This release fixes two minor issues:

1. Allow a UDB file to be written to a pipe or stdout (issue#599).
2. Correct computation of median cluster size after dereplication (issue#611).

This version incorporates many code improvements, more extensive testing, better documentation and some minor bug fixes. List of changes:

• fix: use-after-free introduced in commitde6c1d8 (Jun 13, 2024),
• fix: (harmless) out-of-bounds memory issue in--derep_prefix (commit8a0a508),
• fix: (harmless) memory leak in--fastx_getseqs --label_field (commita9c4271),
• fix: (harmless) memory leak when using option--userfields (--allpairs_global commit03b95bc;--cluster_* commit2fde547;--search_exact commit45cd56d;--usearch_global commitd83bfee),
• fix: (harmless) valgrind error, use of uninitialized values (commit8bab244), also eliminates a pesky compilation warning,
• change: passing a negative value to--fastq_truncee_rate is now an error (commita120f37),
• change: passing a negative value to--fastq_minqual is now an error (commitff5b0c9),
• change: passing a non-ASCII symbol to--join_padgap or--join_padgapq is now an error (commita708f5b),
• change: when using--gapopen "*" to forbid gap opening, the penalty is now set toINT_MAX (rather than 1,000). This might change alignment results for users who relied on the old behavior (thanks to Denis Filloux, issue#602, commit96e9cf9),
• change: when using command--chimeras_denovo,--tabbedout or--alnout can be the only output files specified (commitd51f0a4),
• change: when using command--chimeras_denovo, option--lengthout is now accepted (commit3f55cc6),
• change: when using command--chimeras_denovo, option--xlength is now accepted (commit0fd346c),
• add: compilation optionGLIBCXX_DEBUG when compiling for debugging (commit5cf4a6c, option activates more runtime checks),
• add: official support for clang 20,
• add: initial support for clang 21, initial support for GCC 15,
• add: experimental support for clang 22,
• improve: more accurate line number when reporting illegal characters in fastq headers (commit539084e),
• improve: more accurate line number when reporting non-ASCII characters in fastq headers (commit98a851e),
• improve: remove checks for unneeded libraries during compilation (commit249bb5d),
• improve: code quality (8,536 clang-tidy warnings eliminated),
• improve: documentation and help messages (issues#604),
• improve: complete refactoring and modernization of the command--fastq_stats,
• improve: command--fastq_stats is now up to twice faster (tested on x86-64),
• improve: extensive test-suites for--fastq_stats and--sff_convert,
• improve: code coverage of our test-suite

Add options--n_mismatch,--fastq_minqual, and--fastq_truncee_rate.

The--n_mismatch option will count N's as mismatches in alignments, which may be useful to get sensible alignments for sequences with lots of N's. By default N's are counted as matches. Both the scoring and the counting of matches are affected.

The new--fastq_minqual option for thefastq_filter andfastx_filter commands will discard sequences with any nucleotide with a quality score below the given value. The new--fastq_truncee_rate option for the same commands will truncate sequences at the first position where the number of expected errors per base is below the given value.

Adjust the window size used for chimera detection down from 64 to 32. The window size was by accident increased from 32 to 64 in version 2.23.0, leading to somewhat fewer chimeras being predicted. In addition, a compiler pragma has been included inalign_simd.cc to further protect the compiler from generating wrong code.

This version is released in order to mitigate a bug that occurs when compiling thealign_simd.cc file on x86_64 systems with the GNU C++ compiler version 9 or later with the-O3 optimization option. It results in incorrect code that may cause bad alignments in some circumstances. So far it has only been seen with theallpairs_global command. We are investigating this issue further, but for now we recommend compiling with the-O2 flag. The README.md file and the Dockerfiles have been updated to reflect this. The binaries released with this version will include this fix.

Fix a segmentation fault during clustering when the set of clusters is empty.
Initial documentation in markdown format available on GitHub Pages.

Fix a segmentation fault that could occur during alignment in version 2.29.0, for example with--uchime_ref. Some improvements to code and documentation.

vsearch 2.29.0 fixes seven bugs (see changelog below), adds initial support for RISC-V architectures, and improves code quality and code testing (1,210 new tests).

Changelog:

• add: experimental support for RISCV64 and other 64-bit little-endian architectures, thanks to Michael R. Crusoe and his fellow Debian developers (PR#566),
• add: official support for clang-19 and gcc 14,
• add: beta support for clang-20,
• remove: unused--output option for command--fastq_stats (issue#572),
• fix: bug in--sintax when selecting the best lineage (only low confidence values below 0.5 were affected) (issue#573),
• fix: out-of-bounds error in--fastq_stats when processing empty reads (issue#571),
• fix: bug in--cut, patterns with multiple cutting sites were not detected (commit4c4f9fa),
• fix: memory error (segmentation fault) when using--derep_id and--strand (issue#565),
• fix:--fastq_join now obeys to--quiet and--log options (commit87f968b),
• fix:--fastq_join quality padding is now also set to Q40 when quality offset is 64 (commitbe0bf9b),
• fix: (partial)--fastq_join's handling of abundance annotations (commitf2bbcb4),
• improve: additional safeguards to validate input values and to make sure that they are within acceptable limits. Changes concern options--abskew (commita530dd8) and--fastq_maxdiffs (commit4b254db),
• improve: code quality (1.3k+ commits, 6k+ clang-tidy warnings eliminated),
• improve: documentation and help messages (issue#568),
• improve: complete refactoring and modernization of a subset of commands (--sortbylength,--sortbysize,--shuffle,--rereplicate,--cut,--fastq_join,--fasta2fastq,--fastq_chars),
• improve: code-coverage of our test-suite for the above-mentioned commands (1,210 new tests, 4,753 in total).

Thesintax command has been improved in several ways in this version of vsearch. Please note that several details of this algorithm is not clearly described in the preprint, and the implementation in vsearch differs from that in usearch.

The former vsearch version did not always choose the most common taxonomic entity over the 100 bootstraps among the database sequences with the highest amount of word similarity to the query. Instead, if several sequences had an equal similarity with the query, the sequence encountered in the earliest bootstrap was chosen. The confidence level was calculated based on this sequence compared to the selected sequences from the other 99 bootstraps. This could lead to a suboptimal choice with a low confidence. In the new version, the most common of the sequences with the highest amount of word similarity across the 100 bootstraps will be selected, and ties will be broken randomly.

Another problem with the old implementation was that if several sequences had the same amount of word similarity, the shortest one in the reference database would be chosen, and if they were equally long, the earliest in the database file would be chosen. A new option calledsintax_random has now been introduced. This option will randomly select one of the sequences with the highest number of shared words with the query, without considering their length or position. This avoids a bias towards shorter reference sequences. This option is strongly recommended and will probably soon be the default.

Furthermore, a ninth taxonomic rank, strain (letter t), is now recognized. The speed of the sintax command has also been significantly improved at least in some cases. Run vsearch with therandseed option and 1 thread to ensure reproducibility of the random choices in the algorithm.

This version fixes theweak_id option and makes searches report weak hits in some cases. It also updates the names of the compression libraries tolibz.so.1 andlibbz2.so.1 on Linux to make them work on common Linux distributions without installing additional packages.README.md has been updated with information about compression libraries on Windows.