                      Changes in version 0.0.9004                       

  - Fix centroid m/z/intensity alignment bug.
  - Add regression tests for the reported m/z 164.0528 case.
  - Reduce centroid allocation overhead for faster, leaner processing.
  - Split dense profile peaks at local valleys so close peaks centroid
    separately. (#52)
  - pattern in centroid_one_file() is now honoured as the regular
    expression the documentation describes, instead of being matched
    literally.
  - Fix the command line script, which recursed infinitely on every run
    because its null-coalescing operator called itself.
  - The output file is now published with an atomic rename, so a run
    that fails part way through no longer leaves a truncated file that
    later runs would mistake for finished work.
  - min_datapoints_ms1 and min_datapoints_ms2 must now be whole numbers
    of at least 1. Zero and negative thresholds used to be accepted and
    centroided peak groups that held no data at all, and a fractional
    threshold such as 1.5 can never be reached.
  - centroid_one_file() now returns the logger to the console when it
    finishes. A finished run no longer keeps capturing messages that
    other code logs later, and a log file whose directory has since been
    deleted no longer turns every subsequent logging call in the session
    into an error.

                      Changes in version 0.0.9003                       

  - Update package metadata for version 0.0.9003 and dependencies.

                      Changes in version 0.0.9002                       

  - Fix NaN values in mzML files originating from .wiff2 (#6).

CentroidR 0.0.9001

  - Fixed NaN in mzMLs originating from .wiff2 (#6), see
    https://github.com/sneumann/mzR/issues/307

CentroidR 0.0.9000

  - Initial version.
