- Temporary workaround to issue #8 : Providing
pmd-maskwith an invalid or corrupted fasta index now leads to an uncoverable error. This workaround. - Remove debug and memory profiling artifacts (e.g.
#[inline(never)]) on some performance critical functions.
- Complete genome::Strand documentation + doctests.
- Complete genome::Orientation documentation + doctests.
- Complete genome::coordinate::Position and genome::coordinate::ChrName documentation + doctests.
- Improved logger::Logger documentation
- Complete documentation and doctests for genome::mask::MaskEntry, genome::mask::MaskThreshold and genome::mask::Masks.
- Improved genome::misincorporation::Misincorporations and genome::misincorporation::MisincorporationRecord documentation + doctests.
- Improved parser::Cli and
pmd_maskhigher-level functions documentation.
- Fixes issue #13 :
mask_sequence()function now correctly handles out of index errors when attempting to retrieve positions that lie beyond the chromosome's length. AReferenceOutOfIndexErroris now raised to themask_5p()andmask_3p()functions, whileapply_pmd_mask()now recover from such an error if and only if the read happens to be labeled as 'segment unmapped' (0x4). - Temporary workaround to issue #15 : The current patch directly uses
libc::freeto manually clean up memory between each iteration. A more elegant solution may come in the future, if issue rust-bio/rust-htslib#401 ever resolves.. - Use of
attyensures users are actually feeding the program with an input using piping, when they did not use--bamargument. This ensures a comprehensive error message gets printed, instead of letting the program endlessly hanging. - Improved error handling:
pmd-masknow outputs a comprehensive error message when users forget to provide with a valid misincorporation file path.
Additional -M|--metrics-file argument allows to optionally specify an output summary file, where the software will keep a record of which positions were selected as masking thresholds. This file is headed, tab-separated, lexicographically ordered and structured according to four fields:
Chr: Chromosome name (string)Std: Strand orientation (+|-)5p: Position in base pair whereC>Tmasking stops applying, starting from the5pend.3p: Position in base pair whereG>Amasking stops applying, starting from the3pend.
Note that 5p and 3p positions where the masking threshold was never met are denoted as NA.
- Indels are now taken into account.
- Current implementation is a bit hacky, and could use a lot of optimization (~25% runtime increase has been noticed when applying the software on a single thread. This slowdown has been partially mitigated through link-time optimisation, through ~12%)
- Looks like it works, but still need to implement specific unit tests for indel cases.
- A Small benchmark of
apply_pmd_masking()has been implemented through criterion (See README)
- A More trustworthy codebase : ~91.86% line coverage at this point.
- Basic suite of Unit tests (PRs #4 #6 #9)
- Basic suite of integration tests (PRs #6 and #9)
- Basic Continuous Integration workflow for Ubuntu and MacOS (PRs #5 #7)
- Some very basic doctests within certain user libraries.
- Better separation between libraries and the main logic.
--helpis now a bit more verbose.
- Fixed threadpool construction and cpu_count (#2)
- Restructured code architecture
- Implemented basic error handling using
thiserrorandanyhow. Might attempt to go fromanyhowtomiettein the future, just for the 'gist of it. - Masking is now more aggressive, and will be applied on the whole read length in case of an invalid or missing MaskingThreshold.
- Command line interface
--helpmessage now supports linewrap.
- Fixed threshold validation. (#1)
- Fixed overflowing error when the
3p G>Athreshold length was greater than the read length
- (#2) Threadpool construction and cpu count is broken + thread preemption appears under-utilized.
- very light documentation. Need to take some time to add doctests, now that we have a semblance of structure within the source directory.
- An (Absolutely barbaric) prototype.
- (#1) Threshold validation falsely triggers if the user defined threshold is to low. This is because we currently have no way of distinguishing 'Invalid thresholds' from those which were not found (and thus implies we should apply masking on the whole sequence.)