+        """Returns the real name (object name) of this identifier."""
+        # a.b.c -> real name is the component after the *last* dot
+        dot_idx = None
+        for idx, tok in enumerate(self.tokens):
+            if tok.match(T.Punctuation, '.'):
+                dot_idx = idx
+        return self._get_first_name(dot_idx, real_name=True)
+
+    def get_alias(self):
+        """Returns the alias for this identifier or ``None``."""
+
+        # "name AS alias"
+        kw_idx, kw = self.token_next_by(m=(T.Keyword, 'AS'))
+        if kw is not None:
+            return self._get_first_name(kw_idx + 1, keywords=True)
+
+        # "name alias" or "complicated column expression alias"
+        _, ws = self.token_next_by(t=T.Whitespace)
+        if len(self.tokens) > 2 and ws is not None:
+            return self._get_first_name(reverse=True)
+
+
+class Token:
+    """Base class for all other classes in this module.
+
+    It represents a single token and has two instance attributes:
+    ``value`` is the unchanged value of the token and ``ttype`` is
+    the type of the token.
+    """
+
+    __slots__ = (
+        'is_group',
+        'is_keyword',
+        'is_newline',
+        'is_whitespace',
+        'normalized',
+        'parent',
+        'ttype',
+        'value',
+    )
+
+    def __init__(self, ttype, value):
+        value = str(value)
+        self.value = value
+        self.ttype = ttype
+        self.parent = None
+        self.is_group = False
+        self.is_keyword = ttype in T.Keyword
+        self.is_whitespace = self.ttype in T.Whitespace
+        self.is_newline = self.ttype in T.Newline
+        self.normalized = value.upper() if self.is_keyword else value
+
+    def __str__(self):
+        return self.value
+
+    # Pending tokenlist __len__ bug fix
+    # def __len__(self):
+    #     return len(self.value)
+
+    def __repr__(self):
+        cls = self._get_repr_name()
+        value = self._get_repr_value()
+
+        q = '"' if value.startswith("'") and value.endswith("'") else "'"
+        return "<{cls} {q}{value}{q} at 0x{id:2X}>".format(
+            id=id(self), **locals())
+
+    def _get_repr_name(self):
+        return str(self.ttype).split('.')[-1]
+
+    def _get_repr_value(self):
+        raw = str(self)
+        if len(raw) > 7:
+            raw = raw[:6] + '...'
+        return re.sub(r'\s+', ' ', raw)
+
+    def flatten(self):
+        """Resolve subgroups."""
+        yield self
+
+    def match(self, ttype, values, regex=False):
+        """Checks whether the token matches the given arguments.
+
+        *ttype* is a token type as defined in `sqlparse.tokens`. If it does
+        not match, ``False`` is returned.
+        *values* is a list of possible values for this token. For match to be
+        considered valid, the token value needs to be in this list. For tokens
+        of type ``Keyword`` the comparison is case-insensitive. For
+        convenience, a single value can be given passed as a string.
+        If *regex* is ``True``, the given values are treated as regular
+        expressions. Partial matches are allowed. Defaults to ``False``.
+        """
+        type_matched = self.ttype is ttype
+        if not type_matched or values is None:
+            return type_matched
+
+        if isinstance(values, str):
+            values = (values,)
+
+        if regex:
+            # TODO: Add test for regex with is_keyword = false
+            flag = re.IGNORECASE if self.is_keyword else 0
+            values = (re.compile(v, flag) for v in values)
+
+            return any(pattern.search(self.normalized) for pattern in values)
+
+        if self.is_keyword:
+            values = (v.upper() for v in values)
+
+        return self.normalized in values
+
+    def within(self, group_cls):
+        """Returns ``True`` if this token is within *group_cls*.
+
+        Use this method for example to check if an identifier is within
+        a function: ``t.within(sql.Function)``.
+        """
+        parent = self.parent
+        while parent:
+            if isinstance(parent, group_cls):
+                return True
+            parent = parent.parent
+        return False
+
+    def is_child_of(self, other):
+        """Returns ``True`` if this token is a direct child of *other*."""
+        return self.parent == other
+
+    def has_ancestor(self, other):
+        """Returns ``True`` if *other* is in this tokens ancestry."""
+        parent = self.parent
+        while parent:
+            if parent == other:
+                return True
+            parent = parent.parent
+        return False
+
+
+class TokenList(Token):
+    """A group of tokens.
+
+    It has an additional instance attribute ``tokens`` which holds a
+    list of child-tokens.
+    """
+
+    __slots__ = 'tokens'
+
+    def __init__(self, tokens=None):
+        self.tokens = tokens or []
+        [setattr(token, 'parent', self) for token in self.tokens]
+        super().__init__(None, ''.join(token.value for token in self.tokens))
+        self.is_group = True
+
+    def __str__(self):
+        return ''.join(token.value for token in self.flatten())
+
+    # weird bug
+    # def __len__(self):
+    #     return len(self.tokens)
+
+    def __iter__(self):
+        return iter(self.tokens)
+
+    def __getitem__(self, item):
+        return self.tokens[item]
+
+    def _get_repr_name(self):
+        return type(self).__name__
+
+    def _pprint_tree(self, max_depth=None, depth=0, f=None, _pre=''):
+        """Pretty-print the object tree."""
+        token_count = len(self.tokens)
+        for idx, token in enumerate(self.tokens):
+            cls = token._get_repr_name()
+            value = token._get_repr_value()
+
+            last = idx == (token_count - 1)
+            pre = '`- ' if last else '|- '
+
+            q = '"' if value.startswith("'") and value.endswith("'") else "'"
+            print(f"{_pre}{pre}{idx} {cls} {q}{value}{q}", file=f)
+
+            if token.is_group and (max_depth is None or depth < max_depth):
+                parent_pre = '   ' if last else '|  '
+                token._pprint_tree(max_depth, depth + 1, f, _pre + parent_pre)
+
+    def get_token_at_offset(self, offset):
+        """Returns the token that is on position offset."""
+        idx = 0
+        for token in self.flatten():
+            end = idx + len(token.value)
+            if idx <= offset < end:
+                return token
+            idx = end
+
+    def flatten(self):
+        """Generator yielding ungrouped tokens.
+
+        This method is recursively called for all child tokens.
+        """
+        for token in self.tokens:
+            if token.is_group:
+                yield from token.flatten()
+            else:
+                yield token
+
+    def get_sublists(self):
+        for token in self.tokens:
+            if token.is_group:
+                yield token
+
+    @property
+    def _groupable_tokens(self):
+        return self.tokens
+
+    def _token_matching(self, funcs, start=0, end=None, reverse=False):
+        """next token that match functions"""
+        if start is None:
+            return None
+
+        if not isinstance(funcs, (list, tuple)):
+            funcs = (funcs,)
+
+        if reverse:
+            assert end is None
+            indexes = range(start - 2, -1, -1)
+        else:
+            if end is None:
+                end = len(self.tokens)
+            indexes = range(start, end)
+        for idx in indexes:
+            token = self.tokens[idx]
+            for func in funcs:
+                if func(token):
+                    return idx, token
+        return None, None
+
+    def token_first(self, skip_ws=True, skip_cm=False):
+        """Returns the first child token.
+
+        If *skip_ws* is ``True`` (the default), whitespace
+        tokens are ignored.
+
+        if *skip_cm* is ``True`` (default: ``False``), comments are
+        ignored too.
+        """
+        # this on is inconsistent, using Comment instead of T.Comment...
+        def matcher(tk):
+            return not ((skip_ws and tk.is_whitespace)
+                        or (skip_cm and imt(tk, t=T.Comment, i=Comment)))
+        return self._token_matching(matcher)[1]
+
+    def token_next_by(self, i=None, m=None, t=None, idx=-1, end=None):
+        idx += 1
+        return self._token_matching(lambda tk: imt(tk, i, m, t), idx, end)
+
+    def token_not_matching(self, funcs, idx):
+        funcs = (funcs,) if not isinstance(funcs, (list, tuple)) else funcs
+        funcs = [lambda tk, func=func: not func(tk) for func in funcs]
+        return self._token_matching(funcs, idx)
+
+    def token_matching(self, funcs, idx):
+        return self._token_matching(funcs, idx)[1]
+
+    def token_prev(self, idx, skip_ws=True, skip_cm=False):
+        """Returns the previous token relative to *idx*.
+
+        If *skip_ws* is ``True`` (the default) whitespace tokens are ignored.
+        If *skip_cm* is ``True`` comments are ignored.
+        ``None`` is returned if there's no previous token.
+        """
+        return self.token_next(idx, skip_ws, skip_cm, _reverse=True)
+
+    # TODO: May need to re-add default value to idx
+    def token_next(self, idx, skip_ws=True, skip_cm=False, _reverse=False):
+        """Returns the next token relative to *idx*.
+
+        If *skip_ws* is ``True`` (the default) whitespace tokens are ignored.
+        If *skip_cm* is ``True`` comments are ignored.
+        ``None`` is returned if there's no next token.
+        """
+        if idx is None:
+            return None, None
+        idx += 1  # alot of code usage current pre-compensates for this
+
+        def matcher(tk):
+            return not ((skip_ws and tk.is_whitespace)
+                        or (skip_cm and imt(tk, t=T.Comment, i=Comment)))
+        return self._token_matching(matcher, idx, reverse=_reverse)
+
+    def token_index(self, token, start=0):
+        """Return list index of token."""
+        start = start if isinstance(start, int) else self.token_index(start)
+        return start + self.tokens[start:].index(token)
+
+    def group_tokens(self, grp_cls, start, end, include_end=True,
+                     extend=False):
+        """Replace tokens by an instance of *grp_cls*."""
+        start_idx = start
+        start = self.tokens[start_idx]
+
+        end_idx = end + include_end
+
+        # will be needed later for new group_clauses
+        # while skip_ws and tokens and tokens[-1].is_whitespace:
+        #     tokens = tokens[:-1]
+
+        if extend and isinstance(start, grp_cls):
+            subtokens = self.tokens[start_idx + 1:end_idx]
+
+            grp = start
+            grp.tokens.extend(subtokens)
+            del self.tokens[start_idx + 1:end_idx]
+            grp.value += ''.join(token.value for token in subtokens)
+        else:
+            subtokens = self.tokens[start_idx:end_idx]
+            grp = grp_cls(subtokens)
+            self.tokens[start_idx:end_idx] = [grp]
+            grp.parent = self
+
+        for token in subtokens:
+            token.parent = grp
+
+        return grp
+
+    def insert_before(self, where, token):
+        """Inserts *token* before *where*."""
+        if not isinstance(where, int):
+            where = self.token_index(where)
+        token.parent = self
+        self.tokens.insert(where, token)
+
+    def insert_after(self, where, token, skip_ws=True):
+        """Inserts *token* after *where*."""
+        if not isinstance(where, int):
+            where = self.token_index(where)
+        nidx, next_ = self.token_next(where, skip_ws=skip_ws)
+        token.parent = self
+        if next_ is None:
+            self.tokens.append(token)
+        else:
+            self.tokens.insert(nidx, token)
+
+    def has_alias(self):
+        """Returns ``True`` if an alias is present."""
+        return self.get_alias() is not None
+
+    def get_alias(self):
+        """Returns the alias for this identifier or ``None``."""
+        return None
+
+    def get_name(self):
+        """Returns the name of this identifier.
+
+        This is either it's alias or it's real name. The returned valued can
+        be considered as the name under which the object corresponding to
+        this identifier is known within the current statement.
+        """
+        return self.get_alias() or self.get_real_name()
+
+    def get_real_name(self):
+        """Returns the real name (object name) of this identifier."""
+        return None
+
+    def get_parent_name(self):
+        """Return name of the parent object if any.
+
+        A parent object is identified by the first occurring dot.
+        """
+        dot_idx, _ = self.token_next_by(m=(T.Punctuation, '.'))
+        _, prev_ = self.token_prev(dot_idx)
+        return remove_quotes(prev_.value) if prev_ is not None else None
+
+    def _get_first_name(self, idx=None, reverse=False, keywords=False,
+                        real_name=False):
+        """Returns the name of the first token with a name"""
+
+        tokens = self.tokens[idx:] if idx else self.tokens
+        tokens = reversed(tokens) if reverse else tokens
+        types = [T.Name, T.Wildcard, T.String.Symbol]
+
+        if keywords:
+            types.append(T.Keyword)
+
+        for token in tokens:
+            if token.ttype in types:
+                return remove_quotes(token.value)
+            elif isinstance(token, (Identifier, Function)):
+                return token.get_real_name() if real_name else token.get_name()
+
+
+class Statement(TokenList):
+    """Represents a SQL statement."""
+
+    def get_type(self):
+        """Returns the type of a statement.
+
+        The returned value is a string holding an upper-cased reprint of
+        the first DML or DDL keyword. If the first token in this group
+        isn't a DML or DDL keyword "UNKNOWN" is returned.
+
+        Whitespaces and comments at the beginning of the statement
+        are ignored.
+        """
+        token = self.token_first(skip_cm=True)
+        if token is None:
+            # An "empty" statement that either has not tokens at all
+            # or only whitespace tokens.
+            return 'UNKNOWN'
+
+        elif token.ttype in (T.Keyword.DML, T.Keyword.DDL):
+            return token.normalized
+
+        elif token.ttype == T.Keyword.CTE:
+            # The WITH keyword should be followed by either an Identifier or
+            # an IdentifierList containing the CTE definitions;  the actual
+            # DML keyword (e.g. SELECT, INSERT) will follow next.
+            tidx = self.token_index(token)
+            while tidx is not None:
+                tidx, token = self.token_next(tidx, skip_ws=True)
+                if isinstance(token, (Identifier, IdentifierList)):
+                    tidx, token = self.token_next(tidx, skip_ws=True)
+
+                    if token is not None \
+                            and token.ttype == T.Keyword.DML:
+                        return token.normalized
+
+        # Hmm, probably invalid syntax, so return unknown.
+        return 'UNKNOWN'
+
+
+class Identifier(NameAliasMixin, TokenList):
+    """Represents an identifier.
+
+    Identifiers may have aliases or typecasts.
+    """
+
+    def is_wildcard(self):
+        """Return ``True`` if this identifier contains a wildcard."""
+        _, token = self.token_next_by(t=T.Wildcard)
+        return token is not None
+
+    def get_typecast(self):
+        """Returns the typecast or ``None`` of this object as a string."""
+        midx, marker = self.token_next_by(m=(T.Punctuation, '::'))
+        nidx, next_ = self.token_next(midx, skip_ws=False)
+        return next_.value if next_ else None
+
+    def get_ordering(self):
+        """Returns the ordering or ``None`` as uppercase string."""
+        _, ordering = self.token_next_by(t=T.Keyword.Order)
+        return ordering.normalized if ordering else None
+
+    def get_array_indices(self):
+        """Returns an iterator of index token lists"""
+
+        for token in self.tokens:
+            if isinstance(token, SquareBrackets):
+                # Use [1:-1] index to discard the square brackets
+                yield token.tokens[1:-1]
+
+
+class IdentifierList(TokenList):
+    """A list of :class:`~sqlparse.sql.Identifier`\'s."""
+
+    def get_identifiers(self):
+        """Returns the identifiers.
+
+        Whitespaces and punctuations are not included in this generator.
+        """
+        for token in self.tokens:
+            if not (token.is_whitespace or token.match(T.Punctuation, ',')):
+                yield token
+
+
+class TypedLiteral(TokenList):
+    """A typed literal, such as "date '2001-09-28'" or "interval '2 hours'"."""
+    M_OPEN = [(T.Name.Builtin, None), (T.Keyword, "TIMESTAMP")]
+    M_CLOSE = T.String.Single, None
+    M_EXTEND = T.Keyword, ("DAY", "HOUR", "MINUTE", "MONTH", "SECOND", "YEAR")
+
+
+class Parenthesis(TokenList):
+    """Tokens between parenthesis."""
+    M_OPEN = T.Punctuation, '('
+    M_CLOSE = T.Punctuation, ')'
+
+    @property
+    def _groupable_tokens(self):
+        return self.tokens[1:-1]
+
+
+class SquareBrackets(TokenList):
+    """Tokens between square brackets"""
+    M_OPEN = T.Punctuation, '['
+    M_CLOSE = T.Punctuation, ']'
+
+    @property
+    def _groupable_tokens(self):
+        return self.tokens[1:-1]
+
+
+class Assignment(TokenList):
+    """An assignment like 'var := val;'"""
+
+
+class If(TokenList):
+    """An 'if' clause with possible 'else if' or 'else' parts."""
+    M_OPEN = T.Keyword, 'IF'
+    M_CLOSE = T.Keyword, 'END IF'
+
+
+class For(TokenList):
+    """A 'FOR' loop."""
+    M_OPEN = T.Keyword, ('FOR', 'FOREACH')
+    M_CLOSE = T.Keyword, 'END LOOP'
+
+
+class Comparison(TokenList):
+    """A comparison used for example in WHERE clauses."""
+
+    @property
+    def left(self):
+        return self.tokens[0]
+
+    @property
+    def right(self):
+        return self.tokens[-1]
+
+
+class Comment(TokenList):
+    """A comment."""
+
+    def is_multiline(self):
+        return self.tokens and self.tokens[0].ttype == T.Comment.Multiline
+
+
+class Where(TokenList):
+    """A WHERE clause."""
+    M_OPEN = T.Keyword, 'WHERE'
+    M_CLOSE = T.Keyword, (
+        'ORDER BY', 'GROUP BY', 'LIMIT', 'UNION', 'UNION ALL', 'EXCEPT',
+        'INTERSECT', 'HAVING', 'RETURNING', 'INTO')
+
+
+class Over(TokenList):
+    """An OVER clause."""
+    M_OPEN = T.Keyword, 'OVER'
+
+
+class Having(TokenList):
+    """A HAVING clause."""
+    M_OPEN = T.Keyword, 'HAVING'
+    M_CLOSE = T.Keyword, ('ORDER BY', 'LIMIT')
+
+
+class Case(TokenList):
+    """A CASE statement with one or more WHEN and possibly an ELSE part."""
+    M_OPEN = T.Keyword, 'CASE'
+    M_CLOSE = T.Keyword, 'END'
+
+    def get_cases(self, skip_ws=False):
+        """Returns a list of 2-tuples (condition, value).
+
+        If an ELSE exists condition is None.
+        """
+        CONDITION = 1
+        VALUE = 2
+
+        ret = []
+        mode = CONDITION
+
+        for token in self.tokens:
+            # Set mode from the current statement
+            if token.match(T.Keyword, 'CASE') or (skip_ws and token.ttype in T.Whitespace):
+                continue
+
+            elif token.match(T.Keyword, 'WHEN'):
+                ret.append(([], []))
+                mode = CONDITION
+
+            elif token.match(T.Keyword, 'THEN'):
+                mode = VALUE
+
+            elif token.match(T.Keyword, 'ELSE'):
+                ret.append((None, []))
+                mode = VALUE
+
+            elif token.match(T.Keyword, 'END'):
+                mode = None
+
+            # First condition without preceding WHEN
+            if mode and not ret:
+                ret.append(([], []))
+
+            # Append token depending of the current mode
+            if mode == CONDITION:
+                ret[-1][0].append(token)
+
+            elif mode == VALUE:
+                ret[-1][1].append(token)
+
+        # Return cases list
+        return ret
+
+
+class Function(NameAliasMixin, TokenList):
+    """A function or procedure call."""
+
+    def get_parameters(self):
+        """Return a list of parameters."""
+        parenthesis = self.token_next_by(i=Parenthesis)[1]
+        result = []
+        for token in parenthesis.tokens:
+            if isinstance(token, IdentifierList):
+                return token.get_identifiers()
+            elif imt(token, i=(Function, Identifier, TypedLiteral),
+                     t=T.Literal):
+                result.append(token)
+        return result
+
+    def get_window(self):
+        """Return the window if it exists."""
+        over_clause = self.token_next_by(i=Over)
+        if not over_clause:
+            return None
+        return over_clause[1].tokens[-1]
+
+
+class Begin(TokenList):
+    """A BEGIN/END block."""
+    M_OPEN = T.Keyword, 'BEGIN'
+    M_CLOSE = T.Keyword, 'END'
+
+
+class Operation(TokenList):
+    """Grouping of operations"""
+
+
+class Values(TokenList):
+    """Grouping of values"""
+
+
+class Command(TokenList):
+    """Grouping of CLI commands."""

tests/test_grouping.py

@@ -0,0 +1,687 @@
+import pytest
+
+import sqlparse
+from sqlparse import sql, tokens as T
+
+
+def test_grouping_parenthesis():
+    s = 'select (select (x3) x2) and (y2) bar'
+    parsed = sqlparse.parse(s)[0]
+    assert str(parsed) == s
+    assert len(parsed.tokens) == 7
+    assert isinstance(parsed.tokens[2], sql.Parenthesis)
+    assert isinstance(parsed.tokens[-1], sql.Identifier)
+    assert len(parsed.tokens[2].tokens) == 5
+    assert isinstance(parsed.tokens[2].tokens[3], sql.Identifier)
+    assert isinstance(parsed.tokens[2].tokens[3].tokens[0], sql.Parenthesis)
+    assert len(parsed.tokens[2].tokens[3].tokens) == 3
+
+
+@pytest.mark.parametrize('s', ['foo := 1;', 'foo := 1'])
+def test_grouping_assignment(s):
+    parsed = sqlparse.parse(s)[0]
+    assert len(parsed.tokens) == 1
+    assert isinstance(parsed.tokens[0], sql.Assignment)
+
+
+@pytest.mark.parametrize('s', ["x > DATE '2020-01-01'", "x > TIMESTAMP '2020-01-01 00:00:00'"])
+def test_grouping_typed_literal(s):
+    parsed = sqlparse.parse(s)[0]
+    assert isinstance(parsed[0][4], sql.TypedLiteral)
+
+
+@pytest.mark.parametrize('s, a, b', [
+    ('select a from b where c < d + e', sql.Identifier, sql.Identifier),
+    ('select a from b where c < d + interval \'1 day\'', sql.Identifier, sql.TypedLiteral),
+    ('select a from b where c < d + interval \'6\' month', sql.Identifier, sql.TypedLiteral),
+    ('select a from b where c < current_timestamp - interval \'1 day\'', sql.Token, sql.TypedLiteral),
+])
+def test_compare_expr(s, a, b):
+    parsed = sqlparse.parse(s)[0]
+    assert str(parsed) == s
+    assert isinstance(parsed.tokens[2], sql.Identifier)
+    assert isinstance(parsed.tokens[6], sql.Identifier)
+    assert isinstance(parsed.tokens[8], sql.Where)
+    assert len(parsed.tokens) == 9
+    where = parsed.tokens[8]
+    assert isinstance(where.tokens[2], sql.Comparison)
+    assert len(where.tokens) == 3
+    comparison = where.tokens[2]
+    assert isinstance(comparison.tokens[0], sql.Identifier)
+    assert comparison.tokens[2].ttype is T.Operator.Comparison
+    assert isinstance(comparison.tokens[4], sql.Operation)
+    assert len(comparison.tokens) == 5
+    operation = comparison.tokens[4]
+    assert isinstance(operation.tokens[0], a)
+    assert operation.tokens[2].ttype is T.Operator
+    assert isinstance(operation.tokens[4], b)
+    assert len(operation.tokens) == 5
+
+
+def test_grouping_identifiers():
+    s = 'select foo.bar from "myscheme"."table" where fail. order'
+    parsed = sqlparse.parse(s)[0]
+    assert str(parsed) == s
+    assert isinstance(parsed.tokens[2], sql.Identifier)
+    assert isinstance(parsed.tokens[6], sql.Identifier)
+    assert isinstance(parsed.tokens[8], sql.Where)
+    s = 'select * from foo where foo.id = 1'
+    parsed = sqlparse.parse(s)[0]
+    assert str(parsed) == s
+    assert isinstance(parsed.tokens[-1].tokens[-1].tokens[0], sql.Identifier)
+    s = 'select * from (select "foo"."id" from foo)'
+    parsed = sqlparse.parse(s)[0]
+    assert str(parsed) == s
+    assert isinstance(parsed.tokens[-1].tokens[3], sql.Identifier)
+
+    for s in ["INSERT INTO `test` VALUES('foo', 'bar');",
+              "INSERT INTO `test` VALUES(1, 2), (3, 4), (5, 6);",
+              "INSERT INTO `test(a, b)` VALUES(1, 2), (3, 4), (5, 6);"]:
+        parsed = sqlparse.parse(s)[0]
+        types = [l.ttype for l in parsed.tokens if not l.is_whitespace]
+        assert types == [T.DML, T.Keyword, None, None, T.Punctuation]
+        assert isinstance(parsed.tokens[6], sql.Values)
+
+    s = "select 1.0*(a+b) as col, sum(c)/sum(d) from myschema.mytable"
+    parsed = sqlparse.parse(s)[0]
+    assert len(parsed.tokens) == 7
+    assert isinstance(parsed.tokens[2], sql.IdentifierList)
+    assert len(parsed.tokens[2].tokens) == 4
+    identifiers = list(parsed.tokens[2].get_identifiers())
+    assert len(identifiers) == 2
+    assert identifiers[0].get_alias() == "col"
+
+
+@pytest.mark.parametrize('s', [
+    '1 as f',
+    'foo as f',
+    'foo f',
+    '1/2 as f',
+    '1/2 f',
+    '1<2 as f',  # issue327
+    '1<2 f',
+])
+def test_simple_identifiers(s):
+    parsed = sqlparse.parse(s)[0]
+    assert isinstance(parsed.tokens[0], sql.Identifier)
+
+
+@pytest.mark.parametrize('s', [
+    'foo, bar',
+    'sum(a), sum(b)',
+    'sum(a) as x, b as y',
+    'sum(a)::integer, b',
+    'sum(a)/count(b) as x, y',
+    'sum(a)::integer as x, y',
+    'sum(a)::integer/count(b) as x, y',  # issue297
+])
+def test_group_identifier_list(s):
+    parsed = sqlparse.parse(s)[0]
+    assert isinstance(parsed.tokens[0], sql.IdentifierList)
+
+
+def test_grouping_identifier_wildcard():
+    p = sqlparse.parse('a.*, b.id')[0]
+    assert isinstance(p.tokens[0], sql.IdentifierList)
+    assert isinstance(p.tokens[0].tokens[0], sql.Identifier)
+    assert isinstance(p.tokens[0].tokens[-1], sql.Identifier)
+
+
+def test_grouping_identifier_name_wildcard():
+    p = sqlparse.parse('a.*')[0]
+    t = p.tokens[0]
+    assert t.get_name() == '*'
+    assert t.is_wildcard() is True
+
+
+def test_grouping_identifier_invalid():
+    p = sqlparse.parse('a.')[0]
+    assert isinstance(p.tokens[0], sql.Identifier)
+    assert p.tokens[0].has_alias() is False
+    assert p.tokens[0].get_name() is None
+    assert p.tokens[0].get_real_name() is None
+    assert p.tokens[0].get_parent_name() == 'a'
+
+
+def test_grouping_identifier_invalid_in_middle():
+    # issue261
+    s = 'SELECT foo. FROM foo'
+    p = sqlparse.parse(s)[0]
+    assert isinstance(p[2], sql.Identifier)
+    assert p[2][1].ttype == T.Punctuation
+    assert p[3].ttype == T.Whitespace
+    assert str(p[2]) == 'foo.'
+
+@pytest.mark.parametrize('s', ['foo as (select *)', 'foo as(select *)'])
+def test_grouping_identifer_as(s):
+    # issue507
+    p = sqlparse.parse(s)[0]
+    assert isinstance(p.tokens[0], sql.Identifier)
+    token = p.tokens[0].tokens[2]
+    assert token.ttype == T.Keyword
+    assert token.normalized == 'AS'
+
+def test_grouping_identifier_as_invalid():
+    # issue8
+    p = sqlparse.parse('foo as select *')[0]
+    assert len(p.tokens), 5
+    assert isinstance(p.tokens[0], sql.Identifier)
+    assert len(p.tokens[0].tokens) == 1
+    assert p.tokens[2].ttype == T.Keyword
+
+
+def test_grouping_identifier_function():
+    p = sqlparse.parse('foo() as bar')[0]
+    assert isinstance(p.tokens[0], sql.Identifier)
+    assert isinstance(p.tokens[0].tokens[0], sql.Function)
+    p = sqlparse.parse('foo()||col2 bar')[0]
+    assert isinstance(p.tokens[0], sql.Identifier)
+    assert isinstance(p.tokens[0].tokens[0], sql.Operation)
+    assert isinstance(p.tokens[0].tokens[0].tokens[0], sql.Function)
+    p = sqlparse.parse('foo(c1) over win1 as bar')[0]
+    assert isinstance(p.tokens[0], sql.Identifier)
+    assert isinstance(p.tokens[0].tokens[0], sql.Function)
+    assert len(p.tokens[0].tokens[0].tokens) == 4
+    assert isinstance(p.tokens[0].tokens[0].tokens[3], sql.Over)
+    assert isinstance(p.tokens[0].tokens[0].tokens[3].tokens[2],
+                      sql.Identifier)
+    p = sqlparse.parse('foo(c1) over (partition by c2 order by c3) as bar')[0]
+    assert isinstance(p.tokens[0], sql.Identifier)
+    assert isinstance(p.tokens[0].tokens[0], sql.Function)
+    assert len(p.tokens[0].tokens[0].tokens) == 4
+    assert isinstance(p.tokens[0].tokens[0].tokens[3], sql.Over)
+    assert isinstance(p.tokens[0].tokens[0].tokens[3].tokens[2],
+                      sql.Parenthesis)
+
+
+@pytest.mark.parametrize('s', ['foo+100', 'foo + 100', 'foo*100'])
+def test_grouping_operation(s):
+    p = sqlparse.parse(s)[0]
+    assert isinstance(p.tokens[0], sql.Operation)
+
+
+def test_grouping_identifier_list():
+    p = sqlparse.parse('a, b, c')[0]
+    assert isinstance(p.tokens[0], sql.IdentifierList)
+    p = sqlparse.parse('(a, b, c)')[0]
+    assert isinstance(p.tokens[0].tokens[1], sql.IdentifierList)
+
+
+def test_grouping_identifier_list_subquery():
+    """identifier lists should still work in subqueries with aliases"""
+    p = sqlparse.parse("select * from ("
+                       "select a, b + c as d from table) sub")[0]
+    subquery = p.tokens[-1].tokens[0]
+    idx, iden_list = subquery.token_next_by(i=sql.IdentifierList)
+    assert iden_list is not None
+    # all the identifiers should be within the IdentifierList
+    _, ilist = subquery.token_next_by(i=sql.Identifier, idx=idx)
+    assert ilist is None
+
+
+def test_grouping_identifier_list_case():
+    p = sqlparse.parse('a, case when 1 then 2 else 3 end as b, c')[0]
+    assert isinstance(p.tokens[0], sql.IdentifierList)
+    p = sqlparse.parse('(a, case when 1 then 2 else 3 end as b, c)')[0]
+    assert isinstance(p.tokens[0].tokens[1], sql.IdentifierList)
+
+
+def test_grouping_identifier_list_other():
+    # issue2
+    p = sqlparse.parse("select *, null, 1, 'foo', bar from mytable, x")[0]
+    assert isinstance(p.tokens[2], sql.IdentifierList)
+    assert len(p.tokens[2].tokens) == 13
+
+
+def test_grouping_identifier_list_with_inline_comments():
+    # issue163
+    p = sqlparse.parse('foo /* a comment */, bar')[0]
+    assert isinstance(p.tokens[0], sql.IdentifierList)
+    assert isinstance(p.tokens[0].tokens[0], sql.Identifier)
+    assert isinstance(p.tokens[0].tokens[3], sql.Identifier)
+
+
+def test_grouping_identifiers_with_operators():
+    p = sqlparse.parse('a+b as c from table where (d-e)%2= 1')[0]
+    assert len([x for x in p.flatten() if x.ttype == T.Name]) == 5
+
+
+def test_grouping_identifier_list_with_order():
+    # issue101
+    p = sqlparse.parse('1, 2 desc, 3')[0]
+    assert isinstance(p.tokens[0], sql.IdentifierList)
+    assert isinstance(p.tokens[0].tokens[3], sql.Identifier)
+    assert str(p.tokens[0].tokens[3]) == '2 desc'
+
+
+def test_grouping_nested_identifier_with_order():
+    # issue745
+    p = sqlparse.parse('(a desc)')[0]
+    assert isinstance(p.tokens[0], sql.Parenthesis)
+    assert isinstance(p.tokens[0].tokens[1], sql.Identifier)
+    assert str(p.tokens[0].tokens[1]) == 'a desc'
+
+
+def test_grouping_where():
+    s = 'select * from foo where bar = 1 order by id desc'
+    p = sqlparse.parse(s)[0]
+    assert str(p) == s
+    assert len(p.tokens) == 12
+
+    s = 'select x from (select y from foo where bar = 1) z'
+    p = sqlparse.parse(s)[0]
+    assert str(p) == s
+    assert isinstance(p.tokens[-1].tokens[0].tokens[-2], sql.Where)
+
+
+@pytest.mark.parametrize('s', (
+    'select 1 where 1 = 2 union select 2',
+    'select 1 where 1 = 2 union all select 2',
+))
+def test_grouping_where_union(s):
+    p = sqlparse.parse(s)[0]
+    assert p.tokens[5].value.startswith('union')
+
+
+def test_returning_kw_ends_where_clause():
+    s = 'delete from foo where x > y returning z'
+    p = sqlparse.parse(s)[0]
+    assert isinstance(p.tokens[6], sql.Where)
+    assert p.tokens[7].ttype == T.Keyword
+    assert p.tokens[7].value == 'returning'
+
+
+def test_into_kw_ends_where_clause():  # issue324
+    s = 'select * from foo where a = 1 into baz'
+    p = sqlparse.parse(s)[0]
+    assert isinstance(p.tokens[8], sql.Where)
+    assert p.tokens[9].ttype == T.Keyword
+    assert p.tokens[9].value == 'into'
+
+
+@pytest.mark.parametrize('sql, expected', [
+    # note: typecast needs to be 2nd token for this test
+    ('select foo::integer from bar', 'integer'),
+    ('select (current_database())::information_schema.sql_identifier',
+     'information_schema.sql_identifier'),
+])
+def test_grouping_typecast(sql, expected):
+    p = sqlparse.parse(sql)[0]
+    assert p.tokens[2].get_typecast() == expected
+
+
+def test_grouping_alias():
+    s = 'select foo as bar from mytable'
+    p = sqlparse.parse(s)[0]
+    assert str(p) == s
+    assert p.tokens[2].get_real_name() == 'foo'
+    assert p.tokens[2].get_alias() == 'bar'
+    s = 'select foo from mytable t1'
+    p = sqlparse.parse(s)[0]
+    assert str(p) == s
+    assert p.tokens[6].get_real_name() == 'mytable'
+    assert p.tokens[6].get_alias() == 't1'
+    s = 'select foo::integer as bar from mytable'
+    p = sqlparse.parse(s)[0]
+    assert str(p) == s
+    assert p.tokens[2].get_alias() == 'bar'
+    s = ('SELECT DISTINCT '
+         '(current_database())::information_schema.sql_identifier AS view')
+    p = sqlparse.parse(s)[0]
+    assert str(p) == s
+    assert p.tokens[4].get_alias() == 'view'
+
+
+def test_grouping_alias_case():
+    # see issue46
+    p = sqlparse.parse('CASE WHEN 1 THEN 2 ELSE 3 END foo')[0]
+    assert len(p.tokens) == 1
+    assert p.tokens[0].get_alias() == 'foo'
+
+
+def test_grouping_alias_ctas():
+    p = sqlparse.parse('CREATE TABLE tbl1 AS SELECT coalesce(t1.col1, 0) AS col1 FROM t1')[0]
+    assert p.tokens[10].get_alias() == 'col1'
+    assert isinstance(p.tokens[10].tokens[0], sql.Function)
+
+
+def test_grouping_alias_ctas_lowercase_as():
+    # 'as' is case-insensitive; a lowercase keyword in CREATE TABLE ... AS
+    # SELECT must not disable function grouping in the SELECT body.
+    p = sqlparse.parse('create table tbl1 as select coalesce(t1.col1, 0) as col1 from t1')[0]
+    assert p.tokens[10].get_alias() == 'col1'
+    assert isinstance(p.tokens[10].tokens[0], sql.Function)
+
+def test_grouping_subquery_no_parens():
+    # Not totally sure if this is the right approach...
+    # When a THEN clause contains a subquery w/o parenthesis around it *and*
+    # a WHERE condition, the WHERE grouper consumes END too.
+    # This takes makes sure that it doesn't fail.
+    p = sqlparse.parse('CASE WHEN 1 THEN select 2 where foo = 1 end')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Case)
+
+
+@pytest.mark.parametrize('s', ['foo.bar', 'x, y', 'x > y', 'x / y'])
+def test_grouping_alias_returns_none(s):
+    # see issue185 and issue445
+    p = sqlparse.parse(s)[0]
+    assert len(p.tokens) == 1
+    assert p.tokens[0].get_alias() is None
+
+
+def test_grouping_idlist_function():
+    # see issue10 too
+    p = sqlparse.parse('foo(1) x, bar')[0]
+    assert isinstance(p.tokens[0], sql.IdentifierList)
+
+
+def test_grouping_comparison_exclude():
+    # make sure operators are not handled too lazy
+    p = sqlparse.parse('(=)')[0]
+    assert isinstance(p.tokens[0], sql.Parenthesis)
+    assert not isinstance(p.tokens[0].tokens[1], sql.Comparison)
+    p = sqlparse.parse('(a=1)')[0]
+    assert isinstance(p.tokens[0].tokens[1], sql.Comparison)
+    p = sqlparse.parse('(a>=1)')[0]
+    assert isinstance(p.tokens[0].tokens[1], sql.Comparison)
+
+
+def test_grouping_function():
+    p = sqlparse.parse('foo()')[0]
+    assert isinstance(p.tokens[0], sql.Function)
+    p = sqlparse.parse('foo(null, bar)')[0]
+    assert isinstance(p.tokens[0], sql.Function)
+    assert len(list(p.tokens[0].get_parameters())) == 2
+    p = sqlparse.parse('foo(5) over win1')[0]
+    assert isinstance(p.tokens[0], sql.Function)
+    assert len(list(p.tokens[0].get_parameters())) == 1
+    assert isinstance(p.tokens[0].get_window(), sql.Identifier)
+    p = sqlparse.parse('foo(5) over (PARTITION BY c1)')[0]
+    assert isinstance(p.tokens[0], sql.Function)
+    assert len(list(p.tokens[0].get_parameters())) == 1
+    assert isinstance(p.tokens[0].get_window(), sql.Parenthesis)
+
+
+def test_grouping_function_not_in():
+    # issue183
+    p = sqlparse.parse('in(1, 2)')[0]
+    assert len(p.tokens) == 2
+    assert p.tokens[0].ttype == T.Keyword
+    assert isinstance(p.tokens[1], sql.Parenthesis)
+
+
+def test_grouping_varchar():
+    p = sqlparse.parse('"text" Varchar(50) NOT NULL')[0]
+    assert isinstance(p.tokens[2], sql.Function)
+
+
+def test_statement_get_type():
+    def f(sql):
+        return sqlparse.parse(sql)[0]
+
+    assert f('select * from foo').get_type() == 'SELECT'
+    assert f('update foo').get_type() == 'UPDATE'
+    assert f(' update foo').get_type() == 'UPDATE'
+    assert f('\nupdate foo').get_type() == 'UPDATE'
+    assert f('foo').get_type() == 'UNKNOWN'
+
+
+def test_identifier_with_operators():
+    # issue 53
+    p = sqlparse.parse('foo||bar')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Operation)
+    # again with whitespaces
+    p = sqlparse.parse('foo || bar')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Operation)
+
+
+def test_identifier_with_op_trailing_ws():
+    # make sure trailing whitespace isn't grouped with identifier
+    p = sqlparse.parse('foo || bar ')[0]
+    assert len(p.tokens) == 2
+    assert isinstance(p.tokens[0], sql.Operation)
+    assert p.tokens[1].ttype is T.Whitespace
+
+
+def test_identifier_with_string_literals():
+    p = sqlparse.parse("foo + 'bar'")[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Operation)
+
+
+# This test seems to be wrong. It was introduced when fixing #53, but #111
+# showed that this shouldn't be an identifier at all. I'm leaving this
+# commented in the source for a while.
+# def test_identifier_string_concat():
+#     p = sqlparse.parse("'foo' || bar")[0]
+#     assert len(p.tokens) == 1
+#     assert isinstance(p.tokens[0], sql.Identifier)
+
+
+def test_identifier_consumes_ordering():
+    # issue89
+    p = sqlparse.parse('select * from foo order by c1 desc, c2, c3')[0]
+    assert isinstance(p.tokens[-1], sql.IdentifierList)
+    ids = list(p.tokens[-1].get_identifiers())
+    assert len(ids) == 3
+    assert ids[0].get_name() == 'c1'
+    assert ids[0].get_ordering() == 'DESC'
+    assert ids[1].get_name() == 'c2'
+    assert ids[1].get_ordering() is None
+
+
+def test_comparison_with_keywords():
+    # issue90
+    # in fact these are assignments, but for now we don't distinguish them
+    p = sqlparse.parse('foo = NULL')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    assert len(p.tokens[0].tokens) == 5
+    assert p.tokens[0].left.value == 'foo'
+    assert p.tokens[0].right.value == 'NULL'
+    # make sure it's case-insensitive
+    p = sqlparse.parse('foo = null')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+
+
+def test_comparison_with_floats():
+    # issue145
+    p = sqlparse.parse('foo = 25.5')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    assert len(p.tokens[0].tokens) == 5
+    assert p.tokens[0].left.value == 'foo'
+    assert p.tokens[0].right.value == '25.5'
+
+
+def test_comparison_with_parenthesis():
+    # issue23
+    p = sqlparse.parse('(3 + 4) = 7')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    comp = p.tokens[0]
+    assert isinstance(comp.left, sql.Parenthesis)
+    assert comp.right.ttype is T.Number.Integer
+
+
+@pytest.mark.parametrize('operator', (
+    '=', '!=', '>', '<', '<=', '>=', '~', '~~', '!~~',
+    'LIKE', 'NOT LIKE', 'ILIKE', 'NOT ILIKE',
+))
+def test_comparison_with_strings(operator):
+    # issue148
+    p = sqlparse.parse(f"foo {operator} 'bar'")[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    assert p.tokens[0].right.value == "'bar'"
+    assert p.tokens[0].right.ttype == T.String.Single
+
+
+def test_like_and_ilike_comparison():
+    def validate_where_clause(where_clause, expected_tokens):
+        assert len(where_clause.tokens) == len(expected_tokens)
+        for where_token, expected_token in zip(where_clause, expected_tokens):
+            expected_ttype, expected_value = expected_token
+            if where_token.ttype is not None:
+                assert where_token.match(expected_ttype, expected_value, regex=True)
+            else:
+                # Certain tokens, such as comparison tokens, do not define a ttype that can be
+                # matched against. For these tokens, we ensure that the token instance is of
+                # the expected type and has a value conforming to specified regular expression
+                import re
+                assert (isinstance(where_token, expected_ttype)
+                        and re.match(expected_value, where_token.value))
+
+    [p1] = sqlparse.parse("select * from mytable where mytable.mycolumn LIKE 'expr%' limit 5;")
+    [p1_where] = [token for token in p1 if isinstance(token, sql.Where)]
+    validate_where_clause(p1_where, [
+        (T.Keyword, "where"),
+        (T.Whitespace, None),
+        (sql.Comparison, r"mytable.mycolumn LIKE.*"),
+        (T.Whitespace, None),
+    ])
+
+    [p2] = sqlparse.parse(
+        "select * from mytable where mycolumn NOT ILIKE '-expr' group by othercolumn;")
+    [p2_where] = [token for token in p2 if isinstance(token, sql.Where)]
+    validate_where_clause(p2_where, [
+        (T.Keyword, "where"),
+        (T.Whitespace, None),
+        (sql.Comparison, r"mycolumn NOT ILIKE.*"),
+        (T.Whitespace, None),
+    ])
+
+
+def test_comparison_with_functions():
+    # issue230
+    p = sqlparse.parse('foo = DATE(bar.baz)')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    assert len(p.tokens[0].tokens) == 5
+    assert p.tokens[0].left.value == 'foo'
+    assert p.tokens[0].right.value == 'DATE(bar.baz)'
+
+    p = sqlparse.parse('DATE(foo.bar) = DATE(bar.baz)')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    assert len(p.tokens[0].tokens) == 5
+    assert p.tokens[0].left.value == 'DATE(foo.bar)'
+    assert p.tokens[0].right.value == 'DATE(bar.baz)'
+
+    p = sqlparse.parse('DATE(foo.bar) = bar.baz')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Comparison)
+    assert len(p.tokens[0].tokens) == 5
+    assert p.tokens[0].left.value == 'DATE(foo.bar)'
+    assert p.tokens[0].right.value == 'bar.baz'
+
+
+def test_comparison_with_typed_literal():
+    p = sqlparse.parse("foo = DATE 'bar.baz'")[0]
+    assert len(p.tokens) == 1
+    comp = p.tokens[0]
+    assert isinstance(comp, sql.Comparison)
+    assert len(comp.tokens) == 5
+    assert comp.left.value == 'foo'
+    assert isinstance(comp.right, sql.TypedLiteral)
+    assert comp.right.value == "DATE 'bar.baz'"
+
+
+@pytest.mark.parametrize('start', ['FOR', 'FOREACH'])
+def test_forloops(start):
+    p = sqlparse.parse(f'{start} foo in bar LOOP foobar END LOOP')[0]
+    assert (len(p.tokens)) == 1
+    assert isinstance(p.tokens[0], sql.For)
+
+
+def test_nested_for():
+    p = sqlparse.parse('FOR foo LOOP FOR bar LOOP END LOOP END LOOP')[0]
+    assert len(p.tokens) == 1
+    for1 = p.tokens[0]
+    assert for1.tokens[0].value == 'FOR'
+    assert for1.tokens[-1].value == 'END LOOP'
+    for2 = for1.tokens[6]
+    assert isinstance(for2, sql.For)
+    assert for2.tokens[0].value == 'FOR'
+    assert for2.tokens[-1].value == 'END LOOP'
+
+
+def test_begin():
+    p = sqlparse.parse('BEGIN foo END')[0]
+    assert len(p.tokens) == 1
+    assert isinstance(p.tokens[0], sql.Begin)
+
+
+def test_keyword_followed_by_parenthesis():
+    p = sqlparse.parse('USING(somecol')[0]
+    assert len(p.tokens) == 3
+    assert p.tokens[0].ttype == T.Keyword
+    assert p.tokens[1].ttype == T.Punctuation
+
+
+def test_nested_begin():
+    p = sqlparse.parse('BEGIN foo BEGIN bar END END')[0]
+    assert len(p.tokens) == 1
+    outer = p.tokens[0]
+    assert outer.tokens[0].value == 'BEGIN'
+    assert outer.tokens[-1].value == 'END'
+    inner = outer.tokens[4]
+    assert inner.tokens[0].value == 'BEGIN'
+    assert inner.tokens[-1].value == 'END'
+    assert isinstance(inner, sql.Begin)
+
+
+def test_aliased_column_without_as():
+    p = sqlparse.parse('foo bar')[0].tokens
+    assert len(p) == 1
+    assert p[0].get_real_name() == 'foo'
+    assert p[0].get_alias() == 'bar'
+
+    p = sqlparse.parse('foo.bar baz')[0].tokens[0]
+    assert p.get_parent_name() == 'foo'
+    assert p.get_real_name() == 'bar'
+    assert p.get_alias() == 'baz'
+
+
+def test_qualified_function():
+    p = sqlparse.parse('foo()')[0].tokens[0]
+    assert p.get_parent_name() is None
+    assert p.get_real_name() == 'foo'
+
+    p = sqlparse.parse('foo.bar()')[0].tokens[0]
+    assert p.get_parent_name() == 'foo'
+    assert p.get_real_name() == 'bar'
+
+
+def test_aliased_function_without_as():
+    p = sqlparse.parse('foo() bar')[0].tokens[0]
+    assert p.get_parent_name() is None
+    assert p.get_real_name() == 'foo'
+    assert p.get_alias() == 'bar'
+
+    p = sqlparse.parse('foo.bar() baz')[0].tokens[0]
+    assert p.get_parent_name() == 'foo'
+    assert p.get_real_name() == 'bar'
+    assert p.get_alias() == 'baz'
+
+
+def test_aliased_literal_without_as():
+    p = sqlparse.parse('1 foo')[0].tokens
+    assert len(p) == 1
+    assert p[0].get_alias() == 'foo'
+
+
+def test_grouping_as_cte():
+    p = sqlparse.parse('foo AS WITH apple AS 1, banana AS 2')[0].tokens
+    assert len(p) > 4
+    assert p[0].get_alias() is None
+    assert p[2].value == 'AS'
+    assert p[4].value == 'WITH'
+
+def test_grouping_create_table():
+    p = sqlparse.parse("create table db.tbl (a string)")[0].tokens
+    assert p[4].value == "db.tbl"