changes2.txt

功能强大的 DBF 数据库操作 dll,可以让 VB 和 POWERBASIC 调用

Changes in Cheetah

2.5.4 August 12, 2006
- Corrected incorrect record count being stored to a database when a duplicate index key error occurs.


2.5.3 April 15, 2006
- Corrected xdbEOF, xdbBOF issue when multiuser is enabled.
- Added code in %DLL_PROCESS_DETACH to attempt to delete any orphaned locks in multiuser mode.


2.5.1 March 26, 2006

- Corrected multiuser issue whereby index headers may not be refreshed.


2.5.0 March 8, 2006

- Totally re-wrote the multiuser routines. Cheetah no longer uses a lock file (*.lck). 
Instead, Cheetah now creates temporary lock files that will delete themselves when the
lock is lifted. The actual syntax for the multiuser functions has not changed.


2.2.3 February 23, 2006

- Corrected query problem whereby index used UCASE but the database values were lower case.


2.2.0 April 20, 2005

- Corrected xdbMove GPF when multiuser was enabled.

- Corrected binary field types not being reset to zero during xdbClearBuffer.


2.1.9 July 2, 2004

- Maximum number of query conditions was inadvertantly being limited to 6 instead of 50.

- Fixed bug that caused xdbPack to fail on encrypted databases.

- Corrected xdbSkipDeleted so that it correctly reports result on a database that only
contains all deleted records.  

- Added support for programming languages that only handle Null Terminated Strings. Every
Cheetah function now has a second version of the calling syntax. The send version of the
syntax has "_Z" appended to the end of the function name. This denotes that the function
is specially designed to accept Null Terminated Strings instead of the standard OLE Dynamic
Strings.                                    

- Corrected the Help for "xdbSetEditLock". Previous Help did not indicate the RecordNumber
parameter.

- Added the xdbIsEditLock function.

- Greatly increased the speed of data retrieval when using indexes to read the database in
multi-user mode.  

- Fixed bug that would cause a small memory leak whenever an index was repeatedly opened
a closed many times.  

- Added new "_Z" wrapper function called xdbAssignMemo_Z. That wrapper allows users of non-OLE
string programming languages to assign memo field data to Cheetah Type Z memo fields even in
cases where that data may contain embedded Null (Chr(0)) characters.

- Fixed sort expression parser that would incorrectly compile sort expressions that contained
more than one consecutive function. For example, UPPER(LNAME)+UPPER(FNAME)




2.1.8 March 18, 2003

- Corrected bug originating from the 2.1.7 release. Type M memo entries *could* get 
corrupted if their length pushed the final byte to a new memo block (Cheetah Type M memos
are written in 512 byte blocks) and that memo was subsequently edited. This is a somewhat 
rare occurance. Also, Type Z memo files are *not* affected by this problem.

- Corrected problem whereby it would be possible to create a database containing Type "M"
memo fields with a different block size than 512 bytes (i.e. using xdbCreateExtended). All
databases with Type "M" fields can only have memo files with 512 byte blocks. Type "Z"
memo fields can be between 17 and 4096 bytes (also, remember that you can not mix these
two types in the same database).

- Added extra code to flush file headers to disk after a record add or update. This ensures
that the file's headers remain consistent in the event of power failure.

- Added code to support xdbCreateExtended for Visual Basic developers. The previous
declaration did not deal correctly with the field string array being passed to the DLL.




2.1.7 January 27, 2003

- Corrected problem whereby xdbRecordNumber would not return the correct record number if
multiple queries are accessing the same database.

- Corrected problem whereby Error 5005 could be triggered if xdbSkipDeleted is TRUE and
you loop through a database using xdbMoveNext.

- Increased the maximim size of an index key from 256 bytes to 4000 bytes. It is impossible
to create a larger key because the maximum size of any single database record is 4000 bytes.

- The DTOS and CTOD date functions will now return an empty string if the date passed to
them is also empty (no error code will be set).

- Minor cleanup to database structure to ensure the Chr$(26) End of File character is present
at the end of the database. 

- Code change to create Type "M" memo files 100% dbase compatible. Previously, memo files would
not display properly in other xbase products (e.g. FoxPro) due to Cheetah's "garbage collection"
logic that would re-use deleted memo space. The code has been changed to create compatibility,
and still enable the garbage collection. The result is that Type "M" memo files are now slightly
smaller. Cheetah has a built-in conversion routine that will detect the old-style memo file
and create a new one automatically. For safety, the original memo file is renamed with a ".bak"
extension. This process is transparent to the programmer - Cheetah needs only to open the
database and it will make the necessary upgrade.



2.1.6 December 28, 2002

- Corrected problem where memo field *could* corrupt when an existing memo was extended past its
original memo length. For example, expanding a 512 byte memo to 1000 bytes.

- Added support for VB.Net thanks to code supplied by Bob Plymale.

- Added support for Liberty Basic and IBASIC thanks to wrapper code supplied by Walt Decker.
            
            

2.1.5 November 14, 2002

- Corrected memory leak that would occur by opening and closing the same index file hundreds
of times, or reindexing the same index file many times without closing it.

- Correct GPF that would occur on certain Win95, Win98, WinME systems. 

- Corrected reindex percentages during callback.



2.1.4 October 13, 2002

- Corrected xdbCallBack percentages for multiple indexes being Reindexed during an xdbReindexAll call.

- Corrected bug whereby memo fields would corrupt if null values assigned to them prior to any 
regular text being assigned.

- Corrected problem with query processing Type "M" and Type "Z" memo fields using conditions. Previously,
text used for conditions was being limited to 10 characters. 

- Corrected problem where GPF would occur if query contained no results but you attempt to access
the record buffer using xdbFieldValue. The GPF would most often occur when accessing memo fields.

-Corrected problem where calling xdbZap (a few hundred times) would result in a GPF in databases
containing a memo field.



2.1.3 September 5, 2002

- Due to the new reindexing routine introduced in Version 2.1.2, xdbZAP was not resetting the state 
of the open indexes. This has now been fixed. After a ZAP, any open indexes should correctly report 
xdbKeyCount's of zero and their disk file size should be 4K.

- Error in xdbReindex routine would write an incorrect temporary sort file. This would affect huge
indexes that could not be sorted in main memory. Smaller index files worked correctly. If you have
experienced problems reading extremely large indexes after a reindex, then this fix will benefit you.



2.1.2 August 4, 2002   

- Queries now accept parenthesis. This allows you to group several conditions. Previously, Cheetah would
evaluate all conditions based on the %QUERY_AND or %QUERY_OR as they were encountered. The new format
allows subexpressions to be generated and evaluated and then linked to other subexpressions. Refer to the
User Manual under "Query Functions - Using Queries". 

- Aggregate query functions (%MINIMUM, %MAXIMUM, %SUM, %AVERAGE) no longer require a join phrase. Previously,