~vcs-imports/pxlib/head

Version 0.6.4
	- fixed bug in PX_timestamp2string()

Version 0.6.3
	- some man page updates
	- check for more header files
	- some code cleanups

Version 0.6.2
	- added support for cmake
	- various small fixes to make it compile in a mingw environment
	- works on amd64 (and probably other 64 bit architectures) again

Version 0.6.1
	- much better handling of databases with corrupted block number
	- fixed rpm spec file (Copyright is now License)
	- read tablename propperly from .Xnn and .XGn files
	- set pxdoc->px_head to NULL in PX_new3() (allows to detect PX_new() without
	  PX_open_fp())
	- Fixed various error messages issued by pxlib
	- PX_get_blob_data() returns 0 if blob is null.
	- PX_insert_record() and PX_update_record() treat null values propperly.
	- PX_insert_record() automatically counts up fields of type pxfAutoInc
	  if the value is NULL. PX_create_fp() checks for more than one pxfAutoInc
	  field and issues a warning, because the automatic incrementation works
	  only with one field of that type.
	- various man page updates
	- PX_delete_record() now updates the internal index propperly
	- fixed some compile errors when g++ is used, fixed some compile warnings
	- fixed memory leak in PX_insert_record() and PX_update_record()

Version 0.6.0
	- fixed minor memory leak cause by memory used for the pxs_stream
	- fixed reading of fields of type BCD
	- conversion from multi byte to one byte strings in PX_put_data_alpha()
	  terminates the resulting string propperly
	- close file which was opened by PX_create_file()
	- Add get_short_le_s() which returns signed short int because the size
	  of a datablock as stored in the datablock header can be negative.
	- fixed overflow in px_head->px_fileblocks which is now an unsigned
	  int, as well as px_firstblock and px_lastblock. Also make get_short_le()
	  return unsigned int. Thanks to Arkadiusz.
	- fixed calculation for number of blocks needed by a blob which
	  is stored in block of type 2
	- _px_put_data_blob() doesn't store more than 64 blobs in block
	  of type 3
	- use a cache for reading encrypted blobs
	- make databases opened with PX_open_file() writable
	- added new functions PX_delete_record(), PX_update_record(),
	  PX_insert_record() allowing to modify an existing database
	- added new function PX_retrieve_record() which delivers a record
	  as an array of (pxval_t *). This is a functional replacement for
	  PX_get_record()

Version 0.5.1
	- fixed bug when reading header of blob (Thanks to Giovanni Giacobbi)
	- fixes bug when reading blobs of a certain size from an encrypted
	  .mb file.
	- reading and writing encrypted blob files is more efficient.

Version 0.5.0
	- support for encrypted databases and blob files

Version 0.4.4
	- PX_get_data_alpha() may now return strings which are longer than
	  the database field (e.g. when converting to a multibyte string)
	- fixed reencoding which failed when the input string passed to
	  iconv contains zeros
	- codepage can be set when creating a database
	- fixed copy and paste error in doc/pxlib.sgml
	- added new function PX_create_blob_fp()
	- fixed many compiler warnings and some potential errors
	- added missing prototypes in paradox.h.in
	- allow to turn off warnings
	- PX_set_blob_file() works when reading and writing a file
	- added missing man page for PX_create_blob_file()
	- added overall man page for pxlib with many examples
	- PX_get_value() now returns many fields from the header which
	  makes accessing the header struct directly mostly superfluous

Version 0.4.3
	- added PX_timestamp2string() for conversion of paradox timestamps
	  into a readable representation
	- various fixed for writing blob files
	- return value for bcd fields uses decimal_point from locale
	- added spanish translations (thanks to jrodriguez@innox.com.mx)

Version 0.4.2
	- fixed possible errors in header of mb files
	- added support for type 3 blocks in mb files

Version 0.4.1
	- various updates of debian files
	- small changes to the build process

Version 0.4.0
	- support for writing blobs
	- some man page updates

Version 0.3.1
	- fixed segm fault in put_px_datahead() when second datablock is written
	- free memory when pxblob_t is deleted
	- separated code for recoding into encode.c

Version 0.3.0
	- various man page updates
	- Blob data stored in the record itself can be read
	- added function PX_set_blob_file(), PX_has_blob_file()
	- added function PX_get_data_bcd(), PX_put_data_bcd(), PX_get_data_blob(),
	  PX_get_data_graphic()

Version 0.2.11
	- PX_read_blob_data() has been completely rewritten and can now handle
	  any type of blob data. Watch for the new prototype.

Version 0.2.10
	- creating a paradox db without records led to a corrupted file
	  because the header was not fill with zeros to its full size.

Version 0.2.9
	- new function PX_strdup() to duplicate a string but use the
	  memory management functions of the paradox document.

Version 0.2.8
	- new function PX_get_opaque()
	- fixed all references to PX_open_filename which is now PX_open_file
	- pass const char to API function if possible
	- updated build process
	- Better checking whether px file belongs to db file

Version 0.2.7
	- added man pages for blob file handling
	- set some more header fields. db and px files are now read correctly
	  by the Paradox software.
	- changed default file version to 7.0

Version 0.2.6
	- added new function PX_get_recordsize()
	- set indexRoot and numIndexLevels in .PX files
	- fixed segm fault when writing a primary index file
	- fixed reading of blobs
	- two new header fields are known (indexRoot, numIndexLevels)
	- many more translations
	- some more tweaking to finally store the file update time in the
	  header without messing it up

Version 0.2.5
	- fixed man pages of PX_set_value|parameter. Both functions return int now.
	- PX_set_parameter() returns -1 on error and 0 otherwise
	- Issue a warning if setting a value which does not exist
	- better error checking when setting the number of primary keys
	- fixed some errors in writing primary index files
		level 2 index entries are now supported
	- readded time of modification when writing the header

Version 0.2.4
	- Paradox tables can be read by the Paradox software if the
	  database is stored as version 5.0
	- added prototypes for PX_[get|set]_[parameter|value]()
	- all strings passed to PX_[get|set]_[parameter|value]() are of
	  type const char
	- PX_get_[parameter|value]() return value in passed parameter and
	  return an error code instead
	- Implemented function PX_get_parameter()
	- All PX_get_data_xxx() functions return -1 on error, 0 if the
	  field value is NULL and 1 otherwise.

Version 0.2.3
	- updated many error messages
	- fixed segm fault when PX_close() was called more than once
	- initial support for writing a primary index (experimental)

Version 0.2.2
	- initial multi language support
	- added memory profiling
	- free some more memory in PX_delete()
	- renamed internal memory management functions from px_malloc, px_realloc,
	  px_free to _px_malloc, _px_realloc, _px_free
	- call pxdoc->free instead of px_free
	- fixed potential buffer overrun
	- PX_set_[target|input]encoding returns -2 if recoding is not supported

Version 0.2.1
	- PX_create_fp uses a fixed header size which was to small for databases with
	  many fields. The header size is now calculated and set accordingly.

Version 0.2.0
	- update to latest autoconf tools
	- added support for pkg-config
	- added new function PX_new3() which allows to also set user data which
	  is passed to the error handler
	- added generic way to set and get parameters
	- added support for external read, write, tell, seek functions

Version 0.1.9
	- encoding for data, that is to be written in a record can be set
	- build internal primary index for faster access. This will not
	  speed up searching but accessing records by record number.
	- fixed reading of primary index, now that I understand how it works
	- fixed data block handling (still not perfect, but works if records are
	  stored sequentially)
	- fixed several errors in fileformat which where due to lack of information
	- provide header field 'fileUpdateTime'
	- access data block as linked in list and not as stored in file
	- new functions to manipulate record data
	- made many error message better understandable
	- initial support to write db files

Version 0.1.2
	- primary index files can be read
	- added recno to struct px_datablockinfo
	- PX_get_record returns in referenced parameter deleted whether a
	  record is deleted or not

Version 0.1.1
	- fixed segm fault when calling PX_get_record()

Version 0.1.0
	- New function PX_get_record2() which has two extra parameter.
	  One to enable reading of deleted records and one to return
		internal data about the data block from which the record has
		been read.
	- added pxh->theonumrecords which is the number of records which
	  theoretically fit into all data blocks. This number can be
		much bigger that the number of records in the header. Also
		increases the max. record number to theonumrecords
		when reading a record.
	- DB files can contain many more records than the number of records
	  in the header. The remaining records are left overs, which were
		deleted but not removed from the file. pxlib can now handle those
		files.

Version 0.0.9
	- man pages are not created if docbook-to-man is missing
	- fixed many endianess bugs
	- new function PX_is_bigendian()
	- new function PX_has_recode_support()
	- fixed compile error with older versions of gcc
	- fixed two potential segm faults

Version 0.0.8
	- more documentation
	- recoding of data fields with iconv

Version 0.0.7
	- read more fields from the header
	- added spec file to build rpms
	- alpha fields are automatically recoded if required
	- function to access alpha field data

Version 0.0.6
	- more documentation
	- more functions to access database schema (PX_get_field(), PX_get_fields())
	- much better check for errors as not passing a document or out of
	  range values

Version 0.0.5
	- added PX_delete()
	- many memory leaks fixed
	- proper use of memory allocation functions

Version 0.0.4
  - fixed segm fault if input file does not exist
	- maxTableSize may now be larger as 32
	- add check for endianess
	- added data conversion functions

Version 0.0.3
  - check for zero record size and header size
	- added missing header files in distribution
	- added conversion routine from Gregorian date to serial date number