Prior to this change a value of `'{}'` would split into the list `['']`.

is random; therefore it may be called against the subclass mapper first, so
need to check .concrete on both sides, references #3630

when the construct contains non-standard sql elements such as
returning, array index operations, or dialect-specific or custom
datatypes.  a string is now returned in these cases rendering an
approximation of the construct (typically the postgresql-style
version of it) rather than raising an error. fixes #3631
- add within_group to top-level imports
- add eq_ignore_whitespace to sqlalchemy.testing imports

a base class and a concrete-inherited subclass would raise an error
if those relationships were set up using "backref", while setting up the
identical configuration using relationship() instead with the conflicting
names would succeed, as is allowed in the case of a concrete mapping.
fixes #3630

the :class:`.Engine` to which the :class:`.Session` is bound, when
generating the string form of the SQL, so that the actual SQL
that would be emitted to the database is shown, if possible.  Previously,
only the engine associated with the :class:`.MetaData` to which the
mappings are associated would be used, if present.  If
no bind can be located either on the :class:`.Session` or on
the :class:`.MetaData` to which the mappings are associated, then
the "default" dialect is used to render the SQL, as was the case
previously.  fixes #3081

the exception to itself as the "cause"; while the Python 3 interpreter
is OK with this, it could cause endless loops in iPython.
fixes #3625
- add tests for reraise, raise_from_cause
- raise_from_cause is the same on py2k/3k, use just one function

method, and its interaction with result-row processing, now allows
the columns passed to the method to be positionally matched with the
result columns in the statement, rather than matching on name alone.
The advantage to this includes that when linking a textual SQL statement
to an ORM or Core table model, no system of labeling or de-duping of
common column names needs to occur, which also means there's no need
to worry about how label names match to ORM columns and so-forth.  In
addition, the :class:`.ResultProxy` has been further enhanced to
map column and string keys to a row with greater precision in some
cases.  fixes #3501
- reorganize the initialization of ResultMetaData for readability
and complexity; use the name "cursor_description", define the
task of "merging" cursor_description with compiled column information
as its own function, and also define "name extraction" as a separate task.
- fully change the name we use in the "ambiguous column" error to be the
actual name that was ambiguous, modify the C ext also

for SQL fragments; we've deprecated that in 1.0 and its a good idea
to document like this for 0.9 as well.