[2.1.11]
- Improved the performance and sensitivity of the `cluster`, `deepclust` and
  `linclust` workflows.
- The `--faster` mode will by default use a minimizer sketch of fixed size
  per sequence instead of window-based minimizers.
- Added the option `--sketch-size` to enable seeding using a minimizer
  sketch of the given size per sequence.
- Cascaded clustering and iterated search will by default use the `--fast`
  mode with linearization in the second round.
- The `--round-coverage` parameter is now also applied to uni-directional
  coverage clustering.
- Cluster output files will correctly contain carriage returns on Windows.
- Fixed generation of the Docker container against the latest version of the
  NCBI toolkit.
- Fixed a bug that caused target coordinates not to be reported correctly
  in the tabular format in frameshift alignment mode.
- Added the options `--ungapped-evalue` and `--ungapped-evalue-short` to set
  e-value thresholds for the ungapped hit filter.
- Linearization of search or clustering rounds is limited to seed shape
  configurations of weight >= 10.
- Fixed an issue that could cause an `array size overflow` error when using
  very large `.dmnd` databases with taxonomic annotation.
- Fixed a bug that caused query letters to be printed as `ARND` instead of
  `ACGT` in the `view` workflow.
- Fixed a bug that caused using paired end input files to malfunction with
  an error message.
- Fixed a bug that could produce clustering errors when clustering at
  sequence identities >= 50% and processing the database in multiple super
  blocks.
- Fixed a bug that could cause a crash in global ranking mode.
- Accession parsing rules applied to database sequence accessions for the
  purpose of matching them to accessions in the taxonomy mapping file are now
  by default also applied to the accessions in the mapping file (disable using
  `--no-parse-seqids`).
- Fixed an issue that could cause increased memory use in the hash join stage.
- Added support for FASTA headers containing multiple sequence IDs separated
  by blank spaces (so far only the `\1` character was supported as a
  separator).
- Fixed an issue that could cause hanging or crashes in the `Computing
  alignments` stage.
- `--linsearch` can now be used in conjunction with `--iterate`.
- Fixed a compiler error for GCC 4.8.5.
- Fixed a compiler error on Solaris.
- Fixed compiler errors on systems that do not support the sysinfo function.
- Fixed `Bus error` occuring on Sparc systems.
- Compilation on Sparc systems can be performed without setting `-DX86=OFF`.
- Fixed two issues that could cause increased memory use in the computing
  alignments stage.
- Fixed a bug that caused superfluous quote characters in the JSON output
  format.
- Linear search modes will by default use full-matrix extension.
- Fixed an issue that could cause reduced performance in the masking sequences
  stage.
- Fixed a bug that could cause a crash when using mutual coverage thresholds
  in blastx mode.
- Fixed a bug that could cause a crash when the `--include-lineage` option
  was used.
- When reading protein sequences that unexpectedly only contain DNA letters,
  an error message is only produced if the first 10 sequences in the input
  file all exhibit the problem.
- Fixed a bug that caused setting `--top 100` not to function correctly.
- Fixed a bug that caused target coordinates not to be reported correctly in
  the output of the `realign` workflow.
- Fixed a bug that did not permit using the '--memory-limit/-M' option for
  the `realign` workflow.
- Fixed an issue that could cause non-deterministic output in frameshift
  alignment mode.
- Fixed a bug that could cause a crash when using the XML output format in
  the `view` workflow.
- Added the sensitivity modes `--linclust-20` and `--linclust-40`.
- Fixed an issue that could cause non-deterministic output for identically-scoring
  HSPs in the same target.
- Disabled the default use of increased coverage and identity cutoffs in
  earlier clustering rounds.
- Optimized the performance of the extension stage when coverage or approximate
  identity filters are used.
- Optimized the performance of the extension stage when not using output fields
  that require alignment traceback.
- Fixed an issue that could cause an incorrect order of cascaded clustering
  rounds.

[2.1.10]
- Fixed a bug that could cause a crash when using a bi-directional coverage
  cutoff in query-indexed mode.
- Fixed a bug that caused the `--include-lineage` option to malfunction for
  targets with no taxonomic assignment available.

[2.1.9]
- Corrected the prefix of the query length field for the SAM format.
- Added the size modifiers 'T', 'M' and 'K' for the `--memory-limit`/`-M`
  option.
- Added the option `--mutual-cover` to cluster sequences by mutual coverage
  percentage of the cluster representative and member sequence.
- Added the option `--symmetric` for computing greedy vertex cover with
  symmetric edges.
- Fixed an issue that caused the `--approx-id` option and the `approx_pident`
  output field not to work correctly when using the `--anchored-swipe`
  option.
- Added the option `--no-reassign` to prevent reassignment to closest
  representative for the greedy vertex cover and clustering workflows.
- Added the option `--connected-component-depth` to activate clustering
  of connected components at a given maximum depth for the greedy vertex
  cover and the clustering workflows.
- Fixed a compiler error for Clang v17.
- Improved search performance when searching with mutual coverage threshold
  by filtering for sequence length ratio.
- Added the sensitivity mode `--shapes-30x10` with sensitivity approximately
  equivalent to `--mid-sensitive`.
- Added the options `--round-coverage` and `--round-approx-id` to set per
  round cutoffs for cascaded clustering.
- The CMake switch `-DKEEP_TARGET_ID` is now obsolete and the corresponding
  function is always available.
- Added the option `--include-lineage` to the taxonomic classification format
  to include taxonomic lineage in the output.
- Added native support for the ARM NEON instruction set (contributed by
  Martin Larralde).

[2.1.8]
- Fixed an issue that could cause reduced performance when running in
  query-indexed mode.
- Added support for the JSON output format (option `-f json-flat`).
- Added the option `--sam-query-len` to output query length in SAM format.

[2.1.7]
- Fixed a bug that caused taxonomy names not to be loaded correctly for the
  `makedb` workflow.
- Fixed a bug that caused a crash when using the `--target-indexed` option.
- Fixed an error when using the `--tmpdir` option for the makedb workflow.
- Added a warning message when sequence accessions are shortened due to parsing
  rules for the `makedb` workflow.
- Added the option `--no-parse-seqids` to disable parsing of sequence
  accessions.
- Changed the command line help to print options separated by command.
- Fixed an issue that the `--ignore-warnings` option could not be used for
  the `makedb` workflow.

[2.1.6]
- Fixed compatibility issues on older systems without support for AVX2.
- Fixed linker errors when compiled with `-DX86=OFF`.
- Fixed a compiler error on macOS systems.
- Fixed a bug that could cause missing tags in the XML output format and
  unaligned queries not to be reported correctly.
- Fixed a bug that caused the PAF output format not to work correctly.

[2.1.5]
- Disabled the use of frequency based seed masking when using the linear-time
  search feature with respect to the targets.
- Fixed a bug that caused a `Database file is not a BLAST database` error message
  for the `prepdb` workflow.
- Fixed a bug that caused a segmentation fault when using BLAST databases.
- Added line numbers for error messages when reading taxonomy mapping files.
- Fixed a bug that could cause a crash when using the `greedy-vertex-cover`
  workflow without the `--out` and `--centroid-out` options.
- Fixed a bug that caused the `greedy-vertex-cover` workflow to only produce a
  trivial clustering.
- Fixed a bug that caused the last codon of the -2 reading frame to be translated
  incorrectly.
- Reduced the memory use of the clustering workflow.
- Updated the bundled NCBI toolkit to the latest version.

[2.1.4]
- Leading spaces are now trimmed and tabulator characters escaped as `\t`
  in sequence titles, and a warning message is produced.
- Blank sequence titles are now replaced by `N/A`, and a warning message is
  produced.
- Fixed a bug that could cause a `Traceback error` in certain cases.
- Fixed a bug that caused the `qlen` and `score` output fields not to be reported
  correctly for the `realign` workflow.
- Added an error message when using unsupported output fields for the `realign`
  workflow.
- Fixed an issue that could cause a `Missing fields in input line` error when
  clustering.
- Optimized the performance of the `linclust` workflow.
- Reduced the memory use of the clustering workflow.
- Fixed a bug that caused using standard input as the query not to work.

[2.1.3]
- Fixed compiler errors for GCC 4.8.
- Fixed a GCC compiler error.
- Fixed a segfault issue occuring when compiled using GCC 12 on ARM64 systems.
- Fixed an issue that caused missing support for AVX2.

[2.1.2]
- The iterated search mode (option `--iterate`) now uses a linear-time feature as
  the first search round.
- Added the `linclust` command to cluster using only a single linear-time search
  round.
- Fixed compiler errors on macOS.
- Fixed a bug that caused invalid alignment traceback output for the DAA `view`
  workflow.
- Added the `merge-daa` workflow to merge DAA files.
- Fixed an error when using the `--max-target-seqs/-k` option for the DAA `view`
  workflow.
- Removed AVX2 support from the Windows release binary to ensure compatibility
  with older systems.
- Permitted the `--ignore-warnings` option for the `cluster` and `deepclust`
  workflows.
- Use unlinked temporary files for database blocks in clustering workflows.
- Fixed a bug that could cause invalid results when using a clustering step with
  linearization as the final round in combination with database processing in
  multiple super blocks.
- The `--lin-stage1` option can now be used without compilation using the
  `-DEXTRA=ON` cmake option.
- Added the option to specify the `_lin` suffix for sensitivity keywords for the
  `--iterate` option to activate linear-time feature.
- Added the option `--linsearch` to activate linear-time feature for the search
  workflows.
- Fixed a bug that caused the `ppos` and `positive` output fields not to work
  for the `realign` workflow.
- Fixed an issue that caused motif masking not to work when compiled with link
  time optimization.

[2.1.1]
- Fixed compilation errors on non-x86 systems and for the clang compiler.
- Fixed an error message when running the `recluster` workflow.
- Fixed a bug that could cause an `invalid varint encoding` error when using the
  DAA format.
- Fixed a bug that could cause corrupted DAA output.
- Fixed a bug that caused an error in the `view` workflow.
- Adjusted the hit culling heuristic of the frameshift alignment mode to be less
  aggressive.

[2.1.0]
- Added the `cluster` workflow to cluster protein sequences.
- Added the `realign` workflow to generate clustering output.
- Added the `recluster` workflow to correct errors in clusterings.
- Added the `reassign` workflow to reassign cluster members to their closest centroid.
- Added the option `-M/--memory-limit` to set a memory limit for clustering workflows.
- Added the `--approx-id` option to filter alignments by approximate sequence identity
  and to set an approximate sequence identity threshold for clustering.
- Added the `--member-cover` option to set the coverage threshold of the cluster
  member sequence.
- Added the `--cluster-steps` option to set steps for cascaded clustering.
- Added the `--clusters` option to specify clustering input file.
- The `blastx` mode will now mask any open reading frame below the minimum required
  length as specified by `--min-orf`.
- The `blastx` mode will only count unmasked letters towards the block size.
- Fixed a bug that caused an error when using the global ranking mode.
- Added the fast mode as the first round in iterative searches.
- Fixed a bug that caused the program not to function on systems without support
  for SSE4.1.
- Improved multi-threaded load balancing of gapped extension computations.
- Improved performance of seed extension stage when HSP filter settings are used.
- Added the option `--soft-masking` with possible values `0` and `tantan` to permit
  soft-masking using the tantan algorithm.
- Fixed a bug that could cause an `inflate error` in multiprocessing mode.
- Added the option `--swipe` to compute full Smith Waterman alignments of all
  queries against all targets.
- Added the sensitivity mode `--faster`.
- Added the output fields `approx_pident` and `corrected_bitscore` to the tabular
  format.
- Added the `--lin-stage1` option to linearize comparisons in the seeding stage
  by only considering hits against the longest query sequence for identical seeds
  (only supported when compiled with `-DEXTRA=ON`).
- Added the `--kmer-ranking` option to rank sequences when `--lin-stage1` is used
  (only supported when compiled with `-DKEEP_TARGET_ID=ON`).
- Added the option `--no-block-size-limit` to deactivate upper limits for the block
  size when the `--memory-limit` option is used.
- Added the `greedy-vertex-cover` workflow to compute clustering based on
  alignments.
- Added the `--edge-format` option to set edge format for greedy vertex cover.
- Added the `--edges` option to set input file for greedy vertex cover.
- Added the `--centroid-out` option to output centroid sequences for greedy
  vertex cover.
- Added the `--unaligned-targets` option to generate an output file of unaligned
  targets.
- Fixed an issue that failed compilation using the Intel Compiler.
- Fixed an issue that could cause a segmentation fault in rare cases.
- The `--header` option can now be used with the parameter `simple` to enable
  simple headers for the tabular format, or without a parameter to enable
  headers for the clustering format.
- Added the option `--mp-self` to optimize self-alignment in multiprocessing
  mode.
- Added the option `--query-or-subject-cover` to report alignments if the query
  or the subject cover (or both) are above the given threshold.
- Removed support for the `--comp-based-stats 2` option (now equivalent to
  `--comp-based-stats 3`).
- Removed hit culling in case of overlapping target ranges in frameshift
  alignment mode.
- Added the option `--anchored-swipe` to activate anchored SWIPE extension.

[2.0.15]
- Fixed a bug (present since v2.0.12) that caused the `diamond view` workflow to
  report a zero bit score for all alignments.

[2.0.14]
- Fixed a compiler error on Linux systems that do not define `_SC_LEVEL3_CACHE_SIZE`.
- Fixed an error when using `--unal 1` with the `cigar` output field.
- Fixed an `illegal instruction` error on systems that did not support AVX2.
- Fixed a bug (present since v2.0.12) that could cause an error or suboptimal
  alignments when HSP filter settings were used.

[2.0.13]
- Fixed a bug that caused invalid bit scores in frameshift alignment mode.

[2.0.12]
- Fixed an error when using HSP filter settings together with a BLAST database.
- Optimized the performance of alignment traceback.
- A non-default setting of `--max-hsps` will now recompute a full-matrix Smith Waterman
  alignment with the ranges of the known HSPs masked in the target.
- A non-default setting for `--max-hsps` can now be used together with `--ext full`.
- The sensitivity levels used for iterated searches can now be manually set by
  using a space-separated list after the `--iterate` option.
- Seeds are masked based on complexity instead of frequency by default.
- Added the option `--seed-cut` to set a complexity cutoff for indexed seeds.
- Added the option `--freq-masking` to enable masking seeds based on frequency.
- The fast, default, mid-sensitive and sensitive modes will by default softmask
  a fixed set of highly abundant sequence motifs.
- Added the option `--motif-masking (0,1)` to enable or disable motif masking.
- Added the option `--masking seg` to enable SEG masking of target sequences (BLAST
  default) instead of tantan masking.
- Fixed a bug that caused the `full_sseq` output field to contain invalid information
  or to produce an error when using a BLAST database.
- Changed composition based statistics to use BLOSUM62 background frequencies.
- Fixed the zstd dependency in the Dockerfile.
- Added support for gap letters in BLAST databases.
- Fixed a bug that caused the `--custom-matrix` option not to function correctly.
- Changed the overlap for merging adjoining bands to >0.0.
- Use more moderate filtering of HPSs in the chaining stage.

[2.0.11]
- Fixed a bug that could cause invalid output when using `--masking 0` combined with
  the global ranking mode.
- Enabled lazy repeat masking in the query-indexed and contiguous seed modes when
  using global ranking.
- Added detection of cache size to auto-enable query-indexed mode.

[2.0.10]
- Using BLAST databases now requires a preprocessing step using the command `prepdb`.
- Improved performance of searching small query files.
- Added "iterative" search mode (option `--iterate`) to search the query dataset with
  increasing sensitivity, only searching queries at the target sensitivity that fail
  to align at a lower sensitivity search.
- Added the "global ranking" mode (option `-g`) to set a limit on the number of
  Smith Waterman extensions per query, with the target sequences ranked by their
  ungapped extension scores.
- Added the `--fast` sensitivity mode that is faster and less sensitive than the
  default mode.
- Reduced the time for loading target sequences from BLAST databases.
- Added the contiguous-seed mode (option `--algo ctg`) to improve performance for
  small query files.
- Added support for using `--comp-based-stats (3,4)` in combination with `--ext full`.
- Fixed a bug that could cause a `Traceback error` when using `--comp-based-stats (3,4)`
  in rare cases.
- Changed the `full_sseq` output field to always contain unmasked sequences.
- Fixed an issue that could cause target output order to be nondeterministic in case
  of identically scoring hits.
- Added support for reading zstd-compressed input files (auto-detected) and writing
  zstd-compressed output files (option `--compress zstd`) (requires compilation
  using `cmake -DWITH_ZSTD=ON`).
- Compilation with BLAST database support requires the zstd library.
- Added error message when reading protein sequences from FASTA files that only
  contain DNA letters (can be disabled using `--ignore-warnings`).

[2.0.9]
- Reduced the memory use of database building with taxonomy mapping.
- Removed the limitation of sequence accession length.
- Fixed a bug that could cause using a BLAST database not to function correctly.
- Added support for using BLAST alias databases (created by `blastdb_aliastool`).
- Reduced the memory use of the seed hit sorting stage.
- Improved the consistency of results when running in query-indexed mode (`--algo 1`).
- Added the option `--skip-missing-seqids` to ignore cases of missing sequences
  in the database when using the `--seqidlist` option.
- The `--min-orf` parameter now defaults to 1 in frameshift alignment mode.

[2.0.8]
- Added support for using BLAST database files instead of the Diamond-formatted
  `.dmnd` files.
- Added the option `--seqidlist` to filter the database by sequence accession (only
  supported for BLAST databases).
- Fixed a bug that caused the `--dbsize` option not to function correctly.
- Added the command `makeidx` and the option `--target-indexed` that provide an
  optimisation specialized for small databases (<10 Mb).
- Added the option `--mp-recover` to recover aborted runs in multiprocessing mode.

[2.0.7]
- Added support for computing full-matrix instead of banded Smith Waterman extensions
  (command line option `--ext full`).
- Added support for the new `prot.accession2taxid.FULL.gz` taxonomy mapping file from
  NCBI.
- Added the option `--gapped-filter-evalue` to set the e-value threshold of the gapped
  filter heuristic.
- Added setting the scores of the mask letter according to BLAST rules when a
  compositionally adjusted matrix is used.
- Changed formatting of e-values to print two decimals instead of one.
- Added the output field `qseq_translated` to print the translation of the aligned part
  of the query sequence.
- Added support for providing two input files to `--query/-q` when running alignment
  in blastx mode.
- Added the output field `full_qseq_mate` to print the sequence of the query's mate
  (enabled when using two query files in blastx mode).
- Fixed a bug that could cause a crash in blastx mode for very long queries.

[2.0.6]
- Changed the computation of expected values to use the method described in Park, Y.,
  Sheetlin, S., Ma, N. et al. New finite-size correction for local alignment score
  distributions. *BMC Res Notes* **5**, 286 (2012).
- Enabled the use of a custom scoring matrix without having to specify the statistical
  parameters (option `--custom-matrix`).
- Added support for compositional matrix adjust as described in Yi-Kuo Yu, Stephen F.
  Altschul, The construction of amino acid substitution matrices for the comparison of
  proteins with non-standard compositions, *Bioinformatics*, Volume 21, Issue 7, 1 April
  2005, Pages 902911. Three additional modes have been added that can be enabled by
  setting `--comp-based-stats (2,3,4)` (*the feature is not enabled by default and does
  not support translated searches at the moment*).
- Fixed a bug that could cause incorrect alignment coordinates, gaps counts and sequence
  identities being reported by `diamond view`.
- Targets are sorted by bit score instead of e-value in the alignment output when the
  `--top` parameter is used.
- Disabled support of custom scoring matrices for the DAA format.
- Fixed a bug that caused the use of a custom scoring matrix not to function correctly.
- Fixed an issue that caused the portable binary not to function on systems that did not
  support AVX.
- Added the option `--no-unlink` to prevent unlinking of temporary files.

[2.0.5]
- Fixed an issue that could cause high memory use in frameshift alignment mode.

[2.0.4]
- Fixed a bug that could cause the `--max-target-seqs/-k`, `--ext-chunk-size` and 
  `--file-buffer-size` options not to function correctly on macOS.

[2.0.3]
- Added a new sensitivity mode that is between the default mode and the sensitive mode
  in sensitivity (option `--mid-sensitive`).
- Added counters for total number of reference blocks, shapes and index chunks to
  the status messages.
- Fixed a bug (persisting since v2.0.2) that could cause secondary HSPs within one
  target not to be reported if the `--max-hsps` option was used with a non-default setting.
- Fixed a bug that could cause an invalid error message with regard to the database
  format in certain cases.
- The `--no-self-hits` option is no longer supported in `blastx` mode.
- Changed the semantics of the `--no-self-hits` option to check for equality of both
  sequence and sequence id, independent of the computed alignment.
- The selection of the top hit when using `--top` will respect the identity, coverage
  and no-self-hits filter settings (does not apply when frameshift alignment is
  enabled).
- The inclusion criterion for `--top` is applied to the bit score instead of the raw
  score and is no longer affected by integer rounding (does not apply when frameshift
  alignment is enabled).
- Improved the accuracy of the ranking heuristic.
- Added the options `--ext-chunk-size` and `--no-ranking` to control the ranking
  heuristic.

[2.0.2]
- Fixed a bug (persisting since v2.0.0) that could cause incomplete results in `blastx` mode.
- Reduced the use of temporary disk space.
- Fixed an issue that could cause long runtimes when using the `--taxon-list` option.

[2.0.1]
- Added feature for using the tool in a distributed computing environment.
- Fixed an issue that could cause increased memory usage and runtimes for very long queries.
- Fixed a bug that could cause a crash when using `--comp-based-stats 0`.
- Fixed a bug that could cause a crash for small input files in certain cases.
- Fixed a bug that could cause filtering hits for identity or range cover not to function correctly when using
  the tabular format without traceback being enabled.
- Added warning messages to recommend block size parameters based on system RAM.

[2.0.0]
- Added the sensitivity modes `--very-sensitive` and `--ultra-sensitive`.
- The `--block-size`/`-b` parameter is set to 0.4 and the `--index-chunks`/`-c` parameter is set to 1 by default
  in the new sensitivity modes.
- Improved performance.
- Added the option `--ext` with possible values `banded-fast` and `banded-slow` to adjust band setup for Smith
  Waterman extensions (new default is `banded-fast` for the default and sensitive mode, and `banded-slow` otherwise).
- Added automatic disabling of alignment traceback if not required by the user-defined output fields in tabular
  output format.
- Changed the default value of the `--max-hsps` parameter to 1.
- Changed the default value of the `--freq-sd` parameter from 10 to 20 for the sensitive mode.
- Fixed a compiler error on FreeBSD.

[0.9.36]
- Fixed a bug that could cause `makedb` to produce invalid database files when using taxonomy features.
- Fixed a bug that could cause a crash when running in query-indexed mode.

[0.9.35]
- Fixed a bug in `diamond view` that would cause high memory usage and erroneous output.
- Reduced the use of temporary disk space.
- Fixed a database compatibility issue with big endian architectures.
- Fixed a bug that would cause a crash for query sequences shorter than 5 letters in blastx mode.
- Fixed a bug that would cause a crash when using a FASTA file as database parameter in blastx mode.
- Added support for the following new ranks in the NCBI taxonomy: biotype, clade, forma specialis, genotype, isolate, morph, pathogroup, serogroup, serotype, strain, subvariety.

[0.9.34]
- Fixed a compiler error for native builds.
- Fixed a compiler error for GCC 4.8.
- Fixed a compiler error when support for SSSE3 was enabled without support for SSE4.1.
- Implemented asynchronous loading of seed hits.

[0.9.33]
- Improved performance and sensitivity.
- Increased use of temporary disk space.
- Implemented support for the AVX2 instruction set.
- Fixed a bug on big-endian architectures.
- Fixed bugs for compilers with unsigned char.
- Fixed compiler errors for generic builds.
- Added compatibility of database files between little and big endian architectures.
- Fixed various issues related to `Illegal instruction` errors on macOS.
- Added option `--file-buffer-size` to set the size of the I/O buffers and set the default value to 64 MB.

[0.9.32]
- Fixed a bug that would generate an incorrect count of positive scoring letters in all output formats.
- Fixed a compiler error on macOS.
- Fixed an `illegal instruction` error on macOS.

[0.9.31]
- Improved performance.
- Composition based statistics use integer scoring.
- Option `--quiet` will suppress startup message.
- Added output field `scovhsp` to print the subject coverage per HSP to the tabular format.
- Added option `--culling-overlap` to set the minimum overlap with a higher scoring hit for a hit to be deleted and changed the
  default value from 90% to 50%.
- Added command `diamond test` to run a series of regression tests.
- Fixed an off-by-one error of the query end position in the XML output format.
- (Update 2020/06/08) Due to a bug, since this version DAA files are not backward compatible with previous versions when using frameshift alignment (option `-F`).

[0.9.30]
- Added support for output field `cigar` to the tabular format.
- Changed the maximum repeat period to 50 for tantan masking.
- Changed the tantan masking to ungapped mode.
- Improved the performance of repeat masking.
- Added output fields sskingdoms, skingdoms, and sphylums to print subject super kingdoms, subject kingdoms, and subject phylums.

[0.9.29]
- Fixed a bug that could cause taxonomy features to function incorrectly for databases created by versions 0.9.27 and 0.9.28.

[0.9.28]
- Fixed a bug that could cause alignment score overflows for scores > 65535 in frameshift alignment mode.
- Fixed a clang compiler error.

[0.9.27]
- Improved performance of the seed matching stage.
- Seed frequency counts are computed based on hit seeds.
- Added option `--taxon-exclude` to exclude list of taxon ids from search.
- Compiling from source will no longer perform a native build. Instead, a portable binary that contains code paths
  for multiple architectures will be produced, with dispatch logic that is invoked at runtime.

[0.9.26]
- Fixed a bug that could cause undefined behaviour when using a database file of format version < 2.
- Fixed a compiler error when compiled as generic C++.
- Program will no longer terminate with an error if unlink system call fails.
- Added option `--tantan-minMaskProb` to set minimum repeat probability for tantan masking and changed the default value to 0.9.
- Added option `--tantan-maxRepeatOffset` to set maximum tandem repeat period to consider and changed the default value to 15.
- Added option `--tantan-ungapped` to use tantan in ungapped mode and changed the default to gapped mode.
- Changed score matrix lambda calculation for tantan masking.
- Reference masking is recomputed during alignment runs.

[0.9.25]
- Added support for the `sscinames` output field to print subject scientific names to the tabular output format.
- Fixed a compiler error for GCC 8.2.
- Added option `--stop-match-score` to set the match score of stop codons.
- Fixed a bug that caused the `qqual` output field to not be correctly clipped to the aligned part of the query.
- Added output fields `qseq_gapped` and `sseq_gapped` to the tabular format.
- Raised compiler requirement to GCC 4.8.
- Fixed a bug that caused the final sequence positions to not be printed in the pairwise format.
- Allow using `--min-score` instead of `--top` for the LCA computation of the taxonomy output format.
- Reduced the number of temporary files.
- Added output field `qstrand` to the tabular format.
- Database format version changed to 3.
- Fixed a bug in the `--range-culling` mode that could cause undefined behaviour.

[0.9.24]
- Fixed a compiler error on macOS.
- Added --header option to output header for tabular output format.
- The quality string output in tabular format (qqual field) is clipped to the aligned part of the query.
- Print '*' as quality string if quality values are not available in tabular output format.
- Added field 'full_qqual' to print unclipped query quality values to the tabular format.
- Added field 'full_qseq' to print unclipped query sequence to the tabular format.
- Added support for using the hyphen character '-' to denote the standard input for input file parameters.
- Status messages are written to stderr.
- Fixed a bug that could incorrectly report queries as unaligned in the output of the --un option.
- Added option '--al' to write aligned queries to file.
- Added options '--alfmt' and '--unfmt' to set the format of the aligned/unaligned query file (supported values: fasta, fastq).

[0.9.23]
- Fixed an issue that could cause too high memory usage.
- Added output field 'qqual' to print query FASTQ quality values to the tabular format.
- Changed license to GPL.
- Raised compiler requirement to GCC 4.6.
- Added option to use the DAA output format for diamond view.
- Added support for using a FASTA file as the --db parameter in alignment workflows.
- Added CL (command line) and VN (version) fields to the @PG SAM format header line.
- Fixed a performance issue for very long query sequences.
- Added shortcut --long-reads for --range-culling --top 10 -F 15.

[0.9.22]
- Added output field full_sseq to tabular output format.
- Database sequences that exceed the maximum accession length will no longer cause an error.
- Added support for PAF output format.
- Optimized performance of database taxonomy filtering.

[0.9.21]
- Fixed compiler errors on some systems.

[0.9.20]
- Added Bioconda installation instructions to manual.
- Added official docker release: https://hub.docker.com/r/buchfink/diamond/
- Fixed a bug that could cause corrupted output if compression was activated.
- Fixed an issue that could cause high memory usage by automatic use of the query-indexed algorithm.

[0.9.19]
- Fixed a bug in the --un function to write unaligned reads.
- Fixed an issue in the LCA algorithm that could cause an assignment to a higher node.
- --taxonmap and --taxonnodes parameters now apply exclusively to the makedb command.
- Added option --taxonlist to filter searches by subject taxonomic id.
- Changed database format; rebuilding is required.

[0.9.18]
- Optimized output writing performance.
- Fixed a bug in the XML output format.

[0.9.17]
- Fixed a compiler error on FreeBSD.
- Added option --range-culling.
- Fixed escape sequences in XML output.

[0.9.16]
- Fixed a bug that caused an error for non-SSSE3 builds.

[0.9.15]
- Improved performance of frameshift alignment mode.

[0.9.14]
- Added support for frameshift alignments.

[0.9.13]
- Fixed query positions in pairwise format for translated searches.
- Changed default behaviour of --max-hsps option to report unlimited number of HSPs.

[0.9.12]
- Fixed dbinfo command to be able to read older database formats.
- Adjusted XML format for better compatibility with Blast2Go.
- Fixed a potential error when running multiple instances of Diamond.

[0.9.11]
- Added option --xml-blord-format for alternative-style XML format.
- Fixed a bug that could cause a crash when writing compressed output files.

[0.9.10]
- added --strand option to choose query strand
- added dbinfo command

[0.9.9]
- Added taxonomic classification format.
- Fixed a bug in getseq printing masked residues.
- Fixed parsing of UniRef100_ sequence id prefixes.
- Added support for using the staxids output field in diamond view.

[0.9.8]
- Fixed a compiler error.

[0.9.7]
- Fixed compiler errors.
- Changed XML format to print accessions in the Hit_id and Hit_accession fields.

[0.9.6]
- Fixed compiler errors.

[0.9.5]
- Added support for named pipes.
- Added support for reading input files from stdin.
- Added more elaborate file I/O error messages.

[0.9.4]
- Improved performance.
- Fixed a bug in the query-indexed algorithm.
- Empty sequences are ignored instead of generating an error.

[0.9.3]
- Fixed a bug that could cause hanging.
- Fixed a bug that could cause an error when using the staxids output field and the --unal option.

[0.9.2]
- Fixed a compiler error.
- Improved performance for very small query files.

[0.9.1]
- fixed a performance issue

[0.9.0]
- improved performance
- improved support for alignments with long gaps
- removed SEG masking
- added low complexity masking using tantan
- changed license to AGPL

[0.8.38]
- fixed std::exception error messages
- fixed sequence titles in XML format
- XML and pairwise format contain full length titles by default

[0.8.37]
- fixed a bug that would cause an error message for empty DAA files
- all scoring matrices use the respective default gap penalties from BLAST
- added check for SSSE3 instruction set
- added diamond-sse2 to the binary package
- added staxids field to the tabular format

[0.8.36]
- fixed a compiler error

[0.8.35]
- added a check to detect incomplete database files
- database files will be deleted in case database building fails
- fixed a compiler error on 32 bit systems

[0.8.34]
- fixed a compiler error

[0.8.33]
- modified option --no-self-hits to also require matching sequence titles for filtering of a self hit
- fixed a bug that could cause a crash in the `joining output blocks` stage

[0.8.32]
- improved speed and sensitivity
- fixed an issue that could cause too high memory usage in certain cases

[0.8.31]
- renamed option --run-len to --min-orf
- added compositional score adjustments (option --comp-based-stats (0,1), enabled by default)
- removed --single-domain option and replaced by --max-hsps
- added option --no-self-hits

[0.8.30]
- slightly improved sensitivity
- added option to report unaligned queries (--unal)
- pairwise, XML and SAM format will report unaligned queries by default
- added option to filter alignments by subject cover (--subject-cover)

[0.8.29]
- fixed an issue that could cause a crash when using view on incomplete DAA files

[0.8.28]
- slightly improved sensitivity
- added support for the BLAST pairwise format (option -f 0)

[0.8.27]
- added support for gzip compressed files containing multiple gzip streams

[0.8.26]
- fixed a compiler error
- now compiles as generic C++ code without SSE being available
- added option to write unaligned queries to file (--un)

[0.8.25]
- fixed a bug with the qseq field in the blast tabular format
- added qtitle and btop fields to the blast tabular format
- fixed a bug that could cause a crash when passing a nonexistant input file
- fixed an issue that could cause unexpectedly long runtimes in certain cases

[0.8.24]
- error messages when reading sequence files include line numbers

[0.8.23]
- added option to change the genetic code used for translation in blastx mode

[0.8.22]
- the Hit_id and Hit_accession fields are now filled for the BLAST XML format

[0.8.21]
- fixed an error of 'diamond view' when using a custom scoring matrix

[0.8.20]
- added support for customizing the BLAST tabular output format

[0.8.19]
- ungapped xdrop score and ungapped and gapped filter scores are now set in bits
- added support for custom scoring matrices

[0.8.18]
- fixed a crash bug
- implemented getseq command to retrieve sequences from database files

[0.8.17]
- slightly improved sensitivity of fast and sensitive mode
- added new alignment mode (option --more-sensitive)
- the scoring matrix parameter will accept upper case spelling

[0.8.16]
- optimized performance

[0.8.15]
- fixed a compiler error on Mac
- fixed a crash on some systems

[0.8.14]
- fixed a memory leak

[0.8.13]
- fixed a compiler error for GCC 4.1 and 4.2

[0.8.12]
- changed database format
- block size parameter is set only for the alignment commands
- the program will accept databases created by later versions if the format is compatible
- improved sensitivity of sensitive mode with some loss in performance

[0.8.11]
- fixed a compiler error

[0.8.10]
- added option to directly generate all output formats
- fixed some GCC 6.1 compiler errors

[0.8.9]
- optimized performance
- improved local alignment accuracy
- fixed some fields in the XML format

[0.8.8]
- fixed a clang compiler error

[0.8.7]
- added support for BLAST XML format
- fixed the gzip compressed output option

[0.8.6]
- fixed a problem of the Windows version that could cause errors for larger files
- optimized performance

[0.8.5]
- optimized performance

[0.8.4]
- fixed a problem that could cause very long runtimes for highly repetitive sequences
- optimized performance
- fixed a bug that could cause too high memory usage in blastp mode

[0.8.3]
- fixed a problem that could sometimes cause alignments to be missed
- suppressed reporting of alignments that are enveloped by better alignments

[0.8.2]
- use parameter -k0 to report unlimited number of alignments
- fixed a crash bug on Mac OS X
- fixed a crash when view was used with terminal output

[0.8.1]
- added option to disable auto appending of DAA and DMND file extensions (--no-auto-append)
- fixed some compiler errors for GCC 4.6

[0.8.0]
- fixed a bug that would sometimes cause alignments to be missed
- fixed a bug that would cause alignments on the reverse strand to be missed in blastx mode
- temporary directory now defaults to output directory
- removed dependency on Boost
- fixed a bug that would cause stop codons to be translated into 'X'

[0.7.12]
- fixed header string in SAM output to correctly identify the blastp alignment mode
- added option to filter alignments for a given query cover (option --query-cover)
- added option to print full subject titles in output files (option --salltitles)

[0.7.11]
- added --version switch
- added static build option for CMake build
- fixed a bug that would cause a return code of 1 without an error

[0.7.10]
- --log option works on OS X
- added option to set effective database size (--dbsize)
- added CMake build option
- fixed a Clang compiler warning
- dash characters (-) are ignored (treated as X) in protein sequences

[0.7.9]
- added --single-domain option
- fixed a bug that could cause the program not to use all available threads
- optimized performance

[0.7.8]
- fixed a bug that could produce an incorrect sort order of HSPs

[0.7.7]
- fixed a number formatting error in the SAM output format

[0.7.6]
- fixed a formatting error in CIGAR strings

[0.7.5]
- fixed a compiler error for Clang
- compiles on OS X
- fixed a bug that would not produce the correct sort order for HSPs with e-value 0
- option --run-len defaults to 1 for sequences shorter than 30 letters

[0.7.4]
- removed OpenMP dependency
- using explicit Makefile instead of autoconf
- file extensions can be named explicitly for DAA and database files
- HSPs are grouped by target sequence
- fixed a bug that could produce incorrect output for view
- fixed a bug that could report an incorrect number of alignments
- fixed a compiler error for GCC 4.1.2

[0.7.3]
- changed database format (makedb required)
- optimized multithreading load balancing
- removed a non-determinism in alignment computation
- IUPAC nucleotide ambiguity codes in DNA sequences are accepted and converted into N's
- SEG masking is disabled by default for blastp mode
- fixed a memory leak
- reduced database file sizes

[0.7.2]
- fixed a bug that could create invalid DAA files when using the --id option
- added integrity check of DAA files to view command
- fixed a bug that could cause segmentation fault or file seek errors

[0.7.1]
- fixed a race condition for opening temporary files
- temporary files are deleted on ungraceful termination of the program
- blank lines in FASTA files are tolerated
- fixed a compiler error for GCC 4.4.7
- view command can write to standard output

[0.7.0]
- introduced DAA format
- reduced usage of temporary disk space and memory
- fixed a compiler error for GCC 4.1.2
- fixed a bug that could cause the program to hang when running out of temporary space or memory
- added --forwardonly option to view command