=========================
= SMALT release history =
=========================

Release 0.7.6 (21 March 2014)
=============================

New features:

- For compilation on Intel's Many Integrated Cores
  Architecture compile with ./configure --enable-mic

- Re-organised distributed tests.

- Minor changes allowing to switch between 0-based and 1-based offsets
  for debugging (tracking) and to make vectorised alignment code
  safer.

- Ammended installation instructions with the bambamc library.


Bug fixes:

- Fixed bug due to uninitialised re-used memory. This could result in
  a flipped read orientation of a read (estimated incident rate
  approx. 1 mate in 10^6 Illumina read pairs).

- fixed bug in 'smalt map -p' resulting in the "assertion failed"
  error message. This could occur when a read aligned across two
  successive sequences in the reference file for which the index was
  constructed, and none of the matching segments exceeded the
  threshold of the Smith-Waterman alignment score (set with -m).

- Fixed overcommitted memory in multi-threaded mode. The effect was
  felt when using a large number of threads (> 16)., e.g. on intel MIC
  cards.

- Fixed broken multi-threaded exection on Mac OSX (discovered on OSX
  10.8.5) where the program could fail with 'assertion error' or
  silently produce incomplete output.

=(smalt 0.7.6 21/03/2014 hp3)=


Release 0.7.5 (16 July 2013)
============================

New features:

- Changed buffering in multi-threaded mode to make code run more
  efficient with a high number (>8) of threads. As a consequence,
  the memory footprint increased compared to earlier versions.


Bug fixes:

- Fixed 'assertion failed' error that occurred with -p flag
  and read pairs if no alignment was found for the mate with	
  the higher number of k-mer word hits in the hash index.

=(smalt 0.7.5 15/07/2013 hp3)=


Version 0.7.4.5 (09 July 2013)
==============================

- Fixed error message 'rmap.c:2048 ERROR: unknown error code' which
  could occur with wide insert range boundaries (e.g. -i 80000).

Version 0.7.4.4 (25 June 2013)
==============================

- Fixed broken multi-threaded execution (option -n) on MacOSX (broken
  since version 0.6.2). Program execution stopped prematurely and no
  mappings were output. This was caused by the use of unnamed POSIX
  semaphores (<semaphore.h>) that are not implemented on MacOSX.

- Modified configure script to generate 64-bit binaries on MacOSX Darwin
  on x86_64 architecture.


Version 0.7.4.3 (20 June 2013)
==============================

- Fixed a bug that lead to the program terminating with a "assertion
  failed" error message when one of the mates of a read pair had a
  length shorter than the hashed words, but the other could be
  aligned.


Version 0.7.4.2 (05 June 2013)
==============================

- Removed redundant code, added test data, moved code and repository
  to SoureForge.


Version 0.7.4.1 (22 April 2013)
===============================

- Fixed a bug that led to alignments being missed in output generated
  with the '-d' option. An alignment could be missed if it overlapped
  with another, longer alignment even if that had a lower
  Smith-Waterman score.


Release 0.7.4 (10 April 2013)
=============================

New features:

- added command line option '-S' for the specification of 
  alignment penalty scores. 
 

Bug fixes:

- Fixed SAM input for orphan mates of read pairs.

- Fixed Multiple SAM/BAM lines with the 'secondary alignment' flag
  unset. This occurred when using the '-d' option.

- Fixed spurious bug in CIGAR strings that denote mismatches by the
  letter 'X' (e.g. with '-f smalt:x'). The read length represented by
  the CIGAR string could be 1 bp less then the actual read length if
  reads were mapped against > 512 reference sequences and could be
  aligned across two successive reference sequences in the indexed
  FASTA file. The faulty CIGAR string was issued when the read
  alignment had a mismatch with the last base of the reference
  sequence.

=(smalt 0.7.4 10/04/2013 hp3)=


Release 0.7.3 (15 March 2013)
=============================

New features:

- Added output of CIGAR strings with 'X' for mismatch in SAM and BAM
  output formats with options '-f sam:x', '-f bam:x'

- Added 'NM' tag containing edit distance in SAM/BAM output formats to
  comply with recommendations in SAM spec 1.4. Added @PG line to BAM
  output.

- Updated interface to bambamc library version 0.0.42

Bug fixes:

- Fixed broken BAM output format when reads without base qualities
  (FASTA format) were mapped.

- Fixed bug that scrambled BAM output format.
  (output file was opened twice).

- Fixed bug in BAM ouptut format for paired reads: Bit 0x20 ("SEQ of
  the next segment in the template being reversed") of the FLAG field
  was set according to the 1st rather than the 2nd read of the pair.

- Fixed a bug when paired reads were mapped but only one of the reads
  could be aligned with a Smith-Waterman score above the
  threshold. The bug hat the program stopping with the error message
  "assertion failed".

- Fixed segfault with explicit alignment output (option -a) and long
  PacBio reads.

=(smalt 0.7.3 15/03/2013 hp3)=


Release 0.7.2 (22 February 2013)
================================

New features:

- Added @PG tag to SAM header.

- Overhauled online help options. smalt <task> without further option
  prints short summary of available options.

- Added tests for BAM output format.


Bug fixes:

- Fixed BAM output format which was broken since 0.7.0.
  Reference positions were off  by +1 bp. CIGAR strings
  were incorrect. Output was silent when smalt was compiled
  without the bambamc library.

- Fixed missing SAM header for option -f 'sam' (default).

- Fixed sample task bailing out with an error message when there are
  too few confident mappings and the histogram is therefore empty. Now
  the program finishes without error messages but he output contains
  line 'Histogram of insert sizes is empty'.


=(smalt 0.7.2 22/02/2013 hp3)=


Release 0.7.1 (29 January 2013)
===============================

New features:

- Adopted SAM spec 1.4 for the definition of insert sizes.

- Removed the restriction that the number of reference sequences
  has to be representable by a signed 32-bit integer. The file format
  for the compressed set of reference sequences changed.

- added support for reading gzipped FASTA/FASTQ files.

Bug fixes:

- Fixed inconsistent alignment (CIGAR) strings between mappings on the
  forward strand the reference and alignments on the reverse
  complement strand.  The CIGAR string was inconsistent e.g. for
  indels in homo-polymer runs, where the precises alingment was
  ambiguous.

  This caused problems when calling indels naively directly from the
  mapping output without further processing, e.g. involving multiple
  sequence alignment of the mapped reads in such regions.

- fixed broken '-p' flag for split read mapping.

- Fixed broken 'class' label in cigar output format. If 
  one of the mates was not mapped, the mapped mate was 
  assigned the label 'D' instead of the correct 'S'. 

=(smalt 0.7.1 29/01/2013 hp3)=


Release 0.7.0.1 (05 November 2012)
==================================

- Fixed a number of bugs in verision 0.7.0 affecting indexing of and
  mapping against contigs of an assembly. The index task crashed for
  contigs and hits on the reverse complement strand were not correctly
  collated.

- Fixed crashing check task.

- Progress reports are sent to stderr rather than previously stdout.

=(smalt 0.7.0.1 05/11/2012 hp3)=


Release 0.7.0 (18 October 2012 - withdrawn on 26/10/12)
=======================================================
 
- added support for reading SAM/BAM files (using German
  Tischler's bambam libary and samtools). Added -F option to
  specify input formats.

- Changed default behaviour: Where reads or read pairs have multiple
  equivalent mappings a location is picked at random (as with the
  option '-r 0'). The default output format is now SAM using soft
  clipping. Output written to a file (option -o) in SAM format
  contains the header.

- The option '-r -1' switches off random picking. Reads (pairs) with 
  multiple equivalent mappings are reported as 'not mapped'.

- When mapping paired reads, option '-d 0' now reports all possible
  top ranking pairings.

- Alignments can now be output in BAM format (option '-f bam'). The
  format specification for sam output can be extended by optional
  keywords, e.g. '-f sam:nohead,clip' will omit the header and output
  hard clipped read sequences.

=(smalt 0.7.0 18/10/2012 hp3)=


Release 0.6.4 (25 September 2012)
=================================

- fixed a bug in SAM output format for paired reads. The 'inferred
  insert size' (field 9, ISIZE) was always set to 0 in version 0.6.3.

- fixed broken -a option. Progam bailed out with error message
  "assertion failed" when an unmapped read was encountered.

=(smalt 0.6.4 25/09/2012 hp3)= 


Release 0.6.3 (1 August 2012)
=============================

- Added '-O' flag which preserves the order of the reads in the input
  files for the output when using multiple threads ('-n' option).

- Added GFF2 (Generic Feature Format, version 2) output format with the 
  option '-f gff'.

- Mapping with the option '-d 0' now outputs all possible mappings of
  single reads with the best Smith-Waterman score. This option
  overrides the '-r' option which triggers output of a single mapping
  picked at random.

- Reduced memory footprint for multi-theaded mapping (option -n).

- Reduced the memory footprint when mapping paired reads with
  very long insert sizes (> 100k pb).

- Fixed '-q' option which was not recognised for the 'map' task
  resulting in the error message 'unknown option -q'.

- Fixed a bug in the explicit alignment output (-a option). The
  alignment was scrambled for reads which aligned as the reverse
  complement to the reference.

- Fixed a bug in the ouptput of multiple mappings with identical
  Smith-Waterman scores. Multiple mappings were output even when one of
  those mappings with identical Smith-Waterman scores had a higher
  mapping quality than the rest. This happened when single reads were
  mapped without the '-d' option specified

=(smalt 0.6.3 01/08/2012 hp3)=

Release 0.6.2 (21 May 2012)
===========================

- Fixed long-standing efficiency issues with multi-threaded execution
  (-n option). The maximum number of additional threads that can be
  forked is now 64 (used to be limited to 8).  
  
  WARNING: The order of the reads in the input files is no longer
  preserved for the output. Mappings for read pairs are still output
  in sucessive lines, though.

- Fixed broken 'sample' task. Program would stop with the message
  'command line format error'.

- Fixed a bug affecting mapping to scaffolds. The bug could result in
  missing partial alignments or stop of program execution with the
  error message "ERROR: assertion failed".

=(smalt 0.6.2 21/05/2012 hp3)=


Release 0.6.1 (3 April 2012)
============================

- Corrected reading of hash-index files which was broken in version
  0.6.0. The 'map' task halted with the error message 'error in file
  format' for index files created with certain combinations of k-mer
  lengths and sampling steps by the 'index' task. The error depended
  also on the total length of reference sequences.

- Corrected 'check' task completing with 'end of file' error on
  flawless input files.

=(smalt 0.6.1 03/04/2012 hp3)=


Release 0.6.0 (20 March 2012)
=============================

- Added 'check' and 'sample' tasks. Added '-g' option to the map task
  for using insertion distribution generated with the 'sample' task.

- Ouput in SAM format now includes header lines.

- Fixed '-p' flag for partial mapping of reads which was broken for
  single reads.

- Fixed 'assertion failed' error when attempting to map a read pair
  where the sequences of both mates are shorter than the length k of
  the k-mer words of the index.

=(smalt 0.6.0 20/03/2012 hp3)=


Release 0.5.8 (11 November 2011)
================================

- Added the option '-l pp' for paired read libraries where both
  mates are sequenced on the same strand (e.g. 454 libraries)>

- Fixed segmentation fault error that could occur when multi-threading 
  ('-n' option).

- Fixed a segmentation fault error when the hash index file could not
  be found.

- Fixed program stopping prematurely with the error message 'Overflow
  of Smith-Waterman score' that could occur when aligning very long
  sequences (> 65536 bp).

=(smalt 0.5.8 11/11/2011 hp3)=


Release 0.5.7 (13 Sept 2011)
============================

- Introduced the 'R' label in SSAHA2 and CIGAR output formats obtained
  without the '-r' option for reads that cannot be placed
  unambiguously and until now were labelled 'N'. The label 'N' is now
  restricted to reads for which no mapping location can be found.

- Fixed a bug in the output of paired reads where one mate maps to
  multiple locations and the other mate can be mapped
  unambigously. When mapping without the '-r' option, the unambiguous
  mate was reported as 'unmapped'. 

- Corrected mapping scores for paired reads which have were too
  conservative (from version 0.5.4 onwards).

- Fixed a problem that occurred when there were more than 512
  reference sequences in the index. The program could terminate with
  the error message 'failure' if it encountered reads that had a long
  sequence of nucleotides of unknown type (Ns) and their ends had
  (spurious) alignments with reference sequences.

=(smalt 0.5.7 13/09/2011 hp3)=


Release 0.5.6 (2 Sept 2011)
===========================

- Added '-l' option specifying the type of read pair library used.
  By default it is assumed that reads were produced by the illumina
  paired-end (short insert) library protocol.

- Fixed a bug that caused an unnessecary large number of alignments to
  be carried out for long reads (>~1000 bp). This resulted in 3x speed up
  for long PacBio reads. The sensitivity was also improved.

=(smalt 0.5.6 02/09/2011 hp3)=


Release 0.5.5 (5 July 2011)
===========================

- Fixed core dump that cold occur with read pairs when only a single
  pair was detected wher one mate had a mapping quality score of 0.

=(smalt 0.5.5 05/07/2011 hp3)=


Release 0.5.4 (1 July 2011)
===========================

- Fixed bug in SAM output for mate pairs present in version 0.5.3
  where the insert size was set to 0.

- The '-c' option an now no longer be used unless the '-x' flag is
  also set. It interferes with the heuristics used to filter
  potentially matching segments.

- Fixed a bug introduced with verision 0.5.2 where, on rare occasions
  program execution would stop with the error message 'sequence too
  short to be hashed'.

- Better calibration of mapping quality scores for paired reads.

=(smalt 0.5.4 01/07/2011 hp3)=


Release 0.5.3 (8 June 2011)
===========================
 
- New calibration of mapping scores of paired reads. Mapping scores
  take into account the orientation of the mapped mates. SMALT expects
  paired reads as input as they result from the Illumina paired-end
  (PE) library for short insert lengths (gaps). Reads from Illumina
  mate-pair (MP) libraries with long insert lengths would have to be
  converted to PE format to ensure correct scoring. The same holds for
  paired reads from other sequencing platforms.
  
- Improved sensitivity and accuracy of paired reads at somewhat (~20%)
  decreased speed.

- Output of multi-threaded execution (option '-n') now preserves the
  order of the reads in the input file.

- Fixed a problem with 'mapping category' labels for paired reads in
  CIGAR format.  One of the mates of a pair mapped not in the expected
  orientation could be wrongly labelled as the mate of a 'proper' pair
  ('A' rather than 'C').

- Fixed 'Smith-Waterman score overflow' error that occurred with
  sequences > 255 bp on MacOSX.

=(smalt 0.5.3 08/06/2011 hp3)=


Release 0.5.2 (28 March 2011)
=============================

- Fixed '-r' option for the random picking of on of a set of
  multiple best mappings (broken in version 0.5.1).

- Increased accuracy for paired-end reads.

=(smalt 0.5.2 28/03/2011 hp3)=


Release 0.5.1 (04 March 2011)
=============================

- introduced '-y' option for a threshold in the minimum number of
  identical nucleotides.

- Extended SSAHA2 output format (option -f ssaha2) by one field
  specifying the length of the reference sequence to which the read is
  aligned.

- Changed the parameters for the default paired-read mapping
  strategy. As before, the read with the fewer k-mer hits is aligned
  first. If the mapping quality score is above a certain threshold
  (10), the mate is aligned in a restricted zone around the
  first. Otherwise the mate is aligned indepentently. In this version,
  the insert size distribution is not taken into account for the
  mapping quality or paired reads.

- repaired broken '-p' flag. In the sam output format, the second
  partial alignment now has the 'secondary alignment' bit-flag (0x100)
  of the FLAG field set. In cigar and ssaha output formats such an
  alignment is labelled with 'P'.

- Fixed a few issues with the -m option which was broken, e.g. for '-d
  0' (default).

- -f 'samsoft' was broken for unmapped reads when input files where in
  FASTA rather than in FASTQ format. In this case the field with the
  quality strings was missing in the SAM line.

- Fixed error message ('assertion failed') with FASTA query sequences
  (query sequence without base qualities)

- Fixed error message ('assertion failed') with certain combinations
  of -k and -s parameters when indexing.

=(smalt 0.5.1 04/03/2011 hp3)=


Release 0.5.0 (25 January 2011)
===============================

New features:

- New -r option picks at random one of a set of multiple best
  mappings. This is only active in paired-end mode or if '-d 0' is
  used (the default).

- New -q option sets a threshold of the base qualities. K-mer words
  with nucleotides that have a base quality below this threshold are
  not looked up in the hash index.

- The -m option now specifies a 'hard' absolute threshold of the
  Smith- Waterman (SW) score. If the score of the best alignment of a
  reads falls below the threshold, the read is reported as 'not
  mapped'. Previous versions always reported alignments with the
  overall best SW score even if that score was below the threshold set
  with -m ('soft' threshold).

Bug fixes:

- Fixed 'integer overflow' error that could occur when the total length
  of concatenated reference sequences exceeded 2^32 bp and the number of
  reference sequences exceeded 512.

- Fixed a memory leak with long query sequences (> 1024 nucleotides).

- Fixed a problem with the alignment of very short reads: Program
  execution stopped with the error message "Assertion failed" if the
  read length was less than k + 2*s, where s is the sampling step and
  k the length (indexing parameters '-k' and '-s') of the hashed kmer
  words.
   
=(smalt 0.5.0 25/01/2011 hp3)=


Release 0.4.3 (17 December 2010)
================================

- Fixed a problem with unreliable spurious alignments in version 0.4.2. 
  Now the minimum length of aligned read segments is given by the sum
  of the legth of the hashed kmer words and the sampling step size 
  (indexing parameters '-k' and '-s').

- Fixed a bug with paired reads and an index generated with -s 1 that
  could result in the program stopping with the error "assertion
  failed". This was caused by an integer overflow during the
  calculation of candidate locations for the 2nd aligned mate.

- Fixed 1-nucleotide glitches in the length of the aligned segments of
  the query read (or the CIGAR string). This was a rare event
  that could happen when mapping against a large number of contigs.

=(smalt 0.4.3 17/12/2010 hp3)=


Release 0.4.2 (9 December 2010)
===============================

- Reduced memory footprint for compression of sequences
  while indexing.

- The format of the sequence file (extension *.sma) has changed.
  Reading of the sequence file is backwards compatible with versions
  0.3.x.

- SAM ouput now has the '/1' and '/2' suffices stripped from the read
  names of paired reads (indicate the order of mates). 

  Read names are otherwise not checked by SMALT. This means the
  program won't detect whether the two FASTQ input files (READ_FILE
  and MATE_FILE), which contain the reads of 1st and 2nd mate, were
  swapped. Neither will the program detect whether or not the two
  input files are 'in-sync', i.e. whether the i-th read in READ_FILE
  corresponds to (is the mate of) the i-th read in MATE_FILE. Two
  files could be 'out-of-sync' e.g. because one of the files has an
  additional single read erroneously inserted.

- Output lines printed to standard output are now preceded by a hash
  ('#') if they contain progress reports by the program. This is so
  they easily can be parsed out when alignment coordinates are piped
  to standard output rather than written to a file (with the option
  '-o').

- Fixed core dump that occured when indexing reference sequences with a
  total number of nucleotides exceeding 2^32.

- Fixed core dump occuring with ssaha output format ('-f ssaha' flag).

=(smalt 0.4.2 09/12/2010 hp3)=


Release 0.4.1 (10 November 2010)
===============================

- Accelerated mapping speeds particularly for long (> 500 nt) reads.

- Re-calibrated mapping quality scores for 100bp Illumina reads of the
  human genome.

- Changed the default minimum score (19) to <wordlen> + nskip - 1.

- Fixed the '-w' option which was broken in versions 0.3.x

- Fixed a number of bugs due to unitialised memory. Premature stop of
  execution with error messages or core dumps could occur in
  multi-thread mode (-n) or with the option '-d -1' when aligning
  sequences >~ 1000 nt in length.

=(smalt 0.4.1 10/11/2010 hp3)=


Release 0.4.0 (14 October 2010)
===============================

- Implemented support for k-mer words of up to 20 nucleotides and
  reduced the memory and disk footprint of the hash index for smaller
  genomes. The memory required for the construction of the hash index
  was also reduced.

- The file format for the index (file name extension *.smi)
  changed. Reading of hash indices is backward compatible with
  versions 0.3.x.

- Increased speeds for Sanger capillary reads more than 5-fold over
  versions 0.3.x.

- Fixed a bug in version 0.3.2 that occasionally resulted in the
  program exiting with error messages when the number of reference
  sequences exceeded 256. The bug was limited to query sequences
  (reads) aligning across two reference sequences adjacent in the
  FASTA file from which the hash index was built.

- Fixed a bug in version 0.3.2 which affected the way paired read
  alignments with multiple best mappings were output with the -d 0
  (default) option. Such read pairs should be reported as unmapped,
  but the bug occasionally caused one of the mates being reported as
  mapped to one of the possible locations with a mapping quality of 1.

=(smalt 0.4.0 14/10/2010 hp3)=


Release 0.3.2 (24 August 2010)
==============================

- Fixed broken alignment coordinates when reads aligned across two or
  more successive sequences in the FASTA file of reference sequences.
  This bug occurred when the hash index was constructed for more than
  256 reference sequences.
  
- Fixed broken offsets of aligned segments of the query read which
  occurred predominantly with capillary and 454 reads.

=(smalt 0.3.2 24/08/2010 hp3)=


Release 0.3.1 (23 July 2010)
============================

- Fixed an error message "assertion failed" that occurred with query
  reads with degenerate hits (multiple best hits) if the base call
  qualities where missing (input in FASTA rather than FASTQ format).

- Fixed an error message "could not open file" that occurred in
  multi-threaded mode with the '-o <output file>' option.

- Fixed a racing condition in multi-threaded mode. The maximum number
  of threads for the '-n' option was set to 8.

- Changed the way degenerate hits are output when single reads are
  mapped with the -d 0 (default) option: reads with degenerate hits
  are output as 'unmapped' unless -d <swdiff> with swdiff != 0 is
  used. E.g. -d 1 will output all mapped locations with the same best
  Smith-Waterman score.

=(smalt 0.3.1 23/07/2010 hp3)=


Release 0.3 (06 July 2010)
============================

- First public release of pre-compiled binaries

=(smalt 0.3 06/07/2010 hp3)=