noodles-cram 0.78.0

# Changelog

## 0.78.0 - 2025-02-17

### Changed

  * cram: Merge data container and container.

    Data containers are now the only container type. Header and EOF containers
    are handled as special cases.

    This renames the following:

      * `cram::data_container` => `cram::container`,
      * `cram::DataContainer` => `cram::io::reader::Container`, and
      * `cram::io::Reader::read_data_container` => `cram::io::Reader::read_container`.

  * cram/io/reader: Read container (`Reader::read_container`) into a given
    buffer (`cram::io::reader::Container`).

    Container components (`Container::compression_header` and
    `Container::slices`) are now lazily decoded.

  * cram/io/reader/container/slice: Callers must decode and hold block data
    (`Slice::decode_block`).

    These are subsequently used when reading records (`Slice::records`).

  * cram/io/reader/container/slice: `Slice::records` requires a reference
    sequence repository (`fasta::Repository`) and SAM header (`sam::Header`).

  * cram/io/reader: Record iterators (e.g., `Reader::records` and
    `Reader::query`) return instances of `sam::alignment::RecordBuf`.

  * cram/record: Change `cram::Record` to be read-only.

    noodles-cram no longer has a public, mutable CRAM record. Use a more
    general alignment record, e.g., `sam::alignment::RecordBuf`, instead.
    Fields are only available via the `sam::alignment::Record` implementation.

  * cram/record: Include the read group field in the data iterator
    (`Record::data`).

  * cram/record: Resolve sequence and quality scores when necessary.

    This previously typically happened after slice records were decoded. The
    sequence and quality score iterators now resolve their underlying data when
    necessary, i.e., the sequence is for a mapped record or the quality scores
    are stored as an array.

  * cram/record: Reduce the visibility of `Feature`, `Flags`, `MateFlags`.

    The associated fields are no longer accessible.

  * cram/record/sequence/iter: Maintain reference base case when emitting
    substitution base.

    This previously emitted all substitution bases as uppercase. The case of
    the reference base now determines the case of the substitution base.

### Deprecated

  * cram: Deprecate `cram::data_container`, `cram::DataContainer`, and
    `cram::io::Reader::read_data_container`.

    Use the name container instead of data container instead.

### Fixed

  * cram/async/io/writer: Use inferred format version.

    The format version is inferred from the codecs being used, but the file
    definition writer was always using the default CRAM version (3.0).

  * cram/codecs/gzip: Wrap raw DEFLATE stream in gzip container when using
    libdeflate.

### Removed

  * cram/container/slice: Remove mate resolver (`Slice::resolve_mates`).

    Record mates are now resolved after decoding the slice. It no longer has to
    be manually called.

  * cram/record: Remove `Builder`.

    Build, e.g., a `sam::alignment::RecordBuf` instead.

  * cram/record/convert: Remove conversions
    (`Record::try_from_alignment_record` and
    `Record::try_into_alignment_record`).

    These are now unnecessary. Conversion to a CRAM record is only done for on
    write, and `sam::alignment::RecordBuf::try_from_alignment_record` can be
    used for the conversion into an alignment record buffer.

## 0.77.0 - 2025-02-06

### Changed

  * cram/record/builder: Use record field names in method names.

    This renames the following:

     * `set_next_fragment_reference_sequence_id` => `set_mate_reference_sequence_id`,
     * `set_next_mate_alignment_start` => `set_mate_alignment_start`, and
     * `set_distance_to_next_fragment` => `set_distance_to_mate`.

### Deprecated

  * cram/record/builder: Deprecate
    `Builder::set_next_fragment_reference_sequence_id`,
    `Builder::set_next_mate_alignment_start`, and
    `Builder::set_distance_to_next_fragment`.

    Use the record field names instead.

## 0.76.0 - 2025-01-24

### Changed

  * cram: Move file indexer (`index`) to `fs` module.

### Deprecated

  * cram: Deprecate `index`.

    Use `cram::fs::index` instead.

## 0.75.0 - 2025-01-23

### Changed

  * cram: Sync dependencies.

## 0.74.0 - 2025-01-19

### Added

  * cram/io/reader: Add a header reader adapter (`header::Reader`).

    This can be used to read the raw fields of a CRAM header, e.g., the raw SAM
    header.

  * cram/io/writer: Add missing common methods to get the underlying reader
    (`Reader::get_mut` and `Reader::into_inner`).

### Changed

  * cram: Raise minimum supported Rust version (MSRV) to 1.73.0.
  
  * cram/crai: Move reader (`Reader`) and writer (`Writer`) to `io` module.

  * cram/crai: Move convenience functions (`read` and `write`) to `fs` module.

  * cram/async/io/reader: Change `Reader::read_file_header` to return a parsed
    header (`sam::Header`).

    This no longer returns a raw string and now matches the behavior of
    `cram::io::Reader::read_file_header`.

  * cram/async/io/reader/header: Parse header line by line.

    The async header reader now builds a `sam::Header` by parsing a raw header
    line by line. This makes it so that it is no longer required to read the
    entire raw header into memory before parsing.

  * cram/io/reader/header: Discard trailing padding in raw SAM header.

### Deprecated

  * cram/crai: Deprecate `Reader` and `Writer`.

    Use `crai::io::Reader` and `cram::io::Writer`, respectively, instead.

  * cram/crai: Deprecate `read` and `write`.

    Use `crai::fs::read` and `crai::fs::write`, respectively, instead.

  * cram/crai: Deprecate async re-exports (`AsyncReader` and `AsyncWriter`).

    Use `crai::r#async::io::Reader` and `crai::r#async::io::Writer` instead.

## 0.73.0 - 2024-12-20

### Changed

  * cram: Update to bzip2 0.5.0.

## 0.72.0 - 2024-12-12

### Changed

  * cram: Sync dependencies.

## 0.71.0 - 2024-11-07

### Changed

  * cram: Sync dependencies.

## 0.70.0 - 2024-10-22

### Fixed

  * cram/data_container/slice: Return default template length (0)
    when the record alignment start is missing ([#305]).

    This is possible when an unplaced record has a downstream mate.

  * cram/record: Fix number of CIGAR operations calculation.

    This now returns the number of simplified operations rather than all
    feature to CIGAR operation transformations.

[#305]: https://github.com/zaeleus/noodles/issues/305

### Changed

  * cram/data_container/compression_header/data_series_encoding_map/data_series:
    Normalize names.

    Data series names are now closer to how they're called in noodles rather
    than how they are named in the CRAM specification (§ 8.4.2 "Compression
    header block: Data series encodings" (2024-09-04)).

  * cram/data_container/compression_header/preservation_map: Rename tag IDs
    dictionary (`tag_ids_dictionary`) to tag sets (`tag_sets`).

    This is not a dictionary but a list of sets.

  * cram/record: Rename `NextMateFlags` to `MateFlags`.

### Deprecated

  * cram/record: Deprecate `Record::bases`.

    Use `Record::sequence` instead.

## 0.69.0 - 2024-09-26

### Deprecated

  * cram: Deprecate async re-exports (`AsyncReader` and `AsyncWriter`).

    Use `cram::r#async::io::Reader` and `cram::r#async::io::Writer` instead.

## 0.68.0 - 2024-09-04

### Added

  * cram/async/io/writer: Add alignment record writer
    (`Writer::write_alignment_record`) ([#293]).

  * cram/async/io/writer: Add SAM header writer (`Writer::write_header`)
    ([#294]).

    This is a convenience method that writes the file definition and SAM
    header.

[#293]: https://github.com/zaeleus/noodles/issues/293
[#294]: https://github.com/zaeleus/noodles/issues/294

### Changed

  * cram/io/writer/builder: Rename `Builder::build_with_path` and
    `Builder::build_with_writer` to `Builder::build_from_path` and
    `Builder::build_from_writer`, respectively ([#295]).

[#295]: https://github.com/zaeleus/noodles/issues/295

### Deprecated

  * cram/io/writer/builder: Deprecate `Builder::build_with_path` and
    `Builder::build_with_writer`.

    Use `Builder::build_from_path` and `Builder::build_from_writer`,
    respectively, instead.

## 0.67.0 - 2024-08-04

### Added

  * cram/async/io/reader: Add header reader (`Reader::read_header`).

    This is a convenience method that reads the file definition and header and
    returns a parsed SAM header.

  * cram/async/io/reader: Add builder (`Builder`).

### Changed

  * cram/async/io/reader: Add a reference sequence repository as context.

    `Reader::records` and `Reader::query` no longer accepts a reference
    sequence repository but uses the one attached to the reader. Use
    `cram::r#async::io::reader::Builder` to set the appropriate
    `fasta::Repository`.

  * cram/record/builder: Change `set_name` parameter to be generic over
    `Into<BString>`.

## 0.66.0 - 2024-07-14

### Changed

  * cram: Sync dependencies.

## 0.65.0 - 2024-06-17

### Changed

  * cram: Sync dependencies.

## 0.64.0 - 2024-05-31

### Changed

  * cram: Sync dependencies.

## 0.63.0 - 2024-05-16

### Changed

  * cram: Sync dependencies.

## 0.62.0 - 2024-05-08

### Changed

  * cram: Sync dependencies.

## 0.61.0 - 2024-05-02

### Changed

  * cram: Sync dependencies.

## 0.60.0 - 2024-04-22

### Added

  * cram/io/writer/builder: Add `Builder::build_from_path`.

### Changed

  * cram/async: Move reader (`Reader`) and writer (`Writer`) to `io` module.

## 0.59.0 - 2024-04-04

### Changed

  * cram: Sync dependencies.

## 0.58.0 - 2024-03-28

### Changed

  * cram: Sync dependencies.

## 0.57.0 - 2024-03-12

### Changed

  * cram: Sync dependencies.

## 0.56.0 - 2024-02-22

### Changed

  * cram: Sync dependencies.

## 0.55.0 - 2024-02-15

### Changed

  * cram: Sync dependencies.

## 0.54.0 - 2024-02-08

### Changed

  * cram: Sync dependencies.

## 0.53.1 - 2024-02-04

### Fixed

  * cram: Sync dependencies.

## 0.53.0 - 2024-02-01

### Changed

  * cram: Sync dependencies.

## 0.52.0 - 2024-01-25

### Added

  * cram/record: Implement `sam::alignment::Record`.

  * cram/record/features: Add features to CIGAR operations iterator
    (`Features::cigar`).

### Changed

  * cram: Move readers (`Reader` and `IndexedReader`) and writer (`Writer`) to
    `io` module.

  * cram/record/convert: Change `Record::try_from_alignment_record` to accept
    `&sam::alignment::Record`.

### Removed

  * cram/record/builder: Remove deprecated methods.

    These were deprecated in noodles-cram 0.47.0. Use the following instead:

      * `Builder::add_tag` => `Builder::set_tags`,
      * `Builder::add_base` => `Builder::set_bases`,
      * `Builder::add_feature` => `Builder::set_features`, and
      * `Builder::add_quality_scores` => `Builder::set_quality_scores`.

## 0.51.0 - 2023-12-14

### Added

  * cram/async/reader: Add common methods to access the underlying reader:
    `get_ref`, `get_mut`, and `into_inner` ([#220]).

[#220]: https://github.com/zaeleus/noodles/issues/220

### Changed

  * cram: Raise minimum supported Rust version (MSRV) to 1.70.0.

## 0.50.0 - 2023-11-14

### Changed

  * cram: Sync dependencies.

## 0.49.0 - 2023-11-13

### Changed

  * cram/record: Change sequence to be raw bases.

## 0.48.0 - 2023-11-02

### Changed

  * cram: Sync dependencies.

## 0.47.0 - 2023-10-26

### Deprecated

  * cram/record/builder Deprecate add methods: `Builder::add_tag`,
    `Builder::add_base`, `Builder::add_feature`, and
    `Builder::add_quality_score`.

    Use the setters instead.

## 0.46.0 - 2023-10-23

### Changed

  * cram/record: Change quality scores to be raw scores.

## 0.45.0 - 2023-10-19

### Changed

  * cram: Sync dependencies.

## 0.44.0 - 2023-10-12

### Changed

  * cram: Update to libdeflater 1.19.0 (libdeflate 1.19).

  * cram/reader/header_container: Parse header line by line.

    The header parser can now build a `sam::Header` by parsing a raw header
    line by line. This makes it so that it is no longer required to read the
    entire header into memory before parsing.

## 0.43.0 - 2023-09-21

### Changed

  * cram: Sync dependencies.

## 0.42.0 - 2023-09-14

### Changed

  * cram: Sync dependencies.

## 0.41.0 - 2023-08-31

### Changed

  * cram: Sync dependencies.

## 0.40.0 - 2023-08-24

### Changed

  * cram: Sync dependencies.

## 0.39.0 - 2023-08-17

### Changed

  * cram: Sync dependencies.

## 0.38.0 - 2023-08-03

### Changed

  * cram: Sync dependencies.

## 0.37.0 - 2023-07-27

### Changed

  * cram: Sync dependencies.

## 0.36.0 - 2023-07-20

### Changed

  * cram: Sync dependencies.

## 0.35.0 - 2023-07-06

### Changed

  * cram/writer/header_container: Compress the SAM header with gzip.

    This previously wrote the SAM header data uncompressed. It now uses gzip to
    compress the data at the default compression level.

## 0.34.0 - 2023-06-29

### Added

  * cram/reader: Add `Reader::read_header`.

    This is a convenience method that reads the file definition and header and
    returns a parsed SAM header.

  * cram/writer: Add `Writer::write_header`.

    This is a convenience method that writes the file definition and header.

## 0.33.0 - 2023-06-15

### Changed

  * cram: Sync dependencies.

## 0.32.0 - 2023-06-08

### Added

  * cram/reader: Add builder (`cram::reader::Builder`).

### Changed

  * cram: Update to bzip2 0.4.4 ([#174]).

  * cram/reader: Add a reference sequence repository as context.

    `Reader::records` and `Reader::query` no longer accepts a reference
    sequence repository but uses the one attached to the reader. Use
    `cram::reader::Builder` to set the appropriate `fasta::Repository`.

[#174]: https://github.com/zaeleus/noodles/pull/174

## 0.31.0 - 2023-06-01

### Changed

  * cram: Update to async-compression 0.4.0.

  * cram: Update to libdeflater 0.14.0 (libdeflate 1.18).

## 0.30.0 - 2023-05-18

### Added

  * cram/reader/record: Add Elias gamma decoder for integers.

## 0.29.0 - 2023-05-11

### Changed

  * cram: Sync dependencies.

## 0.28.0 - 2023-05-04

### Changed

  * cram: Sync dependencies.

## 0.27.0 - 2023-04-27

### Changed

  * cram: Update to libdeflater 0.13.0.

## 0.26.0 - 2023-04-06

### Added

  * cram: Add an indexed reader (`cram::IndexedReader`).

### Changed

  * cram: Update to bitflags 2.0.2.

  * cram/reader: Change `Reader::read_file_header` to return a parsed header
    (`sam::Header`).

    This no longer returns a raw string.

## 0.25.0 - 2023-03-14

### Changed

  * cram: Sync dependencies.

## 0.24.0 - 2023-03-03

### Changed

  * cram: Sync dependencies.

## 0.26.0 - 2023-04-06

  * cram: Add `libdeflate` feature to enable using libdeflate for gzip streams.

## 0.23.0 - 2023-02-03

### Added

  * cram/crai/record: Implement `std::error::Error::source` for errors.

### Changed

  * cram: Raise minimum supported Rust version (MSRV) to 1.64.0.

  * cram/record: Return `Name`-`Map<ReferenceSequence>` pair from
    `Record::reference_sequence` and `Record::mate_reference_sequence`.

  * cram/record/builder: Change `Builder::add_tag` to receive a `Tag`-`Value`
    pair.

### Removed

  * cram/data_container/compression_header/preservation_map/tag_ids_dictionary/key:
    Remove `From<&sam::record::data::Field>`.

    Use `Key::new` instead.

  * cram/record/feature/code: Remove `TryFrom<char>` and `TryFromCharError`.

    These were deprecated in noodles-cram 0.13.0. Use `TryFrom<u8>` and
    `TryFromByteError`, respectively, instead.

  * cram/record/feature/code: Remove `From<Code>` for `char`.

    This was deprecated in noodles-cram 0.13.0. Convert to a `u8`
    instead.

  * cram/record/resolve: Remove `resolve_features`.

    This was deprecated in noodles-cram 0.16.0. Use
    `Features::try_into_cigar` instead.

## 0.22.1 - 2022-11-29

### Changed

  * cram: Sync dependencies.

## 0.22.0 - 2022-11-18

### Added

  * cram/async/reader: Add query stream (`Reader::query`).

### Changed

  * cram/reader/container/block: Treat blocks with no data as empty, regardless
    of compression method ([#131]).

  * cram/reader/header_container: Require blocks to be either uncompressed or
    gzipped.

    This was clarified in [samtools/hts-specs@4b61c4d].

[#131]: https://github.com/zaeleus/noodles/issues/131
[samtools/hts-specs@4b61c4d]: https://github.com/samtools/hts-specs/commit/4b61c4d7e40748b274478fd481bbfe60592a969b

### Fixed

  * cram/reader/record: Read quality scores filled with `0xff` as missing
    ([#136]).

[#136]: https://github.com/zaeleus/noodles/issues/136

## 0.21.0 - 2022-10-28

### Added

  * cram/async: Add async writer (`cram::AsyncWriter`).

### Fixed

  * cram/reader/header_container: Truncate data by header length.

    This is required if the header text is padded rather than using a new block
    for padding.

## 0.20.0 - 2022-10-20

### Added

  * cram: Add CRAM 3.1 write support ([#107]).

    The format of CRAM 3.1 is structurally the same as 3.0, but 3.1 includes
    additional block content compression methods: rANS Nx16, adaptive
    arithmetic coder, fqzcomp, and name tokenizer.

  * cram/codecs: Add fqzcomp encoder.

  * cram/codecs: Increase visibilities of `aac::Flags`, `rans_4x8::Order`, and
    `rans_nx16::Flags`.

  * cram/codecs/name_tokenizer/encode: Add `DIGITS0`, `DIGITS`, `DELTA`,
    `DELTA0`, and `MATCH` tokens.

### Changed

  * cram: Raise minimum supported Rust version (MSRV) to 1.62.0.

  * cram/writer: Calculate and write missing MD5 checksums (`M5`) in the
    reference sequence dictionary.

### Fixed

  * cram/codecs/aac/encode: Fix serialization of the max symbol for order-0 and
    1 encoders.

  * cram/codecs/aac/encode: Disable bit packing if the input has more than 16
    symbols.

  * cram/codecs/fqzcomp/decode: Fix reading records with varying lengths.

  * cram/codecs/name_tokenizer/decode: Add missing mutable `DUP` reader.

  * cram/codecs/name_tokenizer/decode: Fix setting `TYPE` reader when reading
    token byte streams.

  * cram/codecs/name_tokenizer/encode: Strip final separator before splitting
    names.

  * cram/codecs/rans_4x8/encode: Fix run-length calculation when writing
    frequencies.

  * cram/codecs/rans_4x8/encode: Avoid scaling frequencies to 0 during
    normalization.

  * cram/codecs/rans_4x8/encode: Ensure the sum of the normalized frequencies
    equals the scaling factor.

  * cram/codecs/rans_4x8/encode/order_1: Fix order of rANS state
    renormalizations.

  * cram/codecs/rans_nx16/encode: Fix run-length calculation when writing
    frequencies.

  * cram/codecs/rans_nx16/encode: Avoid scaling frequencies to 0 during
    normalization.

  * cram/codecs/rans_nx16/encode: Handle 0 RLE symbols.

  * cram/codecs/rans_nx16/encode: Disable rANS encoding when the input is too
    small.

  * cram/codecs/rans_nx16/encode: Disable bit packing if the input has more
    than 16 symbols.

  * cram/codecs/rans_nx16/encode: Ensure the sum of the normalized frequencies
    equals the scaling factor.

  * cram/codecs/rans_nx16/encode/order_1: Fix order of rANS state
    renormalizations.

[#107]: https://github.com/zaeleus/noodles/issues/107

## 0.19.0 - 2022-09-29

### Added

  * cram/codecs: Add rANS Nx16, adaptive arithmetic coder, and name tokenizer
    encoders.

  * cram/data_container: Add block content-encoder map
    (`BlockContentEncoderMap`) to assign encoders to block contents (core, data
    series, and tag values).

  * cram/writer/builder: Add setter for block content-encoder map.

### Changed

  * cram: Increased visibilities of `cram::writer`,
    `data_series_encoding_map::DataSeries`, and `tag_ids_dictionary::Key`.

  * cram/writer/builder: Change `build` to `build_with_writer`.

    `writer::Builder` no longer holds the inner reader.

### Fixed

  * cram/container/block: Add terminators (`NUL`) to decoded names.

### Removed

  * cram/writer: Remove `Writer::builder`.

    Use `cram::writer::Builder` directly instead.

## 0.18.0 - 2022-08-16

### Changed

  * cram: Raise minimum supported Rust version (MSRV) to 1.59.0.

## 0.17.0 - 2022-07-05

### Changed

  * cram: Sync dependencies.

## 0.16.0 - 2022-06-08

### Added

  * cram/async/reader: Add records stream (`AsyncReader::records`).

### Changed

  * cram/async/reader/container/header: Verify the expected CRC32 checksum of
    the container header.

  * cram/data_container/slice: Template size for records with downstream mates
    is now calculated from the leftmost to rightmost base, regardless of
    orientation.

  * cram/data_container/slice: Verify reference sequence checksum.

    § 11 "Reference sequences" (2021-11-15): "All CRAM reader implementations
    are expected to check for reference MD5 checksums and report any missing or
    mismatching entries."

  * cram/data_container/slice/builder: Match mate by read name.

  * cram/reader/container/block: Verify the expected CRC32 checksum of the
    block.

  * cram/reader/container/header: Verify the expected CRC32 checksum of the
    container header.

  * cram/record/convert: Change `Record::try_into_sam_record` to
    `Record::try_into_alignment_record`.

    This now takes a `&sam::alignment::Record`.

### Fixed

  * cram/data_container/builder: Disable encoding alignment start positions as
    deltas when the container has multi-reference slices.

  * cram/data_container/slice: Resolve read group ID.

    This adds the `RG` field with the corresponding read group ID to data.

  * cram/data_container/slice: Resolve read name if none is set.

### Deprecated

  * cram/record/resolve: Deprecate `resolve_features`.

    Use `Features::try_into_cigar` instead.

### Removed

  * cram/record: Remove `sam::AlignmentRecord`.

    Convert to an alignment record (`Record::try_into_alignment_record`)
    instead.

## 0.15.0 - 2022-04-14

### Added

  * cram/reader: Add query iterator (`Reader::query`).

  * cram/reader/data_container/compression_header/encoding: Add Golomb and
    Golomb-Rice encoding decoders.

  * cram/writer/data_container/compression_header/encoding: Add Golomb and
    Golomb-Rice encoding writers.

    These see no usage in practice, but they are still in the spec (§ 13.8 and
    § 13.9, respectively). They are only added for completeness.

### Changed

  * cram/data_container/slice/builder: Slices can now be multi-reference.

### Fixed

  * cram/container: Return error (`TryFromSamHeaderError::InvalidHeaderLength`)
    when header length is too large.

## 0.14.0 - 2022-03-29

### Added

  * cram/reader: Implement `sam::AlignmentReader`.

  * cram/record: Add conversion from an alignment record to a CRAM record.

  * cram/record/features: Add conversion from SAM CIGAR operations.

  * cram/writer: Implement `sam::AlignmentWriter`.

### Changed

  * cram/crai/record: Change reference sequence ID to an `Option<usize>`;
    alignment start, `Option<Position>`; and alignment span, `usize`.

  * cram/record: Move `mapping_quality`, `read_name`, and `quality_scores` to
    the implementation of `sam::AlignmentRecord`.

  * cram/record: Rename `flags` to `cram_flags`.

  * cram/record: Wrap read name as `sam::record::ReadName`.

  * cram/record: Wrap quality scores as `sam::record::QualityScores`.

  * cram/record: Wrap sequence as `sam::record::Sequence`.

  * cram/record: Change tags to `sam::record::Data`.

  * cram/record: Change alignment start and next mate alignment start to a
    `Position`.

  * cram/record: Change distance to next fragment to an `Option<usize>`.

    This distance can only be forward (_CRAM format specification (version
    3.0)_ (2021-10-15) § 8.4 "Compression header block": "... NF can only refer
    to a record later within [the] slice.") and is only set when the "has mate
    downstream" flag is set.

  * cram/record: Change read group ID to an `Option<usize>`.

  * cram/record: Change reference sequence ID and next fragment reference
    sequence ID to an `Option<usize>`.

  * cram/record/convert: Accept `sam::Header` instead of `ReferenceSequences`.

  * cram/record/feature: Change underlying data types:

      * bases as `sam::record::sequence::Base`,
      * lengths as `usize`,
      * positions as `Position`,
      * quality scores as `sam::record::quality_scores::Score`, and
      * substitution as `substitution::Value`.

  * cram/record/resolve: Merge adjacent CIGAR operations.

    For example, instead of 1M1M2M, this is now 4M.

  * cram/record/tag/key: Change tag to a `sam::record::data::field::Tag`.

  * cram/writer: Make a reference sequence repository optional.

    Use `Builder::set_reference_sequence_repository` to set one.

  * cram/writer: A writer no longer has to be constructed with a header.

    However, `Writer::write_record` and `Writer::try_finish` must be called
    with a reference to a `sam::Header` now.

### Fixed

  * cram/data_container/slice: Only attempt to resolve quality scores for
    mapped records.

  * cram/data_container/slice/builder: Detach all records before writing.

    The slice builder does not yet calculate mate distances, so all records are
    currently written as detached.

  * cram/data_container/slice/builder: Normalize sequence for MD5 digest
    calculation.

  * cram/reader/data_container: Disallow unexpected block content types when
    reading a data container.

  * cram/record/convert: Create read group (`RG`) data field if a read group ID
    is set.

  * cram/record/convert: Only convert CIGAR to features for mapped records.

    Unmapped records cannot have features.

  * cram/record/resolve: Handle `Scores` feature when resolving quality scores.

  * cram/writer: Avoid possible wrapping when casting `usize` to `i32`.

### Removed

  * cram/record: Remove `ReadGroupId`.

    Use `usize` instead.

  * cram/record: Remove `Tag`.

    Use `sam::record::data::Field` instead.

  * cram/writer: Remove `Drop`.

    A call to `Writer::try_finish` must be made before the writer is dropped.

## 0.13.0 - 2022-03-02

### Added

  * cram/data_container/slice: Add slice records resolver.

    This resolves mates, read names, bases, and quality scores.

  * cram/record: Add features wrapper (`Features`).

  * cram/record: Implement `sam::RecordExt`.

  * cram/record/feature/code: Implement `TryFrom<u8>`.

  * cram/record/feature/code: Add conversion to `u8`.

  * cram/record/resolve: Add quality scores resolver.

  * cram/record/resolve: Allow base resolution when there is no reference
    sequence available.

    Even when both an external reference sequence and embedded reference
    sequence are missing, it is still possible to resolve bases.

  * cram/writer: Add builder.

  * cram/writer/builder: Add options to change applicable preservation map
    values, i.e., whether to preserve read names and whether to encode record
    alignment start positions as deltas.

### Changed

  * cram/data_container/builder: Decrease max slice count to 1.

  * cram/data_container/slice/builder: Increase max record count to 10240.

  * cram/reader: `records` uses a `fasta::Repository` instead of
    `&[fasta::Record]`.

    This avoids having to do an initial read of all reference sequences. It
    also now requires the SAM header.

  * cram/reader/records: Resolve bases and quality scores when reading records.

    This previously only resolved mates and read names.

  * cram/record/convert: `Record::try_into_sam_record` now assumes the record
    is fully resolved before conversion.

  * cram/record/read_group_id: Change underlying type to a `usize`.

    Use `From<usize>` instead of `From<i32>`.

  * cram/record/resolve: `resolve_bases` takes a `sam::record::Position`
    instead of an `i32` for `alignment_start`.

  * cram/writer: Use reference sequence repository (`fasta::Repository`) and
    SAM header (`sam::Header`) when writing records.

### Deprecated

  * cram/record/feature/code: Deprecate fallible conversion from `char`.

    This also deprecates `TryFromCharError`. Use `TryFrom<u8>` instead.

  * cram/record/feature/code: Deprecate conversion to `char`.

    Convert to a `u8` instead.

### Fixed

  * cram/container: Only set alignment start and span when the container has
    single-reference slices of the same reference.

  * cram/container: Ensure all slices in the container have the same reference
    sequence ID.

  * cram/data_container/slice/builder: Update substitution codes before writing
    records.

  * cram/data_container/slice/builder: Only set alignment start and span when
    the slice is a single-reference slice.

  * cram/data_container/compression_header/preservation_map/substitution_matrix/builder:
    Read read base from read bases.

  * cram/record/convert: Only resolve features when the record is mapped.

  * cram/writer/record: Fix data series type when the features code encoding is
    missing.

### Removed

  * cram/data_container/slice: Remove `resolve_mates`.

    Use `Slice::resolve_records` instead.

  * cram/record/read_group_id: Remove conversion to and from `i32`.

    Convert from `usize` instead.

  * cram/record/resolve: Remove `resolve_bases`.

    Base resolution can only be guaranteed when the slice is available, as it
    may contain an embedded reference sequence.

    Use `Slice::resolve_records` instead.

## 0.12.0 - 2022-02-17

### Added

  * cram: Set minimum supported Rust version (MSRV) to 1.56.0 in package
    manifest.

  * cram/container/block/compression_method: Add decoder support for CRAM 3.1
    (draft) block compression methods: rANS Nx16, adaptive arithmetic coder,
    fqzcomp, and name tokenizer.

  * cram/record/resolve: Add read base (`B`) base resolver.

### Changed

  * cram: Write 0 for alignment start and alignment span for slices with
    unmapped records when indexing.

  * cram/data_container/slice: The mate resolver (`Slice::resolve_mates`) is
    now fallible.

  * cram/record: `Record::alignment_end` now returns
    `Option<io::Result<sam::record::Position>>`.

### Fixed

  * cram/data_container/slice: Fix template size calculation when resolving
    mates.

  * cram/writer/data_container/compression_header/data_series_encoding_map:
    Avoid truncating data length.

## 0.11.0 - 2022-01-27

### Added

  * cram/record/resolve: Add bases (`b`) base resolver.

### Fixed

  * cram/async/reader/record: Read feature code as byte.

  * cram/huffman: Fix read length from bit reader.

  * cram/reader/record: Read feature code as byte.

    This was incorrectly being read as an ITF-8 value.

## 0.10.0 - 2022-01-13

### Added

  * cram/record: Mapping quality is now stored as an `Option`.

    Valid mapping qualities are between 0 and 254, inclusive (`Some`). A
    mapping quality of 255 is considered to be missing (`None`).

## 0.9.0 - 2021-12-16

### Changed

  * cram: Update to [md-5 0.10.0].

[md-5 0.10.0]: https://crates.io/crates/md-5/0.10.0

## 0.8.3 - 2021-12-09

### Fixed

  * cram: Sync dependencies.

## 0.8.2 - 2021-12-02

### Fixed

  * cram: Require tokio's `fs` feature as a dependency ([#62]).

[#62]: https://github.com/zaeleus/noodles/issues/62

## 0.8.1 - 2021-11-18

### Fixed

  * cram: Sync dependencies.

## 0.8.0 - 2021-11-11

### Added

  * cram/crai: Add convenience write function (`crai::write`).

  * cram/crai/async: Add async writer (`crai::AsyncWriter`).

  * cram/crai/async: Add convenience write function (`crai::r#async::write`).

### Changed

  * cram: Update to Rust 2021.

## 0.7.0 - 2021-10-16

### Added

  * cram/data_container/compression_header/data_series_encoding_map/
    data_series: Add legacy TC and TN data series.

    These are no longer used in CRAM 3.0 but still need to be handled. See
    samtools/hts-specs@9a0513783826516fb8086ecf82d13631a2292f75.

  * cram/record/resolve: Handle reference skip feature in sequence resolver.

## 0.6.1 - 2021-10-02

### Fixed

  * cram: Sync dependencies.

## 0.6.0 - 2021-10-01

### Added

  * cram/reader: Add common methods to access the underlying reader: `get_ref`,
    `get_mut`, and `into_inner`.

### Fixed

  * cram/rans/decode/order_1: Fix overflow when reading frequencies.

## 0.5.1 - 2021-09-23

### Fixed

  * cram/async/reader/container/header: Fix reading starting position on the
    reference.

  * cram/async/reader/data_container/slice/header: Fix reading alignment start.

## 0.5.0 - 2021-09-19

### Added

  * cram/crai/record: Implement `Display`.

  * cram/reader: Add data container reader.

    This can be used to manually read records from slices.

  * cram/record: Add conversion to SAM record
    (`cram::Record::try_into_sam_record`).

### Changed

  * cram/record: Change alignment start to a `sam::record::Position`.

  * cram/record: Change next mate alignment start to a `sam::record::Position`.

  * cram/record/resolve: Pass compression header rather than substitution
    matrix.

    The compression header includes the substitution matrix in the preservation
    map.

### Fixed

  * cram/async/reader/data_container/slice/header: Read remainder of stream as
    optional tags.

  * cram/reader/container: Avoid casts that may truncate.

  * cram/reader/data_container/compression_header/encoding: Avoid casts that
    may truncate.

    Buffer sizes that convert from `Itf8` to `usize` now check whether they are
    in range.

  * cram/reader/data_container/slice/header: Read remainder of stream as
    optional tags.

  * cram/record/resolve: Increment feature position with operations that
    consume the read.

  * cram/record/resolve: Include last feature position.

## 0.4.0 - 2021-09-01

### Added

  * cram/async/reader: Add data container reader.

  * cram/reader: Add data container reader.

    This can be used to manually read records from slices.

### Changed

  * cram/record: `Record::read_length` is now stored as a `usize`.

### Fixed

  * cram/reader/data_container/compression_header: Avoid casts that may
    truncate.

    Buffer sizes that convert from `Itf8` to `usize` now check whether they are
    in range.

## 0.3.0 - 2021-08-19

### Added

  * cram/async: Add async header reader (`cram::AsyncReader`).

    This is a partial async CRAM reader that can only read the file definition
    and file header.

  * cram/crai/async: Add async reader (`crai::AsyncReader`).

  * cram/crai/async: Add async writer (`crai::AsyncWriter`).

    Async I/O can be enabled with the `async` feature.

## 0.2.2 - 2021-08-11

### Fixed

  * cram: Sync dependencies.

## 0.2.1 - 2021-07-30

### Fixed

  * cram: Sync dependencies.

## 0.2.0 - 2021-07-21

### Added

  * cram/record/tag: Add conversion from `Tag` to `sam::record::data::Field`.

### Fixed

  * cram: Fixed documentation link in package manifest ([#31]).

[#31]: https://github.com/zaeleus/noodles/issues/31

## 0.1.0 - 2021-07-14

  * cram: Initial release.