Cutadapt

-----------------

* :issue:`555`: Add support for dumping statistics in JSON format using ``--json``.
* :issue:`541`: Add a "Read fate breakdown" section heading to the report, and also
add statistics for reads discarded because of ``--discard-untrimmed`` and
``--discard-trimmed``. With this, the numbers in that section should add up to 100%.
* Add option ``-Q``, which allows to specify a quality-trimming threshold for R2 that is
different from the one for R1.
* :issue:`567`: Add ``noindels`` adapter-trimming parameter. You can now write
``-a "ADAPTER;noindels"`` to disallow indels for a single adapter only.
* :issue:`570`: Fix ``--pair-adapters`` not finding some pairs when reads contain
more than one adapter.
* :issue:`524`: Fix a memory leak when using ``--info-file`` with multiple cores.
* :issue:`559`: Fix adjacent base statistics not being shown for linked adapters.

-----------------

* :issue:`481`: An experimental single-file Windows executable of Cutadapt
is `available for download on the GitHub "releases"
page <https://github.com/marcelm/cutadapt/releases>`_.
* :issue:`517`: Report correct sequence in info file if read was reverse complemented
* :issue:`517`: Added a column to the info file that shows whether the read was
reverse-complemented (if ``--revcomp`` was used)
* :issue:`320`: Fix (again) "Too many open files" when demultiplexing

-----------------

* :issue:`504`: Fix a crash on Windows.
* :issue:`490`: When ``--rename`` is used with ``--revcomp``, disable adding the
``rc`` suffix to reads that were reverse-complemented.
* Also, there is now a ``{rc}`` template variable for the ``--rename`` option, which
is replaced with "rc" if the read was reverse-complemented (and the empty string if not).
* :issue:`512`: Fix issue :issue:`128` once more (the “Reads written” figure in the report
incorrectly included both trimmed and untrimmed reads if ``--untrimmed-output`` was used).
* :issue:`515`: The report is now sent to stderr if any output file is
written to stdout

-----------------

* :issue:`437`: Implement a ``--rename`` option for :ref:`flexible read
name modifications <read-renaming>` such as moving a barcode sequence
into the read name.
* :issue:`503`: The index for demultiplexing is now created a lot faster
(within seconds instead of minutes) when allowing indels.
* :issue:`499`: Fix combinatorial demultiplexing not working when using
multiple cores.

-----------------

* :issue:`443`: With ``--action=retain``, it is now possible to trim reads while
leaving the adapter sequence itself in the read. That is, only the sequence
before (for 5’ adapters) or after (for 3’ adapters) is removed. With linked
adapters, both adapters are retained.
* :issue:`495`: Running with multiple cores did not work using macOS and Python 3.8+.
To prevent problems like these in the future, automated testing has been extended
to also run on macOS.
* :issue:`482`: Print statistics for ``--discard-casava`` and ``--max-ee`` in the
report.
* :issue:`497`: The changelog for 3.0 previously forgot to mention that the following
options, which were deprecated in version 2.0, have now been removed, and
using them will lead to an error: ``--format``, ``--colorspace``, ``-c``, ``-d``,
``--double-encode``, ``-t``, ``--trim-primer``, ``--strip-f3``, ``--maq``,
``--bwa``, ``--no-zero-cap``. This frees up some single-character options,
allowing them to be re-purposed for future Cutadapt features.

-----------------

* Demultiplexing on multiple cores is now supported. This was the last feature that
only ran single-threaded.
* :issue:`478`: Demultiplexing now always generates all possible output files.
* :issue:`358`: You can now use ``-e`` also :ref:`to specify the maximum number of
errors <error-tolerance>` (instead of the maximum error rate). For example, write
``-e 2`` to allow two errors over a full-length adapter match.
* :pr:`486`: Trimming many anchored adapters (for example when demultiplexing)
is now faster by using an index even when indels are allowed. Previously, Cutadapt
would only be able to build an index with ``--no-indels``.
* :issue:`469`: Cutadapt did not run under Python 3.8 on recent macOS versions.
* :issue:`425`: Change the default compression level for ``.gz`` output files from 6
to 5. This reduces the time used for compression by about 50% while increasing file
size by less than 10%. To get the old behavior, use ``--compression-level=6``.
If you use Cutadapt to create intermediate files that are deleted anyway,
consider also using the even faster option ``-Z`` (same as ``--compression-level=1``).
* :pr:`485`: Fix that, under some circumstances, in particular when trimming a
5' adapter and there was a mismatch in its last nucleotide(s), not the entire adapter
sequence would be trimmed from the read. Since fixing this required changed the
alignment algorithm slightly, this is a backwards incompatible change.
* Fix that the report did not include the number of reads that are too long, too short
or had too many ``N``. (This unintentionally disappeared in a previous version.)
* :issue:`487`: When demultiplexing, the reported number of written pairs was
always zero.
* :issue:`497`: The following options, which were deprecated in version 2.0, have
been removed, and using them will lead to an error:
``--format``, ``--colorspace``, ``-c``, ``-d``, ``--double-encode``,
``-t``, ``--trim-primer``, ``--strip-f3``, ``--maq``, ``--bwa``, ``--no-zero-cap``.
This frees up some single-character options,
allowing them to be re-purposed for future Cutadapt features.
* Ensure Cutadapt runs under Python 3.9.
* Drop support for Python 3.5.