Staden Io_lib v1.14.9

Version 1.14.9 (9th February 2017)

Updates:

• BAM: Added CRC checking. Bizarrely this was absent here and in most
  other BAM implementations too. Pure BAM decode of an uncompressed
  BAM is around 9% slower and compressed BAM to compressed BAM is
  almost identical. The most significant hit is reading uncompressed
  BAM (and doing nothing else) which is 120% slower as CRC dominates.
  Options are available to disable the CRC checking incase this is an
  issue (scramble -!).
• CRAM: Now supports bgziped fasta references.
• CRAM/SAM: Headers are now kept in the same basic type order while
  transcoding. (Eg all @pg before all @sq, or vice versa, depending on
  input ordering.)
• CRAM: Compression level 1 is now faster but larger. (The old -1 and
  -2 were too similar.)
• CRAM: Improved compression efficiency in some files, when switching
  from sorted to unsorted data.
• CRAM: Various speedups relating to memory handling,
  multi-threaded performance and the rANS codec.
• CRAM: Block CRC checks are now only done when the block is used,
  speeding up multi-threading and tools that do not decode all blocks
  (eg flagstat).
• Scramble -g and -G options to generate and reuse bgzip indices when
  reading and writing BAM files.
• Scramble -q option to omit updating the @pg header records.
• Experimental cram_filter tool has been added, to rapidly produce
  cram subsets.
• Migrated code base to git. Use github for primary repository.

Bug fixes:

• BAM: Fixed the bin value calculation for placed but unmapped reads.
• CRAM: Fixed file descriptor leak in refs_load_fai().
• CRAM: Fixed a crash in MD5 calculation for sequences beyond the
  reference end.
• CRAM: Bug fixes when encoding malformed @sq records.
• CRAM: Fixed a rare renormalisation bug in rANS codec.
• Fixed tests so make -j worked.
• Removed ancient, broken and unused popen() code.