New functionality:

• Added macros BEGIN_*_NAMESPACE that require semicolon and do not break indentation.

• Added CObjectCounterLocker::MonitorObjectType() for debugging.

• Added CObjectCounterLocker::TransferLock() and updated CRef<>::Swap() to support MonitorObjectType().

• Added ICanceled interface.

• Allow locking input data in stream buffer.

• Added UNICODE support. Internally, the Toolkit works with UTF8-encoded strings; when needed, they are converted into UNICODE and back automatically.

• Added condition variables.

• Added fast double-to-string conversion, which is 5 to 10 times faster than standard sprintf.

• CArg(Allow)_Int8 -- Accept explicitly marked (0x...) hexadecimal integer argument values.

• MemoryAdvise() -- Advise on memory usage for specified memory region (Unix).

• CPerfLogger/CPerfLogGuard -- Performance logger. Measure time spend somewhere and put result to the performance log.

• FileAPI::SetHonorUmask() -- add configurable parameter to honor umask settings (Unix).

• NStr::StringToSizet/SizetToString -- special case for size_t type, thats size depends from platform/OS.

• NStr:: -- added [U]LongToString() variants. Deprecate incompatible types for [U]IntToString().

• CProcess::GetCurrentHandle() added

• CDirEntry::SetOwner() added

• An option for CProcess::Daemonize() to retain parent process

Improvements:

• Changed macro NCBI_EAT_SEMICOLON to extern declaration so that uniqueness is not required anymore.

• Use more efficient direct type_info comparison.

• Optimized errno access in NStr::StringToXxx().

• Optimize MD5 step functions.

• Print stack trace with critical error messages.

• Optimized StringToDoublePosix().

• Fixed processing [+-]infinity and error detection in number parsers.

• Return starting ptr if no number parsed and there are leading spaces.

• Added missing method necessary for CIRef::Swap().

• Added more restrictive bool operator.

• Enable safer iterator-dealing code if _GLIBCXX_DEBUG is defined.

Bug fixes:

• Fixed parsing of "+" and "-" strings in NStr::StringToXxx().

• Fixed sorting of CVersionInfo and SDriverInfo.

New functionality:

• Added support of different forms of namespace qualification as described in XML Schema.

• Added mechanism of matching ASN binary data against set of known data types

• Added a suite of sample programs:

asn_assign, asnwalk_read, asnwalk_type, asnwalk_write, bss_info, hooks_commented hooks_copy_member, hooks_copy_object, hooks_copy_variant, hooks_read_member, hooks_read_object, hooks_read_variant, hooks_skip_member, hooks_skip_object, hooks_skip_variant, hooks_write_member, hooks_write_object, hooks_write_variant, seqannot_splicer

• Allow cancellation via ICanceled callback.

• Implemented global pre-read hooks for class members and choice variants.

Improvements:

• Set internal names for inner classes, choices, and enums.

• Make error message in ThrowUnassidned() more informative for inner classes.

• Make configurable reading/writing of UTF8String tag.

• Allow ambiguous class names until lookup.

• Lock input data in buffer when guessing type.

• Make datatool -t argument optional, guess data type if necessary.

• Reimplemented pre-read member hooks as objects instead of function pointers.

Bug fixes:

• Stop AnyObject reading on closing brace.

• Fixed UTF8 processing in CopyString() and SkipString().

• Use UTF8String tag in ASN.1 binary.

• Correctly process UTF8 strings in datatool.

• Allow trailing spaces in delayed-parsing buffer.

• Reset typeinfo with null object pointer to make CObjectTypeInfo and CConstObjectInfo behave equally in bool context.

XMLWrapp:

• Respect the document encoding (if so) while printing or streaming nodes.

New features:

• Identify HTTP connections with AppName in User-Agent Unicode (UTF-8) system error messages

• SOCK_SetCork() added

• CONN to use new supplement (look ahead) I/O modes for faster I/O SERVICE connectors to maintain NCBI_SID when detected CConn_ServiceStream's ctor to accept customizable user-header CConn_PipeStream to be able to get from the child's both STDOUT and STDERR output streams (merged together)

• CONN_Status() to accept eIO_Open

• Flex firewall mode (per checks with the help of CConnTest). Allow proxy leak configuration (to bypass the proxy if needed). CONN cancellation via callbacks. Allow bound sockets (listening and datagram) to use port "0" for OS auto-assign

Bugfixes:

• SOCK_Poll() to return error on closed / unopened sockets Detect quotes in the CONN environment / registry settings FTP stream to issue errors on incomplete uploads

• ConnNetInfo_ParseURL() to fail for too long URL elements (no truncation)

• ConnNetInfo_ExtendHeader() to check for duplicates (and avoid them)

• CConn_MemoryStream::ToCStr() deprecated and dropped

• CRateMonitor::GetETA() fixed

• Workaround IAC-unclean implementations of FTPD

• CORE_GetUsername() to avoid using getlogin() if at all possible

New functionality:

• Added CSeq_interval, CSeq_loc_mix, CPacked_seqint getters for start and stop.

• Added method CompareOrdered() and predicate PLessOrdered to CSeq_id_Handle.

Improvements:

• CSeq_loc::Compare() now checks Seq-id and circular locations.

• Adjusted CSeq_feat comparison to match C Toolkit and CFeat_CI.

• Added CPDB_seq_id::Compare().

• Prefer gnl|ti|* over gnl|TRACE|*.

• Scoring functions made non-inlined.

• Added matching of integer and string versions of the same general id.

• Changed Seq-table's multi-utf8-string container type to vector.

• Added vector reserve in pre-read hooks for Seq-table and Dense-seg.

• Added reserving pre-read hooks for Seq-graph data.

• Reserve Byte-graph vector too for text formats.

• CSeq_id::IdentifyAccession -- Implement or improve recognition for more prefixes (FN-FO, GB, HV-HX, JJ-JU, KA-KZ, and LA-LJ) and mixed-in EMBL TPA protein accessions.

• Rework publication labeling logic, designating the historic implementation as version 1 and adding a version 2 corresponding to GenBank/GenPept REFERENCE JOURNAL fields and the like. One difference (among many!) between version 1 and 2 labels is that the latter generally leave off item-specific author and title information, which would appear in neighboring flat-file fields.

• CFastaReader -- Rework split sequence packing to support optionally yielding gap literals as appropriate, potentially making for considerable savings (particularly for nucleotide data). NB: Because this feature changes the FASTA reader's traditional interpretation of data, enabling it requires supplying the new fLetterGaps flag. Add another avenue for space saving in the form of a new fNoUserObjs flag, which disables storage of raw deflines.

Bug fixes:

• Fixed CTextseq_id::Compare() for STL requirements.

• Use MT-safe way to get CSeq_id from CSeq_id_Handle.

New Development:

• Object writers and readers: Added support for the VCF and GVF files formats.

Improvements of existing object readers:

• Wiggle: Added support for input files containing multiple tracks.

• Wiggle: Added support for input spanning multiple chromosomes.

• Bed: Chunked down huge (more than 100,000 records) tracks into a sequence of manageably sized (at most 100,000 features) feature tables.

• Bed: Recognized lines beginning with “#” as comment lines.

• Newick: Made node labels in Newick tree input optional.

Improvement of existing object writers:

• GFF3: Added support for writing alignment files.

• GFF3: Added support for a wide variety of ASN.1 attributes to be emitted as GFF3 attributes.

• GFF3: Added support for escaping characters that are considered special by the GFF3 spec.

• GFF3: Added support for source features.

• GFF3: Added support for assembly identifying, sequence region and species pragmas.

• GFF3: Added support for SOFA vocabulary.

• GFF3: Added support for circular sequences.

• GFF3: Added full headers identifying GFF3 revision spec and generating agent.

• Wiggle: Generated more compact byte graphs by pre-analyzing data before final processing.

• GTF: Made output for gene features and exon numbers optional (flag controlled).

Other improvements:

• Enhanced parsing and formatting of RepeatMasker repeat data. The improvements include greater fidelity in preserving information from the native tabular RepeatMasker output, and better standardization in representing repeat features.

• Flatfile generator:

  • Numerous changes to fix various incompatibilities with the C flatfile generator or bugs

• Feature Table Reader:

  • More lenient parsing

  • Better error reporting (e.g. line number)

  • 6th column for features to indicate minus strand

• ASN.1 Record Cleanup:

  • Changed to match C cleanup.

  • Uses autogenerated code, generated by changes to datatool

• CAgpReader:

  • Handle AGP 2.0

• AGP Writer

  • Output AGP 2.0 instead of AGP 1.1

• FASTA reader

  • Can keep track of bad mods and bad residues

• CFastaOstream

  • Now lets you output [key=value] modifiers if requested.  Only a few basic ones currently appear, though.

COBALT

• COBALT now can align two multiple sequence alignments.

• COBALT now can use pre-computed RPS-BLAST alignments saved in BLAST archive format

• Fixed producing alignments in NEXUS format

Improvements:

• Added support for gaps/indels in pairwise, anchored and sparse alignments.

• Redesigned sparse alignment iterator to show more segment types.

New functionality:

• Added API to work with annot zoom tracks.

• Added CSeq_annot_Handle::ReorderFtable().

• Implemented CFeat_CI iterator of features by id/xref.

• Allow cancelling all tasks in prefetch manager.

• Added faster CScope::IsSameBioseq().

Improvements:

• Implemented gap annotation mapping on segset entries.

• Avoid using catch (...). Use guard class for cleanup instead of catch(...).

• Order Object-id.str with numeric value together with Object-id.id.

• Treat Gene-ref.desc as locus if necessary.

• Use gt/lt to indicate completely cut intervals in mapped annotations.

• CTableFieldHandle::TryGet() now returns false instead of throwing an exception.

• Check all mRNA features passing expensive check in GetBestMrnaForCds().

• Try to find mRNA parent even if gene link exists in CFeatTree.

• Keep blob cache size to avoid calling slow list<>::size().

• Make CAnnotCollector::x_MatchLimitObject() non-inlined as it's used in other file.

• Use xref before overlap check in CFeatTree.

• Allow adding chunks to active split info in MT.

• Optimized CTSE_Info::GetAnnotIds().

• Avoid multiple overlap test for multi-id locations in CFeatTree.

• Avoid unnecessary loading split Seq-data via GetInst().

• Heavily redesigned Compare() and TestForOverlap() functions to fix many problems and add more flexibility.

Bug fixes:

• Update annot index before looking for Seq-feat handle.

• Skip empty ranges when looking up for a feature handle.

• Fixed interval matching with circular locations in CHandleRangeMap.

• Fixed IgnoreStrand matching in CHandleRangeMap.

• Fixed loss of features with multi-id and circular locations.

• Fixed self-alignment duplication.

• Fixed alignment duplication on a boundary between chunks.

• Fixed objects::sequence::GetStop() off-by-1 error.

• Do not treat location with null as multi-id.

• Safe-guard against exceptions from TestForOverlap().

• Fixed CMappedFeat::GetLocationId() for mix locations.

• Repeat chunk loading if necessary as they may be added later.

New functionality:

• Allow cache writer without reader.

Improvement:

• Added workaroung to empty skeleton data when it's attached to split-info.

• Use more efficient retrieval of blobs together with versions.

• Added one-time check for ICache version compatibility.

• Added ID2 replies' retry-delay processing.

Bug fixes:

• Avoid slash in cache subkey due to conflict with storing blobs in files.

• Skip zoom-level annotations instead of failing _ASSERTion().

• Fixed loss of chunks' split and blob versions after loading from cache.

• Avoid copying uninitialized iterators.

New functionality:

• Added support for SRZ cfg and graph.

• Added map quality and annot name.

• Added mate pairing getters.

• Added GetFlags() for real BAM files.

• Added configurable SPOT_ID detector.

• Allow independent creation of CSeq_align, CSeq_annot, and short CBioseq.

Improvements:

• Added detailed SRA SDK error reporting.

• Added more CIGAR types.

• Relax alignment size limit.

• Allow direct BAM file name in arguments.

• Convert numeric refseq ids < 1000 to local Seq-id.

• Added CDataLoader::GetPossibleAnnotNames().

• Implemented separate loading of alignments and short sequences.

Bug fixes:

• Fixed skipping entries with low quality.

• Reverse short sequences aligned with minus strand.

• Fixed soft clipping at the end of sequence.

New functionality:

• Implemented SRA and SRZ lookup in readers.

Improvements:

• Added explicit schema registration for XCODE and Windows.

• Separated SRA, SRZ, and BAM paths.

New functionality:

• Implemented cSRA reader for alignments, short reads, coverage graphs, and pileup graphs.

• Added IIdMapper to cSRA reader.

• Added mate read info.

• Added GetShortRead() to get reversed read sequence.

• Added GetMatchAnnot() to get annot with align only.

• Added generation of read quality graph.

• Added generation of reference sequence if necessary.

Improvements:

• Added check for SRA value read success.

• Added query arguments to test applications.

• Added mate ref-id if different.

• Changed local ids to a more efficient ints.

• Explicit mate pair info generation made optional.

• Added Seq-annot.desc for mate match indication.

• Allow SRR accession to be passed to RegisterInObjectManager.

• Added GetIds().

Bug fixes:

• Fixed soft clipping shift.

• Fixed seq-id mapping.

New functionality:

• Added DELTA-BLAST - an new protein sequence alignment tool that constructs a PSSM using the results of a Conserved Domain Database search and searches a sequence database.

• Added a new API CDeltaBlast (see above)

• Added a new command line binary deltablast (see above)

• Added support for bl2seq in blast_formatter

• Expose psiblast command line option to ignore the MSA master sequence

• Added blastdbcp, seedtop, deltablast, and makeprofiledb application

• Added new FSC statistics

• For more details, see BLAST+ 2.2.26 release notes.

Improvements to utilities:

• Multireader: Add support for Newick trees, GVF, VCF and (bare bones) text alignment files.

• Annotwriter: Add support for GVF, and VCF formats.

agp_validate

• AGP version 2.0. Extra checks for chromosome-from-scaffold and scaffold-from-component files (-scaf, -chr).

• It is OK to not include the red-lined commits.

• Added agp/fasta comparison functionality

CHANGES TO COMPILER SUPPORT

• Supported GCC versions now range from 4.0.x (released in 2005) to 4.6.x (the current stable series). In addition, the Toolkit now supports building with LLVM Clang 3.0, at least on Mac OS X.