Release Notes (May, 2009)

CORELIB — Portability and Application Framework

• (*)CTime — removed deprecated operators to add/subtract/increment/decrement days.

• CTime — added CTime ::SetTimeTM() to convert from an arbitrary "struct tm" value; CTime ::GetTimeTM() to convert the current local time to a "struct tm"; and CTime ::CTime(const struct tm& t, ETimeZonePrecision tzp).

• CStaticTls — new class that combines CSafeStaticRef and CTls but requires less overhead.

• NCBI-Boost unit testing framework — test_boost library now includes Boost.Test library as a whole, so there's no need to link against Boost.Test library separately when linking against test_boost.

• NCBI-Boost unit testing framework — now arranges to incorporate Boost.Test's (compiled) code into libtest_boost, thereby requiring only headers from Boost.

• NCBI-Boost unit testing framework – macro NCBI_BOOST_NO_AUTO_TEST_MAIN is now obsolete (everything works whether or not it is used).

• NCBI-Boost unit testing framework — added support for timed out and skipped tests and for timed out units.

• NCBI-Boost unit testing framework — introduced new analogs of BOOST_CHECK* macros that also check the NO_THROW condition.

• CAtomicCounter_WithAutoInit — new class that enables creating counters initialized with a specified value.

• CNcbiResourceInfo and CNcbiResourceInfoFile — new classes providing sensitive data encryption/decryption.

• CNcbiRegistry — reworked once more, with generic functionality factored out into a new CCompoundRWRegistry class; now allows automatic loading of subregistries listed in the ".Inherits" entry of the "NCBI" section, shadowing of set entries on lower layers by explicitly empty entries on higher layers, and the environment variable NCBI_CONFIG_OVERRIDES to name an extra high-priority configuration file.

• CAutoEnvironmentVariable — new class to allow setting environment variables for limited durations, with restoration of previous values when instances go out of scope.

• CExceptionReporter, CDiagMatcher — added diagnostic filtering by error code when reporting exceptions. Diagnostic filtering is set by SetDiagFilter function.

• CStringPairsParser — converted into CStringPairs template to allow using different containers, added methods for merging data into a single string.

• CProcess — new method Daemonize(), formerly a static function resided in ncbi_os_unix.h.

• GetPhysicalMemorySize() — implemented on a wider variety of platforms.

• CDirEntry — A new mode (eRecursiveIgnoreMissing) for the Remove() method was introduced. This mode deletes the files and directories recursively but ignores the errors when the file being deleted does not exist.

• The DeleteReadOnlyFiles CParam<> and the corresponding Set() method of CFileAPI were introduced to work around CDirEntry ::Remove() not being able to remove read-only files on Windows.

CONNECT — Data streaming, Networking, and Dispatching

• Implemented HTTPS protocol support.

• Added new function UTIL_PrintableString[Size]().

• Added new function SERV_ServerPort();

• Retired wsock32 use on Windows; require (and link against) ws2_32 exclusively.

• Fully implemented new TRIGGER primitive.

• CSocket, SOCK_xxx — now can work via SSL (requires GNUTLS API).

• Introduced new (and centralized) ESOCK_Flags for various socket attributes,

• and used them throughout the library (deprecating older equivalents).

• The default connection managing mode was changed to eKeepConnection.

UTIL — Miscellaneous Low-Level APIs

• CZipCompression — added support for concatenated gzip files.

• CFormatGuess — the Format() method now throws an exception in the case of missing input. Format guessing was expanded to include streams other than file streams. FASTA files can now be properly recognized even in the presence of some non-printing characters or very long deflines. Added "guessing" support for the BED, BED15, and WIGGLE file formats. For the AGP format, the presence of line comments starting with a '#' is now allowed.

• CTar — improved to work safely with program pipes (including buggy implementations in certain GLIBC releases).

SERIAL — Data Serialization (ASN.1, XML, JSON)

• CObjectOStreamJson — corrected the writing of UTF8 strings in JSON format.

DATATOOL — Code Generator and Data Converter Tool

• Added possibility to override code generation command line arguments in a DEF file.

• Added support for more data types, including dateTime, time, short, byte, negativeInteger, nonNegativeInteger, positiveInteger, nonPositiveInteger, unsignedInt, unsignedShort, and unsignedByte.

• Fixed several bugs in XML Schema parsing.

CGI — CGI and Fast-CGI Application Framework

• CGI library — applications can no longer send secure cookies over insecure connections.

• Fixed parsing of indexes (arguments without values) in query string.

• A new constructor CCgiRequest ::CCgiRequest(CNcbiIstream&) was defined to allow construction of CCgiRequest by deserialization from a stream.

DBAPI — Generic SQL Database Connectivity

• IResultSet — BindBlobToVariant() now does nothing - how the user wants to read blobs is now automatically determined by Read() (which also means by NcbiIstream) or by GetVariant().

• Fixed incorrect column naming in the SELECT results with Sybase when the column is just renaming a real column in the database.

• Introduced possibility to add hints for bulk insert operations (such as CHECK_CONSTRAINTS, FIRE_TRIGGERS, etc).

• ftds8 driver is not built automatically anymore and its sources will be removed in the next release.

• Added support for connection to mirrored databases for which master/mirror relations can be switched while application is working.

BIO-OBJECTS — Bio-Object Specific Utility Functions (Not Involving OM++)

• CSeq_id ::IdentifyAccession — added or improved recognition for the prefixesAH, AL, BX, CR, CT, CU, DAAA-DZZZ, DM, FT, FU, GJ-GL, GO-GZ, HA-HD, and HAA-HZZ and some more (mixed-in) EMBL TPA protein accessions.

Local data storage (LDS)

• Dramatically reduced memory usage in LDS indexer by processing entries separately.

ALGO/ALIGN — Generic Alignment Algorithms

• CCompartmentFinder - the maximum intron length parameter has been exposed; the default max intron length increased to 1.2M bases.

• CCompartmentAccessor — added AsSeqAlignSet() method to support ASN.1 output.

COBALT — Multiple sequence alignment tool

• (*) Changed API and provided CMultiAlignerOptions class that provides multiple alignment parameters.

• (*) Rearranged the parameters for the COBALT demo application.

• Added query clustering functionality for multiple alignment.

• Added computation time improvements to multiple alignment.

• Added functionality for computing progressive alignment guide tree as a clustering dendrogram.

OM++ — Object Manager — For Retrieving and Processing Bio-Objects

• CSeq_entry_EditHandle now has method TDescr& SetDescr() which works for both bioseqs and bioseq-sets.

• Object manager now recognizes named annotation accessions with versions.

• Object manager now indexes non-feature Seq-tables, and allows extra columns for feature Seq-tables.

• Bioseq and Bioseq_set objects now forbid empty description field by default.

• The default timeout for opening an ID1/ID2 connection was reduced from 20 to 5 seconds. The default timeout after connection is established remains 20 seconds.

• Added sequence::FindLatestSequence() method for searching through a Bioseq's history.

• Fixed mapping of seq-graph data. The old version mapped only graph location, but left the related data array unchanged. The fix creates a new array and copies only the mapped portion of the data. As a result of this, a new valid seq-graph object is created (the old version created seq-graph with incorrect data). Added methods to CGraph_CI for checking mapped ranges without creating the whole mapped data array.

OM++ LOADERS/READERS — Data Retrieval Libraries for OM++

• Added support for the UCSC WIGGLE file format. Enhanced support for UCSC BED and microarray file formats.

• Preliminary version of an idmapper component that translates UCSC sequenceIDs to their corresponding GI IDs.

OBJTOOLS/SEQMASKS_IO — input/output of masked sequences

• Fixed a problem with sequence writer omitting accessions when ID parsing is requested.

ID2 Communication Protocol

• Blob-state information in ID2 reply was changed from ENUMERATED to INTEGER with bit flags. This field is not yet used by ID2 server or ID2 GenBank reader.

BIO-TOOLS

• Implemented a faster non-OM version of sequence::GetTitle(CBioseq), although it may fail if called when OM is necessary.

• format — The output of the flat file generator in the C++ Toolkit is intended to be essentially the same as the output in the C Toolkit. That was not achieved in the last release, so changes were made in this release to make the output more closely match the output obtained using the C toolkit.

PTB — Project Tree Builder

• Improved analysis of library-to-library dependencies; added possibility to enforce build order of static libraries.

• Enhanced the structure of the flat makefile on UNIX to speed up the build by reducing the number of calls to 'make' and avoiding attempts to build the same project more than once.

APPLICATIONS

Location

The documentation is available online as a searchable book "The NCBI C++ Toolkit": http://www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=toolkit.TOC&depth=2.

The C++ Toolkit book also provides PDF version of the chapters; although these are not up to date in this release. The PDF version can be accessed by a link that appears on each page.

Content

Documentation has been grouped into chapters and sections that provide a more logical coherence and flow. New sections and paragraphs continue to be added to update and clarify the older documentation or provide new documentation. The chapter titled "Introduction to the C++ Toolkit" gives an overview of the C++ Toolkit. This chapter contains links to other chapters containing more details on a specific topic and is a good starting point for the newcomer.

A C/C++ Symbol Search query appears on each page of the online Toolkit documentation. You can use this to perform a symbol search on the up-to-date public or in-house versions using source browsers Entrez, LXR, Doxygen and Library - or do an overall search.

HEADS-UP: We have switched our source control system from CVS to SVN (Subversion). Unfortunately, the SVN repository cannot (yet) be accessed from outside NCBI.

UNIX

^a some support

^b nominal support

^c 32-bit only

MS Windows

Mac OS X

Added Platforms

Discontinued Platforms

GCC 3.0.4

• Destructor of constructed class member is not called when exception is thrown from a method called from class constructor body (fixed in 3.3).

• STL stream uses locale in thread-unsafe way which may result in segmentation fault when run in multithread mode (fixed in 3.3).

• Long-file support for C++ streams is disabled/broken (first broken in 3.0; fixed in 3.4).

GCC 3.4.x, 4.0.x

• At least on Linux, ifstream::readsome() does not always work for large files, as it calls an ioctl that doesn't work properly for large files (we didn't test whether 4.0.x fixed this).

• GCC 3.4.4 has a bug in the C++ stream library that affects some parts of our code, notably CGI framework (fixed in 4.0.1).

ICC 8.0

• ICC 8.0 lacks large file support for C++ streams on 32-bit Linux (fixed in 10.1).