DIAMOND v2.1.25

• Improved the performance of the cluster, deepclust and linclust workflows.
• Fixed an issue that could cause an out-of-memory error in the clustering workflows during output generation.
• Added the option --ext none to skip seed extension and output seed hits instead.
• Added the option --min-len-ratio to set length ratio filter in mutual coverage searches.
• Improved performance of loading FASTA files, including when used as database.
• Added the option --hamming-dist-boundary-check to clip the hamming distance filter against sequence boundaries.
• Fixed a bug that could cause undefined behaviour in linear-mode searches with mutual coverage thresholds.
• Added the option --keep-temp-files to prevent deletion of temporary files.
• Added support for clustering BLAST databases.
• Temporarily removed the recluster workflow.
• Removed the --qnum-offset and --snum-offset options.
• Fixed an error when using the taxonomic output format with single-block database processing.

DIAMOND v2.1.24

• Fixed a crash in the greedy-vertex-cover workflow.
• Fixed a bug that could cause a crash when using --swipe.
• Added the option --lin-combo to linearize the search comparing against the longest query or target sequence for identical seeds.
• Added the options --qnum-offset and --snum-offset to set non-zero offsets for ordinal ids (qnum and snum output fields).
• Fixed a compiler error on the CHERI architecture.
• Clustering workflows now support --gapopen, --gapextend, --matrix and --custom-matrix.
• Fixed a potential hanging issue.

DIAMOND v2.1.23

• Clustering in parallel across multiple nodes using the --parallel-tmpdir option now outputs accessions (separated into volumed files) instead of ordinal ids if not using --oid-output, applies tantan, motif and seed complexity masking during seeding by default consistent with the regular clustering workflow, and runs faster.
• Added the option --aln-out to output alignments for clustering workflows when using the option --parallel-tmpdir.
• Fixed a potential crash (thanks @jchorl).
• Fixed a potential crash in the clustering workflows.

DIAMOND v2.1.22

• Fixed compiler errors for GCC 4.8.5.
• Improved performance of the computing alignments stage.
• Fixed a bug that caused the --hit-membuf option not to function correctly.
• Fixed potential crashes when using the --parallel-tmpdir option for clustering.
• Added the option --linclust-chunk-size to set database chunk size for clustering using the --parallel-tmpdir option.
• Fixed a potential hanging problem on macOS.
• Fixed a memory leak.

DIAMOND v2.1.21

• Fixed a potential segmentation fault.

DIAMOND v2.1.20

• Fixed a bug that could cause filtering by taxonomy to malfunction for BLAST databases (present since v2.1.18).
• Fixed a memory leak (present since v2.1.19).
• Fixed a potential deadlock.
• Fixed a potential compilation error on arm64 systems.
• Fixed an error when using BLAST databases that were read-only.
• Fixed missing invocation information and a missing newline when using --header verbose.

DIAMOND v2.1.19

• Improved performance of the searching alignments stage and fixed a potential performance issue present since v2.1.15.
• Fixed search paths for SQLite3 on Ubuntu.
• Fixed a crash when using the --parallel-tmpdir option for clustering.

DIAMOND v2.1.18

• Improved performance when using BLAST databases.
• BLAST format is now the recommended database format for best performance.
• Fixed missing sequence ids for some BLAST databases.
• Fixed CMake 2.8 errors.
• Fixed compiler errors for GCC 4.8.5.
• Added the --load-threads option to set number of CPU threads for file I/O.
• Added the --minichunk option to set chunk size for file I/O.
• Fixed a race condition (present since v2.1.15).
• Fixed a memory leak (present since v2.1.15).
• Added deduplication for the staxids output field for BLAST databases.

DIAMOND v2.1.17

• Added (tabular/json) output fields sRANK to print taxonomy nodes of the given rank associated with the subject sequence, where RANK can be any rank in the NCBI taxonomy, e.g. sdomain, skingdom, sphylum, sorder, sgenus, sspecies, etc.
• Using BLAST databases is now universally supported (including taxonomy features), without depending on the NCBI toolkit.
• SQLite3 is now a required dependency for compilation.
• Added line breaks for the getseq workflow at 80 letters per line.
• Removed the BLAST_LIBRARY_DIR and BLAST_INCLUDE_DIR CMake options.
• Fixed a Missing taxdb error message when using BLAST databases.
• When filtering a database by taxonomy, the letter count for computing e-values will now correctly refer to the filtered subset.
• Removed support for BLAST alias databases filtered by GI numbers and filtered using a binary-formatted seqid list (.bsl file extension).
• Added support for soft masking using the SEG algorithm (--soft-masking seg).
• Added the output field hspnum to print 0-based HSP number within a target to the tabular and JSON output format.
• Added the output field normalized_bitscore to print the bitscore normalized by the maximum of the self alignment scores of query and target sequence to the tabular and JSON output format.
• Added the output field normalized_nident to print the number of identical residues normalized by the maximum of the lengths of query and target sequence to the tabular and JSON output format.
• Fixed the --swipe option not producing any output.

DIAMOND v2.1.16

• Fixed an architecture-dependent inconsistency (present since v2.1.15) of tantan repeat masking.
• Fixed a crash occuring when clustering at 90% identity or above.
• Fixed a potential crash.
• Added the option --hit-membuf for memory buffering of seed hits.
• Fixed a potential crash when using --comp-based-stats 4.
• Fixed an issue that caused line breaks in quality string of FASTQ files to be copied to the output file.