khmer v1.4 release notes

This is the v1.4 release of khmer featuring the results of our March and April (PyCon) coding sprints and the 16 new contributors; the use of the new v0.8 release of screed (the library we use for pure Python reading of nucleotide sequence files); and the addition of @luizirber's HyperLogLog counter for quick cardinality estimation.

Documentation is at

New items of note:

Casava 1.8 read naming is now fully supported and in general the scripts no longer mangle read names. Side benefits: will no longer drop reads with 'bad' names; can generate output in CSV format. #759 #818 @ctb #873 @ahaerpfer

Most scripts now support a "broken" interleaved paired-read format for FASTA/ FASTQ nucleotide sequence files. ` <>`__ has been promoted from the sandbox as well (with streaming support). #759 @ctb #963 @sguermond #933 @standage

The script to transform an interleaved paired-read nucleotide sequence file into two files now allows one to name the output files which can be useful in combination with named pipes for streaming processing #762 @ctb

Streaming everywhere: thanks to screed v0.8 we now support streaming of almost all inputs and outputs. #830 @aditi9783 #812 @mr-c #917 @bocajnotnef #882 @standage

Need a quick way to count total number of unique k-mers in very low memory? the script in the sandbox uses a HyperLogLog counter to quickly (and with little memory) provide an estimate with a controllable error rate. #257 #738 #895 #902 @luizirber can now process both a paired interleaved sequence file and a file of unpaired reads in the same invocation thus removing the need to write the counting table to disk as required in the workaround. #957 @susinmotion

Notable bugs fixed/issues closed:

Paired-end reads from Casava 1.8 no longer require renaming for use in and when used in paired mode #818 @ctb

Python version support clarified. We do not (yet) support Python 3.x #741 @mr-c

If a single output file mode is chosen for we now default to overwriting the output. Appending the output is available by using the append redirection operator from the shell. #843 @drtamermansour

Scripts that consume sequence data using C++ will now properly throw an error on truncated files. #897 @kdmurray91 And while writing to disk we properly check for errors #856 #962 @mr-c no longer fails with small files and many threads. #900 @mr-c

Additional fixes/features

Of interest to users:

Many documentation updates #753 @PamelaM, #782 @bocajnotnef, #845 @alameldin, #804 @ctb, #870 @SchwarzEM, #953 #942 @safay, #929,@davelin1, #687 #912 #926 @mr-c

Installation instructions for Conda, Arch Linux, and Mac Ports have been added #723 @reedacartwright #952 @elmbeech #930 @ahaerpfer

The example script for the STAMPS database has been fixed to run correctly #781 @drtamermansour added -o option to allow specification of an output directory #752 @bede

Fixed a string formatting and a boundary error in #773 @qingpeng #995 @ctb

CSV output added to,, and, and #831 #854 #855 @drtamermansour #959 @anotherthomas

TSV/JSON output of enhanced with the total number of reads processed #996 @kdmurray91 Output files are now also checked to be writable before loading the input files #672 @pgarland @bocajnotnef now prints the output filename nicely #827 @kdmurray91

Cleaned up error for input file not existing #772 @jessicamizzi #851 @ctb

Fixed error in #860 @TheOneHyer

The help text for for the --no-bigcounts/-b flag has been clarified #857 @kdmurray91

@lexnederbragt confirmed an old bug has been fixed with his test for whitespace in sequence identifiers interacting with the script #979

Now safe to copy-and-paste from the user documentation as the smart quotes have been turned off. #967 @ahaerpfer

The script has been restored to the sandbox. #920 @SherineAwad will warn if two of the input files have the same name #932 @elmbeech

Of interest to developers:

Switched away from using --user install for developers #740 @mr-c @drtamermansour & #883 @standage

Developers can now see a summary of important Makefile targets via make help #783 @standage

The unused khmer.load_pe module has been removed #828 @kdmurray91

Versioneer bug due to new screed release was squashed #835 @mr-c

A Python 2.6 and 2.7.2 specific bug was worked around #869 @kdmurray91 @ctb

Added functions hash_find_all_tags_list and hash_get_tags_and_positions to CountingHash objects #749 #765 @ctb

The make diff-cover and ChangeLog formatting requirements have been added to checklist #766 @mr-c

A useful message is now presented if large tables fail to allocate enough memory #704 @mr-c

A checklist for developers adding new CPython types was added #727 @mr-c

The sandbox graduation checklist has been updated to include streaming support #951 @sguermond

Specific policies for sandbox/ and scripts/ content, and a process for adding new command line scripts into scripts/ have been added to the developer documentation #799 @ctb

Sandbox scripts update: corrected #! Python invocation #815 @Echelon9, executable bits, copyright headers, no underscores in filenames #823 #826 #850 @alameldin several scripts deleted, docs + requirements updated #852 @ctb

Avoid running big-memory tests on OS X #819 @ctb

Unused callback code was removed #698 @mr-c

The CPython code was updated to use the new checklist and follow additional best practices #785 #842 @luizirber

Added a read-only view of the raw counting tables #671 @camillescott #869 @kdmurray91

Added a Python method for quickly getting the number of underlying tables in a counting or presence table #879 #880 @kdmurray91

The C++ library can now be built separately for the brave and curious developer #788 @kdmurray91

The ReadParser object now keeps track of the number of reads processed #877 @kdmurray91

Documentation is now reproducible #886 @mr-c

Python future proofing: specify floor division #863 @mr-c

Miscellaneous spelling fixes; thanks codespell! #867 @mr-c

Debian package list update #984 @mr-c

khmer.kfile.check_file_status() has been renamed to check_input_files() #941 @proteasome now uses it to check the input counting table #931 @safay was refactored to not pass the ArgParse object around #965 @susinmotion

Developer communication has been clarified #969 @sguermond

Tests using the 'fail_okay=true' parameter to runscript have been updated to confirm the correct error occurred. 3 faulty tests were fixed and the docs were clarified #968 #971 @susinmotion

FASTA test added for #901 @jessicamizzi

'added silly test for empty file warning' #557 @wltrimbl @bocajnotnef

A couple tests were made more resilient and some extra error checking added in CPython land #889 @mr-c

Copyright added to pull request checklist #940 @sguermond

khmer_exceptions are now based on std::strings which plugs a memory leak #938 @anotherthomas

Python docstrings were made PEP257 compliant #936 @ahaerpfer

Some C++ comments were converted to be Doxygen compliant #950 @josiahseaman

The counting and presence table warning logic was refactored and centralized #944 @susinmotion

The release checklist was updated to better run the post-install tests #911 @mr-c

The unused method find_all_tags_truncate_on_abundance was removed from the CPython API #924 @anotherthomas

OS X warnings quieted #887 @mr-c

Known issues:

All of these are pre-existing.

Some users have reported that will utilize more memory than it was configured for. This is being investigated in

Some scripts only output FASTA even if given a FASTQ file. This issue is being tracked in


@ctb, @kdmurray91, @mr-c, @drtamermansour, @luizirber, @standage, @bocajnotnef, ‡@susinmotion, @jessicamizzi, ‡@elmbeech, ‡@anotherthomas, ‡@sguermond, ‡@ahaerpfer, ‡@alameldin, ‡@TheOneHyer, ‡@aditi9783, ‡@proteasome, ‡@bede, ‡@davelin1, @Echelon9, ‡@reedacartwright, @qingpeng, ‡@SchwarzEM, ‡@scottsievert, @PamelaM, @SherineAwad, ‡@josiahseaman, ‡@lexnederbragt,

‡ Indicates new contributors

Issue reporters

@moorepants, @teshomem, @macmanes, @lexnederbragt, @r-gaia-cs, @magentashades

comments powered by Disqus