fusion 1.3.0

fusion 1.3.0

A round of correctness fixes to parseNMR, and the tests to hold them in
place. Several change the contents of the daE, so runs parsed with an earlier
version are not directly comparable.

Breaking

• obsDescr is built once and holds info, procs and params. The IVDr QC
  tables are opt in through the new tests = TRUE option, and are appended
  after those three. Previously the QC tables were computed and then discarded
  by an unconditional overwrite, so no daE written since 2024-09-30 carried
  them, whatever the input.
• what must name a single measurement. Asking for several silently kept only
  the last one read while the merge step subset a data matrix belonging to a
  different read. An unsupported value now reports itself instead of surfacing
  as an undefined object later.
• sampleID keeps the casing the instrument recorded in USERA2. The local
  folder branch used to lowercase SLTR, LTR, PQC and QC, which the
  rolodex branch did not, so the same sample had two different identifiers
  depending on where it was read from.
• sampleType is classified the same way whatever the source. The local folder
  branch used to leave the whole sampleID in the field for QC samples, so
  getID(da, type = "ltr") matched nothing on a locally parsed daE.
• Rows of the merged tables are now aligned, not merely subset. acqus, the
  data table and loe were filtered independently with nothing keeping the
  same sample on the same row. leftJoinWithAcqus had the same problem through
  merge(), which sorts on the key.
• For 3mm tubes, SPC3_2 and SPC_Glyc are no longer halved. Both the
  numerator and the denominator are already corrected, so the ratio divides the
  tube out. Existing spcglyc results from 3mm tubes have these two columns at
  half their true value and need regenerating.

Fixes

• specOpts$im is read as a value rather than by the presence of its name, so
  im = FALSE no longer produces a complex matrix.
• Subsetting the data matrix keeps two dimensions, so a run where a single
  sample survives the intersection no longer collapses to a vector.
• A cancelled menu aborts instead of yielding a zero length choice, which
  paste() silently dropped from the output file name.
• Interactive project, cohort and matrix choices reach the output file name.
  They were stored in local variables the file name never read.
• Supplying projectName without cohortName no longer fails on an undefined
  projectId.
• outputDir is checked before anything is read, so a bad path costs a second
  rather than a whole parse. An existing daE prompts before being overwritten
  when interactive, and warns when running from a script.
• parseNMR() returns the dataElement invisibly from both paths. The writing
  branch used to return NULL.
• The spcglyc trim no longer discards every column when there is nothing to
  trim, and the TSP region is cut on the same axis it is labelled with.

Performance

• The QC report is only read when tests = TRUE.

Tests

• parseNMR is covered again, against the Bruker fixture shipped with
  nmr.parser so the suite runs anywhere, plus integration tests against the
  COVr49 run that trace a value from the Bruker report through to its position
  in the daE. Nothing is written to the working directory.

fusion 0.0.0.9000

• Added a NEWS.md file to track changes to the package.

1.2.4

1.2.4

1.2.3

daE naming issue solved

1.2.2

release 1.2.2

1.2.1

v1.2.1

1.2.0

1.2.0 feat spyglyc with correct concentrations

1.1.0

add migration to rldx on internal server

1.0.8

Merge branch 'master' of github.com:phenological/fusion

1.0.7

fix import of setDT

1.0.6

Merge pull request #2 from phenological/abe

fixed?: fromVector failing on NA and columns named "type"