Architecture — ROARSounds

ROARSounds — code architecture

Runtime flow

PLAYER_LOGIN
  -> NormalizeName(UnitName("player"))
  -> GetRoarIDForUnit("player")
       -> UnitRace + UnitSex -> raceMale/raceFemale
       -> alias Scourge -> Undead
  -> cache player's matching built-in sound path

UNIT_MODEL_CHANGED for "player"
  -> refresh player sound mapping

CHAT_MSG_TEXT_EMOTE(text=arg1, sender=arg2)
  -> IsRoarText(text) checks English substrings " roars", " roar"
  -> NormalizeName(sender) strips "-realm" suffix
  -> per-normalized-sender GetTime() gate (0.3 seconds)
  -> if sender equals player:
       PlaySoundFile(cached player path), if mapped
     else:
       FindUnitBySender:
         target -> mouseover -> focus -> raid1..40 OR party1..4
       GetRoarIDForUnit(matched token)
       PlaySoundFile(path) if in 16-key mapping

No current CHAT_MSG_ADDON, SendAddonMessage, COMBAT_LOG detection, repeating timer, saved variables, or slash-command interface.

Identity and supported assets

Sound map: 16 fixed combinations of eight race tokens (Dwarf, Gnome, Human, NightElf, Orc, Tauren, Troll, Undead) and two sex flags (2→Male, 3→Female). Scourge is normalized to Undead. Paths have the form:

Sound\\Character\\PlayerRoars\\CharacterRoars<Race><Male|Female>.wav

There are no physical WAV files committed to the repo. All sound files are expected to exist inside the game client and must be tested against the exact client version. Unmapped races yield no playback.

FindUnitBySender only sees unit-token data exposed by the game. It does not enumerate every nearby player, and focus support or other API behavior may vary by client.

Deduplication and edge cases

lastRoarTimeBySender maps normalized sender name to last qualifying text-emote event time. If the same name triggers another matching text emote within 0.3s, it returns before playback. The state is session-local.

Crucially, it records that timestamp before verifying the other player's unit token or sound map. An unmatched/silent roar can therefore suppress a later resolvable event within that interval. Normalizing different name–realm pairs to the same bare name can conflate two speakers, including local and remote characters.

The English substring check may match a phrase where the word roar appears but a full /roar action was not performed, and it may miss translated emote messages. These are testable source-level limitations, not confirmed player-facing bug reports.

Packaging, API target and historical communication experiment

Actual source files: ROAR/ROAR.lua, ROAR/ROAR.toc.
Manifest title ROAR, author Babuni.dev, Interface: 30300; no explicit Version or SavedVariables.
ROAR.zip is present as a separate binary archive; parity with committed source not independently verified here.
Lua uses a legacy callback style with event, arg1, arg2 globals.
On 2026-04-20/22 the project briefly used addon messaging to share roar sounds across party/raid/guild. That implementation was removed by fcc2be8 and does not describe the present architecture.
Interface 30300 is 3.3-era metadata; the source does not establish functioning support on Vanilla 1.12 or later modern clients.

The working audio path selection uses PlaySoundFile; sound output volume, availability, client UI settings and error handling have not been exercised in this audit.

Document information