Version: 1.0-4 (2003-07-19, released)

        * RODBC.c: iodbc was not recognizing the SQLTCHAR cast

        * odbcClose.Rd, odbcConnect.Rd: tidy up


Version: 1.0-3 (2003-05-25, released)

        * RODBC.c: remove unneeded SQLCHAR cast


Version: 1.0-2 (2003-05-13, released)

	* Add inst/ directory and install README.

	* Allow reading of character fields up to 64K bytes (limited
	  as the driver may report silly sizes).

	* Note that convenience wrappers for Excel etc need
	  English-language drivers to be installed.


Version: 1.0-1 (2003-01-20, released)

	* If a character field is truncated when reading from a database,
	  an R warning is given.

	* There was no protection against writing more than 255 chars in a
	  character vector in sqlSave/sqlUpdate(fast=TRUE).  Now the first
	  255 characters are written and a warning given.


Version: 0.99-9 (2003-01-17, test version)

	* Testing with PostgreSQL revealed that
	  - attempting to create a table with a primary key resulted in
	    memory corruption inside the unixODBC PostgreSQL driver and
	  - the special columns in PostgreSQL are pseudo-columns, so ML's
	    stategy in sqlUpdate is doomed.

	* Added `addPK' argument to sqlSave(), defaulting to FALSE, to
	  circumvent the above.  (The latest psqlodbc client behaves.)

	* Changed strategy in sqlUpdate().  Allow `index' argument, and
	  otherwise look first for a primary key in the data frame, then
	  special columns, then "rownames".

	* Changed all C-level tests to allow for SQL_SUCCESS_WITH_INFO as
	  the return value is driver-dependent and psqlodbc was returning
	  SQL_SUCCESS_WITH_INFO from SQLExecute.


Version: 0.99-8 (2003-01-15, test version)

	* Result sets with no rows are now reported as 0-row data frames
	  and not regarded as an error (return value -1).

	* More consistent usage of errors, with new error code -2
	  indicating no data (that is no result from the query, not just
	  no rows available).  For example, failure to create a table in
	  sqlSave() is now a fatal error.

	* On Windows, convenience functions odbcConnectAccess(),
	  odbcConnectDbase() and odbcConnectExcel() based on suggestion of
	  Fan, <xiao.gang.fan1@libertysurf.fr>.

	* A `channel' object contains a randomly generated ID, so even if
	  a channel has been closed and the slot reused, the original R
	  object will be declared invalid.

	* options("dec") is only set when RODBC is loaded if it is not
	  already set.

	* New function odbcCloseAll().

	* Improved documentation.  New README, move ML's to README.orig.


Version: 0.99-6 (2003-01-12, test version)

	* Migration to use .Call where appropriate.

	* `channel' is now a classed object, with a print method and
	  all relevant information is recorded as attributes of the
	  object, and new function odbcReConnect() allows reconnection to
	  a saved `channel' object.

	* New function odbcDriverConnect() provides an alternative
	  interface via ODBC driver connection strings. Under the Windows
	  GUI this allows interactive completion of connections strings
	  via standard ODBC dialog boxes.

	* Names of tables are no longer deparsed, so normal R
	  semantics apply and they should be quoted.

	* odbcNumRows now returns the number of rows, not a list.

	* odbcCaseFlag now returns character in all cases.

	* New argument `believeNRows' to accommodate ORACLE (and also
	  sqlTypeInfo() calls on MySQL).

	* The limit on the number of channels was off by one.

	* sqlSave() and sqlFetch() now default to rownames = TRUE, that is
	  to include rownames in the table, and to use column `rownames'
	  for the rownames if it exists.  It is possible to retrieve
	  columns other than the first as rownames, as documented (but not
	  working).

	* New `safer' option to sqlSave(), which defaults to TRUE.

	* When sqlSave() creates a table and includes rownames, the rownames
	  are created as a primary key (where supported).

	* sqlUpdate() has a new `tablename' argument and many internal
	  changes (e.g. implementing case changes): it is now operational.

	* sqlSave(fast=TRUE) and sqlUpdate() can now create database NULLs
	  even on MS Access, and character strings with value "NA" are no
	  longer automatically regarded as missing.  sqlUpdate() now has a
	  fast=FALSE method, and for both the fast methods avoid
	  converting numeric and integer columns to character.

	* odbcFetchRow() has been removed.

	* sqlCopy() is now implemented.

	* odbcFetchRows() now returns a list, and sqlFetchRows() only
	  returns a data frame, this being the most efficient storage.
	  SQL double, real, integer and smallint columns are transferred
	  directly to R numeric and integer columns (rather than being
	  converted to character and back again).

	* Function sqlDrop(), sqlColumns() and sqlPrimaryKeys() were not
	  respecting case-changing (although case-altering DBMSs usually
	  coped).

	* Various memory leaks have been plugged.

	* Improved documentation.


Version: 0.9-1 (release candidate 2, released to CRAN 2002-10-30)

	* Improved documentation


Version: 0.9-0 (release candidate 1)

	* New maintainer.

	* Bring up to date for R 1.6.0.

	* Several functions were not testing the case-switched version of
	  the table name: now integrated into new function odbcTableExists.

	* New function odbcGetInfo.

	* New function sqlTypeInfo.

	* sqlFetch now does work with Excel spreadsheets.


Version: 0.8-3

Minor changes to make documentation and examples R CMD check compliant
