symbulator 0.6.18

No change to any answer: nothing under symbulator/ moved but one comment
and the version number. This release exists so that PyPI's page carries the
new classifier and README.

• Out of beta (#471). Symbulator 9 left the beta it had been in since its
  public launch on 28 Aug 2026; the classifier is now
  Development Status :: 5 - Production/Stable, and Python 3.13 and 3.14 are
  listed beside 3.9–3.12.
• The repository, ready for review (#470). A GitHub Actions workflow runs
  the suite on Python 3.9 to 3.14, with the ahkab ground-truth tests in a job
  of their own; CONTRIBUTING.md, CODE_OF_CONDUCT.md and CITATION.cff
  added; the JOSS paper and its fact-check in paper/.
• README. The Tests section names the command that counts the suite
  instead of a count that had gone stale ("48 tests across six files"; there
  are 26 files), and says why the ahkab and IPython tests skip in a plain
  install. Two sentences that still said a side of the circuit behind a port
  is "reported floating" now say it is given a local reference, as it has
  been since 0.5.32. The quick start's transient uses 1'u, so the printed
  answer matches its comment exactly. Install names the Python versions and
  the extras.

On PyPI: https://pypi.org/project/symbulator/0.6.18/ (wheel sha256 f93423548d1d...).

symbulator 0.6.17

Added

• evaluate(result, "expression") and solve(result, equations, unknowns) -- the app's Evaluate box and Solve card as package
  functions, so a notebook can do what the calculator did with a solved
  circuit's variables: evaluate(res, "vo/vs"), evaluate(res, "vc", conditions=["t = 2"]), solve(res, ["im(ze) = 0"], ["w"], conditions=["w > 0"], real_only=True). Both return SymPy and read
  answer names in every spelling (i_r1, ir1, IR1), with the same
  shorthand a circuit value gets (2'k, ^, u(t), {...} in FD).
  evaluate also answers pf(...), s2t(...)/t2s(...) and
  limit(...) with the answers substituted in first, takes a condition at
  infinity as a limit, and on a th() result knows the load answers
  irl, vrl and prl in the variable load. solve reads a pinning
  condition (R_3 = 10) as a substitution and a comparison as a filter on
  the roots, and real_only=True is the calculator's solve() against
  its cSolve(). A port() result and a plain mapping of names to values
  are accepted too, the latter being how er()'s single expression is
  handed over. New module symbulator/cards.py; nothing else changes.
• Result.desc, the circuit a result was solved from. It lets
  cards.py derive an element's voltage drop from the two nodes it spans,
  which in fd and tr is not among values -- the third level is computed
  for dc and ac only -- so vc names an answer on the app's page while
  the package had no such key.
• Notebooks. notebooks/books/ holds one executed notebook per built-in
  example book (the Course's problems, the two textbook samplers, the
  Showcase and the monograph's exemplars), A Baker's Dozen, and the
  Manual's circuits, each generated from the app's own files and compared
  with the app answer by answer. Not part of the installed package.

Fixed

• th(), er() and port() accept a frequency written as a string, as
  ac() always has: th(circuit, "2", "0", domain="ac", omega="2*pi*2e3")
  reached the stamping code as a str and failed there with can't multiply
  sequence by non-int of type ImaginaryUnit, naming nothing the caller had
  typed. The one place all of them pass through now sympifies it.

On PyPI: https://pypi.org/project/symbulator/0.6.17/ (wheel sha256 c7108cbd0395...).