CDF Changes Log, CDF V3.9 V2.0.0 11-Feb-91 Release of CDF V2.0. V2.1.0 7-Jun-91 Release of CDF V2.1. V2.2.0 20-May-92 Release of CDF V2.2. V2.3.0 1-Oct-92 Release of CDF V2.3. V2.4.0 26-Jan-94 Release of CDF V2.4. V2.5.0 21-Dec-94 Release of CDF V2.5. V2.6.0* 23-Feb-96 Alpha release of CDF V2.6. V2.6.0# 5-Apr-96 Beta release of CDF V2.6. V2.6.0& 13-Jun-96 Beta release of CDF V2.6. V2.6.0$ 17-Jun-96 Beta release of CDF V2.6. V2.6.0@ 15-Aug-96 Beta release of CDF V2.6. V2.6.0 10-Oct-96 Release of CDF V2.6. V2.7.0 27-Sep-99 Release of CDF V2.7. V2.7.1 16-May-01 Add new ports: Solaris on PC, MacOS X and Linux on DEC/Alpha. Correct bugs in CDFDump program. Correct and enhance CDF-Java APIs and Java Native Interface (JNI). Add new Java-based tool programs for CDFEdit and CDFExport. Add variable name checking in CDFconvert for compression option. Add Cygwin port for win32 on PC. Create individual tool programs for Windows using Cygwin libraries. 05-Dec-01 Add a new set of APIs: CDFgetrVarsRecordData, CDFgetzVarsRecordData, CDFputrVarsRecordData and CDFputzVarsRecordData to allow a full single record read/write for a group of r/zVariables. 26-Jan-02 Correct a bug in cdfcmp.c for showing dimension size differences between two zVariables. 28-Jun-02 Add a new option in cdfcmp.c to allow tolerance checks while comparing two unequal data values. 22-Jul-02 Handle f77 Fortran for Cygwin. 25-Nov-02 Add 64-bit mode for Solaris/sparc64 for sparcv9. Also, gcc for 64-bit is added. V2.7.2 08-Apr-04 Change the way that current variable select is being handled. Keep the current variable selected offset as a reference for any following variable selection. Originally, each time a selection of a variable or getting a variable's field data will start the search for that variable ALWAYS from the beginning of the variable list. It requires too many I/Os for accessing a sequence of variables. 04-May-04 Corrected a bug that caused the f77 to fail under 64-bit environment. V3.0 07-Jan-05 Changed file offset from type long (32-bit) to off_t (64-bit) for the platform on which the data type is supported. Changed the file I/O functions accordingly. The CDF internal file structures are changed. Made the library backward compatible, meaning programs created from V3.0 can still access (read/update) the CDF files of older version. Expanded the length of variable and attribute name from 64 to 256. The CDF/Java was modified to ensure that concurrent accessing a CDF would be thread-safe. Added a new data type, CDF_EPOCH16, to accommodate more refined time resolution within a second. A new set of functions, similar to those of CDF_EPOCH data type, was added. Modified cdfdump tool program to be more efficient in data reading. 25-Mar-05 Changed JNI and Java-CDF APIs to ensure proper operations within the multiple-threaded environment. V3.1.0 27-May-05 Added new sets of APIs to allow Standard Interface to interact with zVariables and other CDF-related information. 11-Jul-05 Added MingW port for PC. 05-Aug-05 Added new functions to allow creating CDF files in older, V2.7, version, not just the default V3.1. 18-Jan-06 Modified code, including tools, to handle file path that has ".cdf" or ".skt" extension while it is not expected to be there. 30-Jan-06 Add FreeBSD port for PCs. 1-Feb-06 Added Intel C++ and Fortran compilers for Linux port on PCs. 22-Jun-06 Added support for HPUX and AIX for both 32 and 64-bit mode. Allow Mac OS X to build code for PPC and x386. V3.1.1 12-Oct-06 Modified to allow upper/lower case CDF name for Windows. Changed Epoch to make 9999-12-31 23:59:999 as encoded date for the filled value of -1.0E31. For Epoch16, a pair of -1.0E31 filled values is encoded as 9999-12-31 23:59:999:999:999:999. V3.2.0 21-Oct-06 Added MD5 checksum feature for data integrity check of the CDF. Modified tools to use the checksum feature. Added a couple of new tool programs: cdfdump and cdfmerge. Renamed the original cdfdump to cdfirsdump. 25-Apr-07 Changed the default size of cache buffers from 512 to 10240 bytes to improve data access performance. 18-Jun-07 Enhanced READONLY mode to improve metadata access performance. When READONLY mode is selected, all metadata is read and stored in internal data structures which are then accessed whenever metadata is requested. V3.2.1 24-Apr-08 Modified the library so a potential buffer overflow vulnerability when reading specially-crafted (invalid) CDF can be avoided. V3.2.2 10-Aug-08 A maintenance release. Modified the Java tools so they can handle CDF files with space(s) in the file path. cdfmerge tool was modified to allow merging "Epoch" variable data just like other variables. V3.2.3 24-Nov-08 Modified CDFdump to add a new output option. V3.3.0 05-Dec-08 Added an optional process to validate data fields in a CDF when it is open. This process addresses a potential issue of library becoming vulnerable when a compromised CDF is accessed. A failed validation from a CDF will cause the application to end gracefully. All CDF tools will have sanity checks on. A new tool, cdfvalidate, is added. Added an option to cdfdump to allow user-specified variables to be dumped. V3.3.1 10-jan-11 Added several new features to the cdfexport tool program: -[include | exclude] " | varsfile=" -[epochrange | recordrange] "" Bug fixed in Java-CDF APIs and a couple of new methods were added. Added an option to cdfdump to allow a range of records to be dumped. MingW port was revised to handle 'pdcurses', if installed, for the curses-based tools. Fixed cdf validation to allow some V2.0 files to be valid. Library bugs fixed: read/delete/read scenario, updated alloc/max record number when adding records to a compressed variable before closing the file, modified last record number in VXRs when records are deleted. Replaced calls to sprintf/vsprintf with snprintf/vsnprintf. Added a new set of functions to allow encode/parse epoch data in ISO 8601 format in the CDF library. 01-Aug-11 For version 3.3.1.1 Modified code to handle 64-bit Fortran on Solaris/Intel port. Minor fixed in the CDF java and cdfml class methods. V3.3.2 05-Aug-11 Added two new data types: CDF_INT8: 8-byte signed integer CDF_TIME_TT2000: 8-byte integer of epoch from TT2000 with leap seconds A new tool, cdfleapsecondsinfo, to show the information i of the leap seconds used in the CDF. Change the default CDF encoding to the most popular hardware of IBMPC_ENCODING. Added an opion to convert epoch between CDF_EPOCH/EPOCH16 and CDF_TIME_TT2000 in cdfconvert tool. 15-Aug-11 For version 3.3.2.1 Fixed the code to handle TT2000 base time. 01-Sep-11 For version 3.3.2.2 Modified code to handle 64-bit Fortran on Solaris/Intel port. Minor fixed in the CDF java and cdfml class methods. A fixed bug in the library. 11-Jan-12 For version 3.3.2.3 Modified email address for CDFsupport. Modified cdfml: DTD and XSD to add CDF_INT8 and CDF_TIME_TT2000 data types. Changed date/time parameters for compute_TT2000 and TT2000_breakdown functions from double to int, similar to compute_epoch and EPOCH_breakdown. Minor change in CDFUtils java code. Revised default pad, dummy and illegal values for TT2000 data type. V3.4.0 17-Feb-12 Convert V3.3.2.3 to new version. Reset pad values for INT8 and TT2000 data types. 10-May-12 For version 3.4.1.0 Bugs fixed in core library. Minor changes in skt2cdf tool program and cdfdump help. Fixed cdfmerge while using text file for input files control. Modified Java CDF class's open method to delay collecting variable and attribute data. V3.5.0 25-Feb-13 The original beta version for version 3.5.0.0 Used open source zlib source code to replace CDF's GZIP compression/decompression, which was modified from original code by Jean-loup Gailly and Mark Adler, in the previous versions. Bugs fixed. Modified the default blocking factor for compressed variable data. New features were added in cdfdump. 15-Sep-13 for version 3.5.0.1 (the official release version) Fixed minor memory leaks in core library and JNI. Added a new option to cdfdump tool for how to show the floating-point values if FORMAT entry is missing. Added support for Visual Basic on Windows. Pad value is set when a variable is created. Fixed code to save CDFid for 64-bit Fortran when a CDF is open/created, even only 4-byte of it is used in Fortran code. Default pad values are set with rather invalid values, so they can be recognized easily. 31-Mar-14 for version 3.5.0.2 Use the latest zlib version 1.2.8 without code changes. Minor code changes for filling space(s) for writing string type pad value. 64-bit Windows package now distributes 64-bit Cygwin-based libraries and tools. Bugs fixed. Added support for nan, inf and -inf for floating point values in the tool programs. V3.6.0 5-Feb-15 for version 3.6.0.3 Added a new leap second for 2015-07-01 to the leap second table. Added a field "LeapSecondLastUpdated" in GDR to record what leap second table a CDF is based upon. A set of functions to read/write/validate the field is added. Added an option to renumber the sparse variable records after a record(s) is deleted. Added sort and detect options in tool cdfconvert. Skip checking the file/path name's characters when reading a CDF file. Enforce ASCII chars for file name (not including the directory portion) only. Minor bug fixes. 1-May-15 for version 3.6.0.4 Modified cdfjava.jar. Extended status text length. V3.6.1 20-Sep-15 for version 3.6.1.0 Used the preserved system temporary folder, e.g., /tmp for Linux/Unix/MacOSX, to hold the temporary file(s) while doing compression/decompression. Set temporary file name using the random number generator with process id and current time as the seed. Added support for computing TT2000 from UTC if the passed day is DOY (day of the year from January 1st). Added a few options to cdfconvert tool program. V3.6.2 20-Mar-16 for version 3.6.2.0 Modified to Makefile and installation process to support Mac OS X El Captain (10.11). Modified the library to handle string data that has a shorter length than defined (number of elements) when reading/writing. It will be filled with spaces, starting from the NUL. 13-May-16 for version 3.6.2.1 Bug fixed for handling TT2000 time computing if only year/month/day is given. Speed up the leap second time/table validation. V3.6.3 20-Oct-16 for version 3.6.3.0 Added a new set of C-based APIs for reading the CDF data. Each of these APIs, identified as CDFread..., will return a pointer to the library allocated space for the data. Caller is responsible to free the space after use. New features were added to cdfstats and cdfdump tool programs. A new leap second is added for 1/1/2017. 10-Jan-17 for version 3.6.3.1 Updated the leap second last updated header in a CDF from the (latest) leap second table if the CDF is created based on an existing master CDF that has an older leap second table. V3.6.4 10-Feb-17 for version 3.6.4.0 For Unix-based systems, use C's "mkstemp" function to create the temporary files, in the template form of "mycdftmp.XXXXXX". On Windows, use VS's "_mktemp_s" to do the same. Use the random number generator to create the temporary file name as the last resort if C's function fails to create one. For Unix-based systems, the temporary files will be created in the current directory if it is writable and neither environment variable CDF_TMP nor TMPDIR is defined. Otherwise, it will try the directory defined by the environment variable CDF_TMP first and then TMPDIR. If the current directory is not writable, then system's "/tmp" is used. On Windows, both "TMP" and "TEMP" directory will be tried if CDF_TMP is not defined. If all fail, the last resort is to use the current directory. Modified cdfexport to not truncate variable name while displaying its data. Have a separate patch for supporting IDL 8.6. V3.7.0 11-May-18 for version 3.7.0.0 For variable attribute, multi-strings are allowed. A header field in the internal data structure for entry is used to stored the number of strings. A function to free the retrieved strings, when reading, is created. The data validation control process, thru CDF_VALIDATE environment variable, is modified to also skip the checksum verification, in addition to the data sanity check, if the validation is turned off. cdfconvert tool allows the output of the converted file to replace the source file. Added support for ARM ports, little or big-endian. The string-typed data for attribute entry or variable data can be Nul-terminated. The default pad value can also be empty. Previously, space(s) is filled if the data length is shorter than the defined number of elements. Added openVMS on Itanium (IA64) ports: IA64VMSi, IA64VMSd and IA64VMSg. Added cdfj.jar, the pure Java package for CDF read and write without needing CDF library and JNI. A revised cdf schema is included for CDF's XML support. V3.7.1 21-Aug-18 for version 3.7.1.0 Modified to make the encoded date/time string for CDF_EPOCH and CDF_EPOCH16 in the form of ISO 8601 as the default, as yyyy-mm-ddThh:mm:ss.ccc or yyyy-mm-ddThh:mm:ss.cccuuunnnppp respectively. Enhanced the CDFML xsd schema. Do not throw an error if the ILLEGAL_TT2000_VALUE is encountered. When encoding its value into date/time string, it is presented as 1707-09-22T12:12:10.961224195. (Read CDF's C document.) Add a set of general encoding/parsing functions for each of the CDF epoch data type. Added a set of time conversion functions between CDF_EPOCH/CDF_EPOCH16/CDF_TIME_TT2000 time and unix time. Reset the rVariables' dimension to zero if a cdf is converted, by cdfconvert, with "zMode 2". V3.8.0 27-Oct-19 for version 3.8.0.0 Modified the code to use a variable's FILLVAL, replacing its PAD value, for all record varying variable's virtual record data if it exists. To use FILLVAL, it must have a data type equivalent to its variable's. String typed variable's pad value is filled with a single space and followed by NUL(s), instead of multiple spaces. Changed skt2cdf tool program to allow handling a skeleton table directly from Windows (with \r\n at the end in each line) on non-Windows systems. Added new options to cdfconvert tool program to 1. remove non-varying dimension(s) from the source zVariable(s) if it does not have DEPEND_* attribute defined to the destination variable. 2. replace any pad value(s) in a variable's data by its FILLVAL value, if FILLVAL attribute exists and has an equivalent data type as variable's. Modified the tool programs to use FORMAT attribute to encode both data and metadata if the format is to be used. These include all C-based and Java-based tools. A new option is added to cdf2skt tool to allow users to choose how to display variable's metadata and data, either with or without format. 7-Jul-20 for version 3.8.0.1 Modified CDF epoch breakdown functions that might cause incorrect day being returned. V3.8.1 22-Mar-22 for version 3.8.1.0 Added a set of easy-to-use read functions: READCDF and READVARIABLE, to read CDF and variable information for Java, C#, Perl and IDL interface in a single call. Allowed building the fat JNI from the main CDF Makefile build. Modified TT2000 type handling routine in C/Java to accept minus leap second in the future. Added Json-CDF converters in Java package. Enhanced netCDF-to-CDF converter. Support characters in UTF-8 encoding. Dynamic space allocation is no longer placed into a linked list. Modified TT2000 handling function to handle reverse leap second (no 59 second). Added support for Apple M1 processor. V3.9.0 22-Jan-23 for version 3.9.0.0 Increased the regular cache buffer size to improve the hit rate for getting the CDF internal information. Modified the variable compression section to avoid making a temporary file by increasing the number of cache buffers, and flushing out the old caches when they reach a point that processing the next variable's compressed data would cause a temporary file to be created as the remaining caches will not be enough. The blocking factor is limited to a value, computed from CDF's maximum caches, so variable's compression operation will not trigger a temporary file(s) being created. A bug fixed for variable's allocated from/to operation for a compressed or sparse variable. Increased the cache buffer size for Linux/Mac to improve the performance. Bugs fixed for curses-based tool programs. Moved src/tests directory to cdf_tests directly within the cdf home to make test programs there more visible.