v1.1.0

Added

• fitab accepts a fitopts struct. Each field overrides the
  same-named option: weights, order, mask, quantile,
  refqtls, Nboot, alpha, or plotfit.
• fitab errors with baseflow:fitab:invalidFitopt for a fitopts
  field of the wrong type and with baseflow:fitab:unknownFitopt for an
  unknown field.
• fitevents and setopts('fitevents') accept fitopts and pass it
  to every fitab call. weights and mask must be scalars there,
  because each event fit has its own points. Any other size raises
  baseflow:fitevents:nonscalarFitopt. fitab expands a scalar
  weights or mask to every point.
• fitevents and setopts('fitevents') accept a ctsmethod option.
  It selects the CTS stencil: B1 (the default), B2, F1, F2,
  C2, or C4. fitevents passes it to getdqdt.
• fitevents forwards a fitorder option to fitab for linear
  reservoir fits.
• struct2varargin converts a name-value struct to a cell array.
  trendplot and formatPlotMarkers call it in place of
  namedargs2cell.
• nonnansegments accepts a vector, a matrix, or a cell array whose
  elements are vectors or matrices, and applies the same rules to each
  element. For a matrix, the option input selects one result per
  column ('each', the default), the rows where every column is non-nan
  ('all'), or the rows where any column is non-nan ('any').
• tests/test_eqstrings.m checks the value and symbolic labels of
  aQbString, QtString, and QtauString.
• fitab fits the 'ols' method on GNU Octave. A weighted
  least-squares solve with t-based confidence intervals replaces the
  Curve Fitting Toolbox fit and confint calls, and on MATLAB the
  results match them to rounding.
• toolbox/docs/baseflow_powerlaw_notation.m maps the exponent
  notation of plfit, r_plfit, the MATLAB generalized Pareto
  distribution, and the toolbox b and tau conventions. It runs a
  worked comparison on synthetic data.
• Vendored helpers yorkfit, nanmean, nanmedian, tocolumn, and
  renametimetabletimevar make the core workflow chain self-contained.
• Example sections in the help of 16 core-workflow functions, and help
  text in many private helper functions.
• New test suites:
  • test_fitcts: every stencil against the analytic derivative of an
    exponential recession, the midpoint times, both errors, and the
    getdqdt path;
  • test_fitopts: the overrides, both errors, precedence, scalar
    expansion, and the fitevents path;
  • test_peakfinder: empty input, directly and through islocalmax;
  • test_plfitb_hanel: the arguments the 'hanel' method passes to
    r_plfit;
  • test_corechain: eventtau, globalfit, fitphi, gpfitb,
    fitphidist, and aQbString;
  • test_dependencies: core-chain self-containment, declared
    products, the option outputs, the 'resolve' file copies, the
    plfitb known-external classification, the loadflow parked
    reader, and the Setup('dependencies') report;
  • test_version: every version source agrees.
• The demo scripts run under the suite (tests/test_demos.m). The
  theory demos skip when the Symbolic Math Toolbox is not installed.
• tests/octave_smoke.m, a plain script with bare asserts, runs the
  core workflow in GNU Octave and MATLAB: load the example data, then
  getevents, fitevents, and fitab with 'nls' and 'ols'.
• Plain unit tests of helpers that the toolbox vendors from matfunclib
  (nanmean, nanmedian, yorkfit, tocolumn, timetablereduce,
  nonnansegments, withcd, listfiles, and mpackagefolders) live
  in the matfunclib library test folders, next to the source functions.
  The matfunclib sources carry the help and fixes from the toolbox
  copies.
• tests/closenewfigs.m closes the figures a test opens, so a full
  suite run leaves zero open figures and keeps the figures a user had
  open.
• A MATLAB project definition file.
• TODO.md audits all work-in-progress signals: sandbox TODO lists,
  in-code markers, and disabled code blocks. It records their
  provenance, classification, and recommendations.
• This changelog and a minimal CONTRIBUTING.md.

Fixed

• The project .octaverc starts the toolbox with
  addpath(fullfile(pwd, 'toolbox')) and Setup('addpath'). It sourced
  Setup.m, which is a function file in toolbox/, so Octave started in
  the repository root did not add the toolbox to the path.
• eventfinder detects hydrograph troughs again. In 1.0.0 its
  Octave-compatible islocalmin and islocalmax wrappers called
  peakfinder with a threshold of 0, which dropped every minimum of
  positive flow and every negative local maximum of dQ/dt. The wrappers
  apply no threshold, which restores the v0.1.0 behavior of the MATLAB
  islocalmax and islocalmin functions, and the vendored peakfinder
  keeps a single interior peak when endpoints are excluded. Event counts
  change: the example data gives 287 events with the setopts defaults
  (327 in 1.0.0). The Kuparuk annual workflow gives 230 events and a
  global b of 1.3541, which matches the published 1.3540 (1.3519 in
  1.0.0).
• QtString and QtauString value labels showed an italic "e" with the
  latex interpreter and printed the wrong mantissa for a >= 10 (for
  example 1250000e^{3} for a = 1250). They build the label the way
  aQbString does.
• Getting Started lists the baseflow.setopts defaults, which direct
  name-value calls also use. It no longer
  labels plotdqdt deprecated. The eventfinder help and the
  setopts fitevents option list match their parsers.
• Continuous integration did not trigger: YAML parsed the
  space-separated branch list as one branch named "main dev". The
  workflow runs on every push and pull request to main and dev.
• The CI workflow pins its actions and uploads the JUnit results and
  Cobertura coverage as artifacts.
• Version metadata disagreed: at the 1.0.0 tag,
  baseflow.internal.version and toolbox/info.xml reported 0.1.0,
  DESCRIPTION reported 0.1.1, and CITATION.cff reported 1.0.0.
• +deps/peakfinder returns empty outputs for empty input. The
  empty-input branch assigned varargout, which left the named outputs
  peakInds and peakMags undefined.
• fitcts computes the C4 stencil as the fourth-order centered
  difference. The draft stencil subtracted and added Q(i+2), which
  cancelled the term, and never used Q(i-2).
• fitab applies fitopts. The parser read parser.Unmatched, which
  is always empty, and fitevents discarded the option.
• plfitb method 'hanel' passes 'cdat' to r_plfit, so r_plfit
  fits the continuous tau sample and does not bin it on an integer grid.
• plfitb method 'hanel' passes the exponent search bounds as
  'exp_min' and 'exp_max'. r_plfit ignores the 'alpha_min' and
  'alpha_max' names and used its default range of 0 to 5.
• nonnansegments returns the correct segments for data with leading or
  trailing nans. It errored or returned wrong indices for them. It keeps
  the nmin filter that eventfinder uses to remove short segments.
  For an all-nan vector, it returns empty (0-by-1) start, end, and
  length outputs.
• fitphi errors with baseflow:fitphi:unsupportedSolution for a
  solution pair with no derived formula. For example, when b2 is
  incompatible with the Rupp and Selker (2005) solution, the non-flat
  branch with soln1 'RS05' falls back to the Boussinesq (1903)
  late-time solution. The resulting pair RS05_BS03 has no formula.
  fitphi returned unassigned outputs for such a pair.
• ccdf reads its makeplot option from the parser results. The option
  caused an error on every call.
• Setup('install') runs the dependency check, which an early return
  skipped. In matlab -batch runs, it does not prompt before a
  re-install.
• fitab defaults the qtl method's polynomial order to 1.
• fitvts computes dt as a number, not as a duration value.
• struct2varargin assigns default values for its outputs.
• Field names in numfits and numevents match the current structures.
• citing_baseflow carries the JOSS citation (doi 10.21105/joss.05492).
  It states the BSD 3-Clause license apart from the citation request.
• toolbox/functionSignatures.json matches the parsers. The
  derivmethod choices for fitevents, getdqdt, and setopts are
  VTS, ETS, and CTS. The fitevents and setopts entries list
  ctsmethod and fitopts. The globalfit and setopts entries name
  drainagedensity.
• The BSD 3-Clause text in +internal/private/withcd.m had corrupted
  characters. It matches the license template.
• The suite builder skipped tests/test_withcd.m because the file
  declared no test output. That test now lives in matfunclib.
• test_conversions checked the b to k conversion against the
  reciprocal of the gpfit relation k = (1-b)/(b-2) at b = 1.5, where both
  equal 1. It checks b = 1.4 and the inverse k to b conversion.

Changed

• getevents, eventfinder, and wrapevents take their name-value
  defaults from baseflow.setopts('getevents'), the values that produced
  the published results and that the demos use. A direct call without an
  options struct uses fmax = 1, rmin = 1, rmnochange = true, and
  rmrain = true. In 1.0.0 these parsers used fmax = 2, rmin = 0,
  and rmrain = false (and rmnochange = false in getevents and
  eventfinder), so direct-call results change.
• globalfit defaults aquiferslope to 0, the setopts value. globalfit
  does not use this input, so results do not change.
• getevents and eventfinder require rmax > 1. The rmnochange
  filter counts each nan as a run of length 1, so rmax <= 1 rejected
  every day.
• aQbString, QtString, and QtauString share one help layout, input
  parser, and label format. Their symbolic labels come from
  baseflow.getstring.
• The runlength and isminlength help state that each nan is a run of
  length 1 and that callers use nan to break runs.
• getdqdt method 'CTS' (private/fitcts.m) is complete. It computes
  dQ/dt with six finite-difference stencils. The ctsmethod option
  selects B1 (backward, first order, the default), B2, F1, F2,
  C2, or C4, and fitcts validates the name.
• fitcts returns the rain on the input time vector. Its tq output
  is the input time vector, and a seventh output, tqmid, holds the
  stencil midpoint times.
• fitcts errors with baseflow:fitcts:nonuniformTime when the time
  vector is not uniform, because one constant time step applies to
  every stencil.
• plotdqdt defaults labelplot to false. Pass 'labelplot', true to
  draw the refline arrow annotations (the 1.0.0 default).
• baseflow.internal.dependencies analyzes the code live with
  matlab.codetools.requiredFilesAndProducts. It needs no saved
  dependencies.mat file.
• With no function name, baseflow.internal.dependencies analyzes the
  public API: the files that listfiles finds in +baseflow and
  +baseflow/+util.
• The baseflow.internal.dependencies options report, check, and
  resolve dependencies. The 'check' option compares the detected
  products with the DESCRIPTION MatlabProducts line and returns
  undeclared_products. The comparison ignores the Signal Processing
  Toolbox and Symbolic Math Toolbox entries that
  requiredFilesAndProducts reports for code that does not call them.
• The dependency report lists r_plfit in known_external whenever it
  analyzes plfitb, and does not count r_plfit as missing. It does not
  count the matfunclib getlist chain as missing either; only the parked
  readflow function in loadflow reaches it.
• Setup('dependencies') and Setup('install') run the live
  dependency check. The dependencies_checked preference is true only
  when no required file is missing. A failed check points to the
  baseflow issues page. On Octave, the check prints a message and skips
  the analysis.
• The DESCRIPTION file lists the dependencies. Depends lists the
  Octave packages, and a MatlabProducts line lists the required
  MATLAB toolboxes.
• The upstream project renamed the Octave package
  statistics-bootstrap to statistics-resampling. Setup and
  .octaverc load the installed name, and DESCRIPTION lists
  statistics-resampling and financial.
• README requirements list the Statistics and Machine Learning and
  Curve Fitting toolboxes. They also list the six Octave packages the
  toolbox loads.
• README states that Octave support covers the core analysis functions
  and the demos. It states that tests/octave_smoke.m verifies the core
  workflow on GNU Octave 11.3.0, and that the toolbox ran on Octave 8.2.0
  in 2023. It lists the install commands for the six Octave packages and
  their io and datatypes dependencies.
• README lists the four public functions that need data files or
  functions that the toolbox does not ship.
• README documents the external r_plfit requirement of the plfitb
  method 'hanel'. r_plfit has no license grant, so the toolbox does
  not vendor it.
• CONTRIBUTING.md states the input-parsing rule. Core analysis
  functions and demos use inputParser. The MATLAB-only +internal
  maintenance tooling and its private helpers may use arguments
  blocks.
• Core analysis functions and demos parse inputs with inputParser. On
  Octave 11.3.0, baseflow_demo_1, baseflow_demo_2, and
  baseflow_demo_3 run. baseflow_demo_kuparuk errors in trendplot
  at its fitlm call, and the theory demos need the symbolic package.
• +deps/arrow allocates numeric handles when gobjects is absent.
  Note: arrow errors on Octave at the MATLAB-only WarpToFill
  axes property, before it reaches that allocation. TODO.md records
  the incompatibility.
• Tests follow the no-magic-variables rule in CONTRIBUTING.md. A test
  names a value that a reader cannot identify, such as a non-obvious
  expected value or a tolerance. A comment says what each case
  checks. Option sweeps use parameterized test classes.
• Every version source reads 1.1.0: baseflow.internal.version,
  DESCRIPTION, CITATION.cff, .zenodo.json, and toolbox/info.xml.
• CITATION.cff cites the Zenodo concept DOI, which covers all versions.
• CITATION.cff lists both authors, Matthew G Cooper and Tian Zhou.
• The two-entry toolbox/+baseflow/private/functionSignatures.json
  (private fitets and fitvts) is deleted.
  toolbox/functionSignatures.json covers the public functions only.
• The contents page source (toolbox/docs/baseflow_contents.m) lists
  the public API, the +util functions, and
  baseflow.internal.version, and links the Examples page. It lists the
  private helpers under a separate heading.
• The Getting Started page source
  (toolbox/docs/baseflow_gettingStarted.m) spells aQbString
  correctly, names baseflow.internal.completions, lists
  privatefunction once, and lists the +util folder.
• Function help names the options that the parsers accept: bootreps
  in plfitb, ax in plotdqdt, and earlyqtls and lateqtls in
  cloudphi. The "See also" lines name fitphidist, aquifertrend,
  loadbasins, fitevents, and getdqdt in place of names that match
  no function, and they omit eventsplitter.
• The getdqdt help lists the VTS, ETS, and CTS methods and the
  options that its parser accepts: ctsmethod, etsparam, vtsparam,
  fitmethod, pickmethod, plotfits, and eventID.
• The Contents.m manifests and the m2html function pages are
  regenerated for 1.1.0.

Removed

• loadcalm, loadghcnd, loadgrace, mapbasins, and mapgages.
  They need data files and functions that the toolbox does not ship, and
  no demo, test, or core function uses them. The analysis project
  (mgcooper/arctic_baseflow) keeps working copies.
• The pending_decision field of the baseflow.internal.dependencies
  report and of Setup('dependencies').
• The Mapping Toolbox from the DESCRIPTION MatlabProducts line, the
  README requirements, Getting Started, and the Setup preferences. No
  toolbox function calls a Mapping Toolbox function.