# Changes in this Release

This release includes major improvements to robustness, compressed I/O, and performance, along with multiple bug fixes and cross-platform updates.

Fixes

• Fixed handling of multiple inputs with a single output
• Fixed Windows compatibility issues
• Fixed bgzip output bugs
• Fixed parsing issues in R .rd interface (length/quality layout)
• Improved handling of empty and edge-case datasets

Compressed I/O

• Reworked readTableCompressed() with stricter validation
• Added checks for corrupted or truncated files
• Enforced consistency between header, payload, and record counts
• Added validation of decompressed payload size
• Improved error handling and diagnostics
• Standardized size handling using uint64_t / size_t
• Ensured consistent serialization/deserialization layout
• Fixed potential out-of-range access

Performance & memory

• Reduced memory allocations and reallocations
• Reused BAM records in output
• Freed batch memory more aggressively
• Added elastic blocking queue to reduce memory overhead
• Optimized sampling input loop
• Precomputed frequently used values

I/O and format handling

• Added bgzip support
• Avoided automatic bgzip output (more explicit behavior)
• Switched FASTA/FASTQ compressed input handling to htslib
• Removed redundant zlib dependency for input

Output handling

• Fixed stream handling with cifiCombinations_flag
• Corrected internal stream counting (numStreams)
• Improved batch ordering and buffering logic

R interface

• Added optional debug mode
• Improved robustness of .rd parsing
• Fixed indexing of length and quality fields to use the new compression format

Build / CI

• Fixed Ubuntu workflow (dependency installation)
• Improved GitHub Actions reliability

Changes in this Release

New Features

• Added full support for PacBio CiFi read preprocessing, including CiFi-specific metrics, biologically precise cut-site handling, examples, and test datasets.
• Report visualization improvements: replaced scatterplot with a 2D density plot, dramatically improved speed using data.table, cowplot, and ggrastr, and removed contours/jitter when unnecessary.

Performance & Parallelization

• Nearly full multithreaded input support with MPMC ring buffer management and user-defined buffer sizes for constant memory allocation.
• Similar implementation for output streaming for improved throughput and reduced allocations.
• Added parallel file processing with deterministic sorted output; improved handling of multiple and mixed input files.

Bug Fixes

• Fixed multiple multi-file and mixed-format issues, including subtle sorting bugs and rd-specific output handling.
• Corrected MD5 computation (including MP-md5 stream closure) and prevented empty reads from being generated.
• Fixed average per-base quality calculation, reproducibility with --sample/--random-seed, and per-thread log clearing.
• Avoid uninitialized bytes in rd file due to padding, to prevent file size being different between runs due to compression. NOTE: The R interface was not fixed in this release, so the reports will not be generated correctly when using this version of the .rd files. Use v0.0.9 instead.

Changes in this Release:

• allowed filters directly on .rd files in all modes
• updated the usage hints (--help)
• allowed multiple output combinations in the same run (e.g. --sequence-report and -o out.bam)
• static linkage for Linux distribution (rdeval.v0.0.6-linux-static.zip)
• fixed minor bugs

Changes in this Release are the same v0.0.7 but this version was not used to avoid conda caching during update

Changes in this Release

Changes in this Release

Changes in this Release
Added:

• homopolymer compression
• .rd file summaries with md5sum
• enhanced filtering capabilities
• support for bam/cram IO/conversion

Fixed/refactored:

• -s option outputs with hashmap

Changes in this Release

rdeval first official release.