v3.4.1

Changed

• Refactor SIMD code to reduce number of required registers, and improve SSE2 performance.
• Refactor Prodigal initialization functions into sparse initializer code to reduce library size.

v3.4.0

Added

• strict argument to Gene.translate to control translation of ambiguous codons with unambiguous translation (#54).
• strict_translation argument to Genes.write_genbank and Genes.write_translation.
• Support for translation tables 26 to 33 in Gene.translate.
• Support for translation tables 26, 29, 30, 32 and 33 in GeneFinder.train.
• Genes.score property to count the total score of all extracted genes.
• full_id parameter to Genes.write_gff, Genes.write_translation and Genes.write_genes to control the ID field written for each gene (#53).

Changed

• Gene.translate now raises a warning when called with a translation table incompatible with the training info.

Fixed

• Bug in code for masking trailing nucleotides (#55).

v3.3.0

Added

• CLI option to disable translation of stop codons (#51, by @zclaas).

Changed

• Scorer internal API to separate connection scoring and overlap disentangling.

Fixed

• Bug with computation of minimum node in connection scoring loop (hyattpd/Prodigal#108).
• Out-of-bounds sequence access in _shine_dalgarno_exact and _shine_dalgarno_mm methods of Sequence.
• Memory leak in Nodes.__setstate__ caused by incorrect reallocation.

v3.2.2

Fixed

• Always mark SSE2 support on x86-64 CPUs independently of archspec-detected features (#49).

v3.2.1

Added

• Option to change argument parser in pyrodigal.cli.main.

v3.2.0

Added

• AVX-512 implementation of the SIMD pre-filter.
• Additional support for reading lz4 and xz and zstd-compressed input in the CLI.
• Option to change gene finder type in pyrodigal.cli.main.

v3.1.1

Fixed

• Incorrect unpickling of GeneFinder causing crashes with multiprocessing (#46).

v3.1.0

Added

• Support for Python 3.12.
• min_mask argument to GeneFinder to control the minimum lenght of masked regions on mask=True.

v3.0.1

Fixed

• Genes.write_scores and Genes.write_gff crashing on empty Genes (#44).

v3.0.0

Added

• MetagenomicBins collection to store a dense array of MetagenomicBin objects.
• metagenomic_bins keyword argument to GeneFinder allowing to control which models are used when running gene finding in meta mode (#24).
• metagenomic_bin attribute to Genes referencing the metagenomic model with which the genes were predicted, if in meta mode.
• Additional TrainingInfo properties (missing_motif_weight, coding_statistics).
• Setters for all remaining TrainingInfo properties.
• Proper TrainingInfo constructor with configuration option for all attributes.
• TrainingInfo.to_dict method to extract all parameters from a TrainingInfo.
• Genes.write_genbank method to write a GenBank record with all predicted genes from a sequence.
• include_stop flag to Gene.translate and Genes.write_translations to allow excluding the stop codon from the translated sequence.
• include_translation_table flag to Genes.write_gff to include the translation table to the GFF attributes of each gene.
• gbk output format to the Pyrodigal CLI.
• Sequence.unknown property exposing the number of unknown nucleotides in the sequence.
• Sequence.start_probability and Sequence.stop_probability to estimate the probability of encountering a start and a stop codon based on the GC%.

Fixed

• Genes.write_gff not properly reporting the number of bytes written.
• Merge several nogil sections in Sequence constructor.
• Several Cython functions missing a noexcept qualifier.

Changed

• BREAKING: Rename OrfFinder to GeneFinder for consistency.
• BREAKING: Use memoryview to expose all TrainingInfo attributes instead manually building lists or tuples.
• Reorganize memory management of the built-in metagenomic models.
• Make the internal Cython model public (pyrodigal.lib) to allow importing the underlying classes in other Cython projects.
• Use typing.Literal for allowed translation table values in pyrodigal.lib annotations
• Cache intermediate log-odds in Nodes._raw_coding_score to reduce calls to pow and log functions.
• Inline connection scoring functions to reduce function call overhead.
• Reorganize struct _node fields to reduce size in memory.
• Make GeneFinder.find_genes and GeneFinder.train reserve memory for the Nodes based on the GC% of the input sequence.
• Avoid storing temporary results in the generic implementation of ConnectionScorer.compute_skippable.
• Use Cython freelist for allocating Node, Gene, MetagenomicBin and Mask.
• Increase minimum allocation for Genes and Nodes to reduce early reallocations.

Removed

• BREAKING: metagenomic_bin attribute of TrainingInfo.