Randomness 1.15.0

Added

• |sep: — choose what goes between multiple results. A roll that
  returns several things has always joined them with , , which is
  right inside a sentence and wrong everywhere else. |sep: sets the
  glue: `rdm:3#monster|sep:<br>|link` puts one monster per line,
  `rdm:3#monster|sep:<br>• |link` bullets them, and
  `rdm:3[[Rumours|line]]|sep: /\_` runs them together with
  slashes. It works on every multi-result inline roll — tag rolls,
  table rolls, line and block rolls — and under the dice: prefix.

  Everything after sep: is the separator, spaces and HTML included.
  Escapes cover what the surrounding syntax would eat: \n newline,
  \t tab, \_ space (needed at the end of a glue, since an inline
  span reaches the parser trimmed — sep: / joins with " /", and
  sep: /\_ is the one that gives /), \\ backslash. On a wikilink roll
  sep: goes outside the brackets — inside them a pipe already means
  "column pick" — and on a tag roll it is a segment like unique, so
  it comes before prop:, which still swallows the rest of the line.
  Separator text is never evaluated, so a glue containing [@table]
  prints as itself. Thanks to Gizmo734 for the request.

Fixed

• dice:*|folder=… rolls instead of silently doing nothing useful.
  The tagless * source — "any note matching these properties", with
  no tag constraint — never reached the tag-roll branch under the
  dice: prefix. It fell through to the formula translator and came
  out as the nonsense expression {*|folder=Bestiary|link}: no error,
  no roll, nothing to indicate the syntax was fine and the plumbing
  wasn't. The reference has claimed since these filters landed that
  they work under the compatibility prefix, so this was a documented
  feature that had never run. * now dispatches exactly like #tag,
  including the repetition prefix, |unique, and prop: templates,
  and a test pins the two prefixes to the same note on the same seed.

• Documentation: [@N table] was described as joining results with
  blank lines. It doesn't, and shouldn't — a repeated sub-table call
  runs its results straight together, because it is nearly always
  embedded in a sentence (The party of [@4 hero] sets out.) where an
  injected blank line would be wrong. The separator is the author's
  call, via >> implode: bare for , , or \n / \n\n / <br> /
  ;\_ for anything else. The reference's "Call variations",
  "Filters" and "Repetitions" sections now say so, with a new note on
  the one place separation is automatic — a whole-file MaxReps:
  roll, whose reps are standalone blocks and so are blank-line
  separated. Behaviour is unchanged; the corpus tests now pin every
  one of these joins so the docs can't drift again.

  Two smaller errors went with it: a filter's glue is trimmed and
  never quoted, so the documented implode ", " would have put the
  quote marks in the result and implode , loses its trailing space.
  Both examples are corrected, and \_ is documented as the way to
  end a glue with a space. Thanks to claudermilk for the report.

Randomness 1.14.0

Changed

• A bare dice formula in an rdm: span now rolls. `rdm:2d10`
  used to render the literal text "2d10" — no roll, no error, nothing
  to suggest anything was wrong. It was the trap waiting for anyone
  converting dice: spans, where the formula is the whole
  expression. An inline call that is nothing but a dice formula is now
  treated as `rdm:{2d10}`, and an omitted die count is filled in
  (`rdm:d20`) for the same reason.

  The test is deliberately narrow: the whole expression must be a
  formula, so `rdm:you take 2d6` is still literal text, and no
  expression containing generator syntax ([@table], {…}, #tag,
  |) can match. Bracing is native, so `rdm:3d6>=10` means what
  `rdm:{3d6>=10}` means — 1 when the total reaches 10 — rather
  than the success-counting a dice: span would read it as. Thanks to
  Anna_B_Meyer for walking into it.

Randomness 1.13.0

Added

• A "Show dice formula" setting. Inline rolls can now show what
  was rolled next to the result — 2d6+3 → 11 instead of a bare 11
  — without adding a flag to every call. Settings → Randomness → Show
  dice formula, off by default. It covers both prefixes, so
  `rdm:{2d6+3}` and `dice:2d6+3` read the same (the {…}
  wrapper is dropped from the display), and it only touches rolls that
  actually rolled dice, so `rdm:[@Weather]` is unaffected. A
  dice: span still overrides it per roll: |form shows the formula
  with the setting off, |noform hides it with the setting on, and
  |text(label) wins over both. The formula is display only — locking
  a roll commits the result, as before. |noform had been accepted
  and ignored until now. Thanks to Anna_B_Meyer for the request.

Randomness 1.12.0

Added

• Roll more than one note at a time. Tag and folder rolls take the
  same repetition prefix the [[Note|line]] rolls have always had —
  `rdm:3#monster|link` or `rdm:{1d4}#rumour` — and the new
  |unique segment draws them as a deck so no note comes up twice:
  `rdm:3#monster|unique|link` is three different monsters.
  Comma-joined, works with prop: templates
  (`rdm:3*|folder=Bestiary|unique|prop:{{name}} (CR {{cr}})`), and
  passes through the dice: prefix. |unique has to come before
  prop:, which swallows the rest of the line. Asking a unique roll for
  more notes than match gives you all of them rather than an error.
  Thanks to Gizmo734 for the suggestion.

Randomness 1.11.0

Added

• api.rollFormula() and api.formulas() — saved dice formulas are
  now reachable from scripts. Formula aliases (Settings → Randomness
  → Dice formula aliases, and the dice tray's ★ button) resolved only
  for inline dice: spans and the tray, so a Templater / Meta Bind /
  QuickAdd script had no way to invoke one — the reported case was an
  initiative button that rolls exploding step dice and writes the
  result to a monster note's frontmatter. api.rollFormula("sneak")
  now rolls the saved alias (matched trimmed and case-insensitively,
  exactly as inline), and any unmatched string is rolled as a raw
  formula in the full Dice Roller grammar — modifiers, special dice,
  [[Note^id]] table rolls, #tag rolls. api.formulas() lists the
  saved aliases. roll(), rollUnscoped() and rollExpression() are
  unchanged: they still do not resolve aliases, so an expression
  sharing a name with one keeps its existing meaning. API version
  1.2.0 → 1.3.0 (additive).
• Print a rolled note's properties, not just a link to it. Tag and
  folder rolls could always filter on frontmatter
  (rdm:*|folder=Bestiary|cr=3|link); now they can output it. End the
  call with prop: and the rest is a template:
  `rdm:*|folder=Bestiary|prop:{{link}} — CR {{cr}}, {{hp}} HP` →
  [[Bestiary/Bog Hag|Bog Hag]] — CR 3, 45 HP. One note is rolled and
  the whole template is filled from it, so the values can never belong
  to different monsters the way two separate `rdm:` spans would.
  prop:cr is shorthand for prop:{{cr}}. {{link}}, {{linkpath}},
  {{path}} and {{name}} describe the note itself; every other
  placeholder is a frontmatter key, and naming one also requires it, so
  a note missing that property is never picked. Works under the dice:
  prefix too.
• api.randomNote() now returns the note's frontmatter, so
  Templater and dataviewjs callers don't need a second metadataCache
  lookup. API version 1.3.0 → 1.4.0 (additive).

Randomness 1.10.0

Added

• count(Table) inside {...} — how many items a table has.
  Mostly useful for rolling an index that stays valid as a table
  grows: {1d{count(npcs)}} instead of hard-coding 1d3 and
  forgetting to update it when you add a row. The argument is a table
  name rather than a value, so dotted column names
  (count(npcs.Job)), quoted names with spaces, and interpolated
  names (count({$whichTable})) all work. It counts items, so a
  lookup table with two range rows counts 2, not the span of its dice
  formula; counting a table that doesn't exist is an error, the same
  as rolling one.

Documentation

• How to pull several columns from the same rolled row. Each
  column of a markdown table is its own rollable table, so
  [@npcs.Job] and [@npcs.Name] land on different people — which
  looks like a bug the first time you write a sentence with two of
  them in it. The fix is to roll the row number once and pick by
  index: The [#{row=1d{count(npcs)}} npcs.Job] was [#{$row} npcs.Secret]. Written up in the reference (Rolling on note
  content → One row, several columns) and in guide note 02, along
  with the reason it can't be split across two `rdm:` spans:
  every inline span is evaluated on its own, so variables don't carry
  from one to the next.
• The [#table] current-index form — pick the item at the position
  of the item currently being rendered, for cross-indexing parallel
  tables — is now in the call-variations table, and the function list
  (if, max, min, round, substr, …) is documented at all.

Randomness 1.9.2

Fixed

• The dice animation now shows the total when a roll has a
  modifier. The graphical dice overlay only drew its = total line
  when a roll had more than one die, so a single die plus a modifier —
  1d20 + 5, 1d6 - 3 — animated the die and showed nothing else.
  The die face was the only number on screen, which read as "the
  modifier was ignored"; the roll itself was always correct, and the
  dice tray's result panel had the right number the whole time. The
  total is now shown whenever it says something the faces don't: more
  than one die, or a flat modifier or dropped die that makes the total
  differ from the dice on screen. A plain 1d20 still animates
  without a redundant total. Thanks to Anna_B_Meyer for the report.

Randomness 1.9.1

Added

• names.rdm in the Fantasy Hub bundle — a much larger NPC name
  pool. Dictionary tables keyed by race_gender
  ([#elf_female TF-PersonName], or rollUnscoped("TF-PersonName", { dictKey: "elf_female" })), plus TF-FirstName and TF-Surname
  for the halves. Surnames blend a curated list with a
  prefix + suffix compound table, giving roughly 16,000–47,000 full
  names per race/gender bucket instead of a few hundred. Covers human,
  elf, half-elf, half-orc, gnome and goblin; half-elves draw from both
  parent cultures. Town Forge's place templates use it to keep every
  NPC in a settlement distinct.

Changed

• Bigger built-in portrait name tables. The names behind
  api.portraits.roll() came from lists of 9–15 entries per race, so
  a town full of NPCs would occasionally hand out the same name twice
  (human males had only 210 possible full names). Each list is now
  32–48 entries — human is 48 × 48 — cutting the chance of a repeat
  across a 40-NPC town from roughly 36% to 4%. This is the fallback
  path now that the Fantasy Hub ships names.rdm.
• Fantasy Hub Personality grown from 21 to 61 beats, so
  characters stop sharing a personality line two or three times per
  generated town.

Randomness 1.9.0

Added

• Reroll button on randomness codeblocks. Fenced ```randomness
  rollers now render a Reroll button beside their prompt controls, so
  an NPC or table roller embedded in a note can be re-rolled in place —
  the same affordance the .rdm file view and inline calls already
  had. Prompt selections are preserved across rerolls, and the button
  forces a fresh roll even when Stable codeblock seeds is on (the
  new result then persists across passive re-renders until the next
  reroll).

Randomness 1.8.0

Changed

• Tag and property |link rolls now show the note's name, not its
  full path. `rdm:#tag|link` and property/folder variants like
  `rdm:*|spelllevel=1st|link` render the wikilink as the note's
  name — Burning Hands rather than Spells/Level 1/Burning Hands —
  while still linking to the correct note wherever it lives in the
  vault. This is the new default for every |link roll.

Added

• |linkpath flag for link rolls that keeps the full vault path
  visible, for anyone who preferred the previous display. Works under
  the dice: compatibility prefix too (`dice:#tag|linkpath`).