GROUP BY, and HAVING clauses, if they exist.
…+call-with functions. This patch makes no functional changes however it is now possible to trace, using plain old cl:trace, the transaction/savepoint stacking macros.
…e the transaction binding.
When computing the column names to use for insert/update methods on dao instance consider the class's column-map and not just the slots' names. When building up s-sql forms in the dao readers/writers make sure to use :raw for the column names (which we have as strings and not symbols)
Willem-Alexander has been the sovereign since the 30th of April 2013, and according to Wikipedia the population is now at about 16800000.
…ransaction-level* correctly reflect the state of the transaction (that is, closed) at the time that any abort or commit hooks run. Also add abort-logical-transaction and commit-logical-transaction methods (and documentation) for use with with-logical-transaction, where the user will not know whether the handle represents a savepoint or transaction.
…te empty arrays.
It can be customized to serialize lisp values directly into the socket stream either as string or as bytes. Defaults to TO-SQL-STRING. E.g. the integration with local-time can use it to emit the binary timestamp format of PostgreSQL and avoid printing and parsing of human readable timestamp strings.
… with cl-postgres:to-sql-string" This reverts commit 80f42e9.
Also refactored so that the macro expands into a function call (call-with-ensured-transaction) so that there's a function that can be traced (and that appears on the call stack) and so that there's less back quoting.
… and abort-hooks
…h starts with #\/ connect to a unix socket in that diretory. To simplify some of the #+ccl and #+sbcl and avoid some code duplication usocket is always used when connecting to an inet socket. The change only affects those connections whose host is a string or a pathname, :unix as a host still works as before (and depends on the conection time value of *unix-socket-dir*).
…te entries. The PostgreSQL protocol defines error fields with some precision, as can be read at http://www.postgresql.org/docs/current/static/protocol-error-fields.html. This patch improves the database-error condition to expose important fields separately and fixes get-error so that news fields are properly filled when an error occurs.
It apparently does not support that.