connorferster/handcalcs v2.0.0rc3

Pre-release: v2.0.0rc3

on GitHub

Pre-release

This is the third pre-release of v2.0.0, to get people testing it and using it in real projects. Bugs are expected, so please raise issues.

Install it with pip install "handcalcs>=2.0.0rc3"; a plain pip install handcalcs still installs 1.x until 2.0.0 is released.

Changes since rc2

Breaking changes

• Comment commands are read, never run (#354). In the assignment form
  (# hc: multiline = False), each value must be a literal: a number, a string, True,
  False or None. An expression such as # hc: format_code = "." + "2f" is now an
  error.
• A value keeps the -f format of the line that calculated it (#357), on the lines
  that substitute it: after DCR = C_f / C_r # hc: -f .3f, ok = DCR <= 1.0 shows
  0.833 <= 1.0, not 0.83292 <= 1.0. A line's own -f still wins, and a name
  assigned again without -f goes back to the format in effect. Calculations with an
  inline -f can show different numbers in the lines below it
  (details).
• Light markup in comments (#356) changes how some existing comments render:
  **bold** and *italic* are now styled (the Markdown renderer used to escape them),
  \* is a literal *, and in the LaTeX and Markdown renderers a pair of $ is math.
  Formulas (w*L**2/8) and prices ("costs $5 to $10") stay as written; see Features.
• The ASCII renderer spaces comparison operators (#357) like the arithmetic ones:
  DCR <= 1.0, not DCR<=1.0, as the HTML renderer does.

Features

• Math in comments (#350, #356): $...$ in a heading, comment line or inline
  comment is LaTeX math in the LaTeX and Markdown renderers, as in v1. \$ is a literal
  $. Turn it off with LatexRenderer(comment_math=False) or
  MarkdownRenderer(comment_math=False)
  (details).
• Bold, italic and code in comments (#356): **bold**, *italic* and `code`
  are styled in every renderer: Markdown, LaTeX, HTML, Word (including inline comments)
  and Excel (rich text). The ASCII renderer shows them as written.
• @handcalc(..., docstring=False) (#339, #355): leave the function's docstring out
  of the render. It is rendered by default, as before.
• find_inputs(..., detail=True) (#358): each input as an InputSpec with its
  value, the units it is written in ("mm", "kN / m"), the comment at the end of its
  line and its line number, for building a form or an input file from a script
  (details).
• handcalcs.testing (#349): a renderer-neutral feature suite. FEATURE_CASES
  holds a calculation for each feature with the names its output must show and hide,
  and check_case(renderer, case) checks a renderer against one, so renderers in other
  packages can test every feature
  (details).

Fixes

• # hc: -s and # hc: -m work after forallpeople's
  si.environment(..., top_level=True) (#354). Before, s (seconds) and m (metres)
  in builtins made them valid Python expressions, and they were silently ignored:
  after # hc: -i, everything that followed stayed hidden.
• Two or more multi-line calculations in a row (after a standalone # hc: -m)
  render one below the other, instead of being joined into one row
  (= 0.8 M_p = 7 * 8), in every renderer (#354).
• In the Excel renderer, a comment command inside an if block (# hc: -s, -i,
  -f) now applies to the lines after the block, as in the other renderers.
• Docs: the migration guide explains that a standalone comment command stays in effect
  for the lines after it (#359).
• Development: uv.lock updates certifi, nbconvert and tornado for security fixes
  (#360). The package's own requirements are unchanged.

The forallpeople side of #352 (a unit with a factor, such as a tonne, shown with the wrong value in a substitution) is fixed in forallpeople, not in handcalcs: connorferster/forallpeople#142 (open; it ships with the next forallpeople release).