"""The plan: a fake community drawn from a spec and its seed, the same on every run (tests/test_gen.py pins its hash). A plan is a list of steps, each {n, phase, round, actor, verb, args, ref}. Accounts are named "/", objects by a ref ("p12"); the seeder turns refs into URIs as it goes. Phases run in order: accounts every account, its profile, PrivaPub's roots and personas groups PrivaPub circles and communities, Lemmy communities graph follows (requests to locked accounts accepted, left pending or rejected), circle and community joins content top-level posts, then reply rounds, each waiting for the round before to arrive interact likes, boosts, reactions, votes, bookmarks mutate edits, deletes, blocks, mutes, reports The planner only asks a platform for what its driver can do (`caps`) and only lets an account touch what it could see. """ import hashlib import json import os import sys from core.rng import Rng GENERATOR_VERSION = 1 HOSTS = {"privapub": "privapub.test", "gts": "gts.test", "mastodon": "mastodon.test", "misskey": "misskey.test", "sharkey": "sharkey.test", "akkoma": "akkoma.test", "lemmy": "lemmy.test", "hollo": "hollo.test"} SHORT = {"privapub": "pp", "gts": "gt", "mastodon": "ms", "misskey": "mk", "sharkey": "sk", "akkoma": "ak", "lemmy": "lm", "hollo": "ho"} MICRO = ("privapub", "gts", "mastodon", "misskey", "sharkey", "akkoma", "hollo") CAPS = { "privapub": {"post", "reply", "cw", "media", "poll", "like", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "edit", "vote", "react", "quote", "profile", "circle", "community", "located"}, "gts": {"post", "reply", "cw", "media", "poll", "like", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "edit", "vote", "profile"}, "mastodon": {"post", "reply", "cw", "media", "poll", "like", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "edit", "vote", "quote", "profile"}, "misskey": {"post", "reply", "cw", "media", "poll", "like", "react", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "vote", "quote", "profile"}, "sharkey": {"post", "reply", "cw", "media", "poll", "like", "react", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "vote", "quote", "profile", "edit"}, "akkoma": {"post", "reply", "cw", "media", "poll", "like", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "edit", "vote", "react", "quote", "profile"}, "lemmy": {"thread", "comment", "upvote", "downvote", "dm", "delete", "community", "edit", "block"}, "hollo": {"post", "reply", "cw", "media", "poll", "like", "boost", "bookmark", "follow", "block", "mute", "report", "dm", "delete", "edit", "vote", "quote", "react", "profile"}, } FIRST = ["ada", "bo", "cleo", "dario", "elif", "femi", "gaia", "hiro", "ines", "jun", "kai", "lia", "milo", "nadia", "omar", "pia", "quinn", "rosa", "sami", "tea", "umbe", "vera", "wren", "xavi", "yuki", "zoe", "aldo", "bea", "cosimo", "dora", "enzo", "flor", "gigi", "hana", "ivo", "jade", "kira", "luca", "mara", "nico"] WORDS = { "en": "the garden river morning bread market neighbour bicycle library rain coffee window festival map bridge " "lantern harvest tide mountain postcard quiet bright small warm early late again together tomorrow".split(), "it": "il giardino fiume mattina pane mercato vicino bicicletta biblioteca pioggia caffè finestra festa mappa ponte " "lanterna raccolto marea montagna cartolina tranquillo luminoso piccolo caldo presto tardi insieme domani".split(), "ja": "庭川朝パン市場隣人自転車図書館雨コーヒー窓祭り地図橋提灯収穫山葉書静か明るい".split(), "ar": "حديقة نهر صباح خبز سوق جار دراجة مكتبة مطر قهوة نافذة مهرجان خريطة جسر فانوس حصاد جبل هادئ".split(), } EMOJI = ["🌱", "🚲", "☕", "🌧", "🎉", "📚", "🌉", "🍞", "🗺", "🏔"] TAGS = ["town", "pasture", "garden", "market", "rain", "bikes", "books", "festival"] REACTIONS = ["🎉", "👍", "❤", "🔥", "😂"] PLACES = [(45.4642, 9.19, "Milano"), (45.07, 7.6869, "Torino"), (41.9028, 12.4964, "Roma"), (48.8566, 2.3522, "Paris"), (52.52, 13.405, "Berlin"), (35.6762, 139.6503, "Tokyo")] def load(path): if not os.path.isabs(path) and not os.path.exists(path): path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "specs", path if path.endswith(".json") else f"{path}.json") with open(path) as f: return json.load(f) class Planner: def __init__(self, spec): self.spec = spec self.rng = Rng(spec["seed"], f"{spec['name']}/v{GENERATOR_VERSION}") self.steps = [] self.accounts = {} # key -> account dict self.follows = {} # follower key -> set of followee keys (accepted) self.objects = {} # ref -> object dict self.groups = {} # ref -> group dict self.n = 0 def step(self, phase, actor, verb, args, ref=None, round_=0): self.n += 1 s = {"n": self.n, "phase": phase, "round": round_, "actor": actor, "verb": verb, "args": args} if ref: s["ref"] = ref self.steps.append(s) return s @staticmethod def acct(key): platform, user = key.split("/", 1) return f"{user}@{HOSTS[platform]}" def text(self, rng, lang, n_words, ref): words = WORDS.get(lang, WORDS["en"]) body = " ".join(rng.choice(words) for _ in range(n_words)) if rng.chance(0.3): body += " " + rng.choice(EMOJI) return f"{body} ({ref})" # -- accounts def plan_accounts(self): rng = self.rng.sub("accounts") prof = self.spec.get("profiles", {}) langs = prof.get("langs", {"en": 70, "it": 15, "ja": 8, "ar": 7}) used = set() for platform, conf in self.spec["peers"].items(): if platform == "privapub": for r in range(conf.get("roots", 1)): # root logins and passwords are canaries: neither may ever appear on another server or in a log line canary = hashlib.sha256(f"{self.spec['seed']}/root{r}".encode()).hexdigest() root = f"rootcanary{r + 1:02d}{canary[:8]}" password = f"Canary-{canary[8:28]}-9!" for _ in range(rng.span(conf.get("personas", [2, 3]))): self._account(rng, platform, used, prof, langs, root=root, password=password) else: for _ in range(conf.get("accounts", 3)): self._account(rng, platform, used, prof, langs) def _account(self, rng, platform, used, prof, langs, root=None, password=None): while True: name = f"{rng.choice(FIRST)}_{SHORT[platform]}{rng.between(1, 99):02d}" if name not in used: used.add(name) break key = f"{platform}/{name}" lang = rng.weighted(langs) locked = platform != "lemmy" and rng.chance(prof.get("locked", 0.2)) account = { "platform": platform, "username": name, "password": password or f"Town-{hashlib.sha256(name.encode()).hexdigest()[:12]}-7!", "name": name.split("_")[0].capitalize() + f" of {platform}", "lang": lang, "bio": self.text(rng, lang, rng.span(prof.get("bioWords", [5, 20])), f"bio-{name}"), "fields": [[f"field{i}", rng.choice(WORDS["en"])] for i in range(rng.span(prof.get("fields", [0, 3])))], "locked": locked, "bot": rng.chance(prof.get("bot", 0.05)), "avatar_seed": f"avatar-{name}" if rng.chance(prof.get("avatar", 0.9)) else None, "header_seed": f"header-{name}" if rng.chance(prof.get("header", 0.5)) else None, "root": root, } self.accounts[key] = account self.step("accounts", key, "account", account) def keys(self, platform=None, micro=False): return [k for k, a in self.accounts.items() if (platform is None or a["platform"] == platform) and (not micro or a["platform"] in MICRO)] # -- groups def plan_groups(self): rng = self.rng.sub("groups") conf = self.spec["peers"] personas = self.keys("privapub") for i in range(conf.get("privapub", {}).get("circles", 0)): owner = rng.choice(personas) ref = f"circle{i + 1}" self.groups[ref] = {"kind": "circle", "owner": owner, "members": {owner}, "name": f"{ref}_{rng.between(100, 999)}"} self.step("groups", owner, "group", {"kind": "circle", "name": self.groups[ref]["name"]}, ref) for i in range(conf.get("privapub", {}).get("communities", 0)): owner = rng.choice(personas) ref = f"ppcomm{i + 1}" self.groups[ref] = {"kind": "community", "host": "privapub", "owner": owner, "members": set(), "name": f"square{i + 1}_{rng.between(100, 999)}"} self.step("groups", owner, "group", {"kind": "community", "name": self.groups[ref]["name"]}, ref) for i in range(conf.get("lemmy", {}).get("communities", 0)): lemmings = self.keys("lemmy") if not lemmings: break owner = rng.choice(lemmings) ref = f"lmcomm{i + 1}" self.groups[ref] = {"kind": "community", "host": "lemmy", "owner": owner, "members": {owner}, "name": f"plaza{i + 1}_{rng.between(100, 999)}"} self.step("groups", owner, "group", {"kind": "community", "name": self.groups[ref]["name"]}, ref) # -- the graph def plan_graph(self): rng = self.rng.sub("graph") g = self.spec.get("graph", {}) micro = self.keys(micro=True) popularity = {k: 1 + rng.below(5) for k in micro} for follower in micro: n = rng.span(g.get("follows", [3, 8])) candidates = [k for k in micro if k != follower] weights = {k: popularity[k] * (3 if k.split("/")[0] == "privapub" or follower.startswith("privapub/") else 1) for k in candidates} chosen = set() while len(chosen) < min(n, len(candidates)): chosen.add(rng.weighted({k: w for k, w in weights.items() if k not in chosen})) for target in sorted(chosen): self._follow(rng, g, follower, target) # mutuals: some of those followed follow back for target in sorted(chosen): if rng.chance(g.get("mutual", 0.4)) and follower not in self.follows.get(target, set()): self._follow(rng, g, target, follower) # circles: members are local siblings' neighbours and remote followers of the owner for ref, grp in sorted(self.groups.items()): if grp["kind"] == "circle": pool = sorted(k for k in micro if k != grp["owner"] and k.split("/")[0] in ("privapub", "gts", "mastodon", "akkoma")) for member in rng.sample(pool, min(len(pool), self.spec.get("circleMembers", 4))): grp["members"].add(member) self.step("graph", member, "join", {"group": ref}) self.step("graph", grp["owner"], "approve", {"group": ref, "member": member}) elif grp["host"] == "privapub": pool = sorted(k for k in self.accounts if k != grp["owner"]) for member in rng.sample(pool, min(len(pool), self.spec.get("communityMembers", 6))): grp["members"].add(member) self.step("graph", member, "join", {"group": ref}) else: pool = sorted(k for k in self.accounts if k != grp["owner"] and self.accounts[k]["platform"] in ("privapub", "lemmy", "mastodon", "gts")) for member in rng.sample(pool, min(len(pool), self.spec.get("communityMembers", 6))): grp["members"].add(member) self.step("graph", member, "join", {"group": ref}) def _follow(self, rng, g, follower, target): self.step("graph", follower, "follow", {"target": target}) if self.accounts[target]["locked"]: fate = rng.weighted({"accept": 1 - g.get("pending", 0.1) - g.get("rejected", 0.05), "pending": g.get("pending", 0.1), "reject": g.get("rejected", 0.05)}) if fate == "accept": self.step("graph", target, "accept", {"target": follower}) elif fate == "reject": self.step("graph", target, "reject", {"target": follower}) return else: return self.follows.setdefault(follower, set()).add(target) def followers(self, key): return {f for f, targets in self.follows.items() if key in targets} def can_see(self, viewer, obj): if viewer == obj["author"]: return True v = obj["visibility"] if v in ("public", "unlisted"): return True if v == "followers": return viewer in obj["followers"] or viewer in obj["mentions"] if v == "direct": return viewer in obj["mentions"] if v == "circle": return viewer in self.groups[obj["group"]]["members"] if v == "community": return True return False # -- content def plan_content(self): rng = self.rng.sub("content") c = self.spec.get("content", {}) mix = c.get("mix", {"text": 40, "cw": 8, "tags": 10, "mention": 10, "image": 8, "poll": 6, "quote": 5}) vis = c.get("visibility", {"public": 55, "unlisted": 15, "followers": 15, "direct": 10, "circle": 5}) micro = self.keys(micro=True) top = [] for i in range(c.get("posts", 60)): author = rng.choice(micro) platform = author.split("/")[0] obj = self._post(rng, f"p{len(self.objects) + 1}", author, platform, mix, vis, c, top) if obj: top.append(obj) # community posts: Lemmy threads, posts into the PrivaPub community for ref, grp in sorted(self.groups.items()): if grp["kind"] != "community": continue posters = sorted(m for m in grp["members"] | {grp["owner"]} if (grp["host"] == "lemmy" and m.startswith("lemmy/")) or (grp["host"] == "privapub" and m.startswith(("privapub/", "lemmy/")))) for _ in range(c.get("communityPosts", 4)): if not posters: break author = rng.choice(posters) oref = f"p{len(self.objects) + 1}" obj = {"ref": oref, "author": author, "visibility": "community", "group": ref, "mentions": set(), "followers": set(), "kind": "thread", "round": 0} self.objects[oref] = obj lang = self.accounts[author]["lang"] self.step("content", author, "post", {"visibility": "community", "group": ref, "title": f"{rng.choice(WORDS['en']).capitalize()} thread {oref}", "text": self.text(rng, lang, rng.between(8, 30), oref)}, oref) top.append(obj) # reply rounds rounds = c.get("replyRounds", 3) previous = top for rnd in range(1, rounds + 1): current = [] for parent in previous: if not rng.chance(c.get("replies", 0.45) if rnd == 1 else c.get("deeperReplies", 0.35)): continue repliers = [k for k in self.accounts if k != parent["author"] and self.can_see(k, parent) and self._can_reply(k, parent)] if not repliers: continue for _ in range(1 if rnd > 1 else rng.between(1, 2)): replier = rng.choice(repliers) obj = self._reply(rng, replier, parent, rnd) if obj: current.append(obj) previous = current def _can_reply(self, key, parent): platform = key.split("/")[0] if parent["visibility"] == "community": grp = self.groups[parent["group"]] return key in grp["members"] or key == grp["owner"] if platform == "lemmy": return False if parent["visibility"] == "circle": return platform == "privapub" or key in self.groups[parent["group"]]["members"] if parent["visibility"] == "located": return False return "reply" in CAPS[platform] def _post(self, rng, ref, author, platform, mix, vis, c, top): caps = CAPS[platform] vweights = {k: w for k, w in vis.items() if k in ("public", "unlisted", "followers", "direct") or (k in ("circle", "located") and k in caps)} visibility = rng.weighted(vweights) circles = [] if visibility == "circle": circles = [r for r, g in sorted(self.groups.items()) if g["kind"] == "circle" and g["owner"] == author] if not circles: visibility = "followers" # circle and located posts go through /clientapi, which takes text, a warning and a title only plain = visibility in ("circle", "located") kind = rng.weighted({k: w for k, w in mix.items() if (k in ("text", "tags", "mention") or k in caps or (k == "image" and "media" in caps)) and not (k == "quote" and not top) and not (plain and k in ("image", "poll", "quote", "mention", "tags"))}) lang = self.accounts[author]["lang"] args = {"visibility": visibility, "text": self.text(rng, lang, rng.between(4, 40), ref), "language": lang} obj = {"ref": ref, "author": author, "visibility": visibility, "mentions": set(), "kind": kind, "round": 0, "followers": set(self.followers(author)), "group": None} if visibility == "circle": obj["group"] = circles[0] args["group"] = circles[0] if visibility == "located": lat, lng, place = rng.choice(PLACES) args["location"] = {"lat": lat, "lng": lng, "range_km": rng.choice([2, 5, 20])} args["text"] += f" near {place}" if visibility == "direct" or kind == "mention": others = [k for k in self.keys(micro=True) if k != author] count = rng.between(1, 3) if visibility == "direct" else 1 obj["mentions"] = set(rng.sample(others, count)) args["mentions"] = sorted(obj["mentions"]) if kind == "cw": args["cw"] = f"{rng.choice(WORDS['en'])} {rng.choice(['spoilers', 'food', 'weather', 'politics'])}" obj["cw"] = args["cw"] elif kind == "tags": args["tags"] = rng.sample(TAGS, rng.between(1, 3)) elif kind == "image": args["media"] = [{"kind": "image", "seed": f"{ref}-{i}", "alt": f"picture {i + 1} of {ref}", "sensitive": rng.chance(0.15)} for i in range(rng.between(1, 3))] elif kind == "poll": opts = rng.sample(WORDS["en"], rng.between(2, 4)) args["poll"] = {"options": opts, "multiple": rng.chance(0.3), "expires_in": 86400} obj["options"] = len(opts) elif kind == "quote": candidates = [o for o in top if o["visibility"] in ("public", "unlisted") and o["kind"] != "quote"] if not candidates or "quote" not in caps: kind = obj["kind"] = "text" else: target = rng.choice(candidates) args["quote"] = target["ref"] obj["quote"] = target["ref"] self.objects[ref] = obj self.step("content", author, "post", args, ref) return obj def _reply(self, rng, replier, parent, rnd): ref = f"p{len(self.objects) + 1}" platform = replier.split("/")[0] if parent["visibility"] == "community" and platform == "lemmy": kind = "comment" elif platform == "lemmy": return None else: kind = "reply" visibility = parent["visibility"] if parent["visibility"] in ("direct", "followers", "circle", "community") else \ rng.weighted({"public": 70, "unlisted": 30}) if visibility == "circle" and platform != "privapub": visibility = "direct" # a remote member answers a circle post as the post's own audience mentions = set() if visibility == "direct": mentions = (parent["mentions"] | {parent["author"]}) - {replier} elif visibility in ("public", "unlisted", "followers") and parent["author"].split("/")[0] in MICRO \ and parent["author"] != replier: mentions = {parent["author"]} # clients mention the account replied to, and servers deliver by mention lang = self.accounts[replier]["lang"] args = {"visibility": visibility, "text": self.text(rng, lang, rng.between(3, 25), ref), "reply_to": parent["ref"], "language": lang} if mentions: args["mentions"] = sorted(mentions) if visibility == "circle": args["group"] = parent["group"] if visibility == "community": grp = self.groups[parent["group"]] if platform == "privapub" and grp["host"] == "privapub": args["group"] = parent["group"] elif platform != "lemmy": args["visibility"] = "public" # a microblog reply to a community thread is a public reply the group relays obj = {"ref": ref, "author": replier, "visibility": visibility, "mentions": mentions, "kind": kind, "round": rnd, "parent": parent["ref"], "followers": set(self.followers(replier)), "group": parent.get("group")} self.objects[ref] = obj self.step("content", replier, "post", args, ref, rnd) return obj # -- interactions def plan_interactions(self): rng = self.rng.sub("interact") conf = self.spec.get("interactions", {}) for ref, obj in sorted(self.objects.items(), key=lambda kv: int(kv[0][1:])): if obj["visibility"] == "located": continue viewers = [k for k in self.accounts if k != obj["author"] and self.can_see(k, obj) and self._can_interact(k, obj)] if not viewers: continue for k in rng.sample(viewers, rng.span(conf.get("likes", [0, 4]))): platform = k.split("/")[0] if platform == "lemmy": self.step("interact", k, "like" if rng.chance(0.75) else "downvote", {"target": ref}) elif "react" in CAPS[platform] and rng.chance(conf.get("react", 0.3)): self.step("interact", k, "react", {"target": ref, "emoji": rng.choice(REACTIONS)}) else: self.step("interact", k, "like", {"target": ref}) if obj["visibility"] in ("public", "unlisted"): for k in rng.sample([v for v in viewers if "boost" in CAPS[v.split("/")[0]]], rng.span(conf.get("boosts", [0, 2]))): self.step("interact", k, "boost", {"target": ref}) if obj.get("options"): for k in rng.sample([v for v in viewers if "vote" in CAPS[v.split("/")[0]]], round(len(viewers) * conf.get("vote", 0.6))): self.step("interact", k, "vote", {"target": ref, "choices": [rng.below(obj["options"])]}) if rng.chance(conf.get("bookmark", 0.1)): k = rng.choice([v for v in viewers if "bookmark" in CAPS[v.split("/")[0]]] or [None]) if k: self.step("interact", k, "bookmark", {"target": ref}) def _can_interact(self, key, obj): platform = key.split("/")[0] if obj["visibility"] == "community": grp = self.groups[obj["group"]] return key in grp["members"] or key == grp["owner"] if platform == "lemmy": return False if obj["visibility"] == "circle": return key in self.groups[obj["group"]]["members"] return True # -- mutations def plan_mutations(self): rng = self.rng.sub("mutate") m = self.spec.get("mutations", {}) for ref, obj in sorted(self.objects.items(), key=lambda kv: int(kv[0][1:])): platform = obj["author"].split("/")[0] if "edit" in CAPS[platform] and rng.chance(m.get("edit", 0.08)) and obj["visibility"] not in ("circle", "community", "located"): lang = self.accounts[obj["author"]]["lang"] # an edit keeps the names the post had, as a client's edit box does; dropping them changes who it is for self.step("mutate", obj["author"], "edit", {"target": ref, "text": self.text(rng, lang, 6, ref) + " (edited)", "mentions": sorted(obj["mentions"]), "cw": obj.get("cw")}) obj["edited"] = True for ref, obj in sorted(self.objects.items(), key=lambda kv: int(kv[0][1:])): if obj["round"] > 0 and rng.chance(m.get("delete", 0.04)) and obj["visibility"] not in ("circle", "located"): self.step("mutate", obj["author"], "delete", {"target": ref}) obj["deleted"] = True micro = self.keys(micro=True) for verb, count in (("block", m.get("blocks", 2)), ("mute", m.get("mutes", 2))): for _ in range(count): a, b = rng.sample(micro, 2) self.step("mutate", a, verb, {"target": b}) for _ in range(m.get("reports", 1)): reporter = rng.choice(micro) theirs = [o for o in self.objects.values() if o["author"] != reporter and o["visibility"] == "public" and o["author"].split("/")[0] in MICRO] if theirs: o = rng.choice(theirs) self.step("mutate", reporter, "report", {"target": o["author"], "refs": [o["ref"]], "comment": "town report"}) def plan(self): self.plan_accounts() self.plan_groups() self.plan_graph() self.plan_content() self.plan_interactions() self.plan_mutations() return self.steps def plan(spec): return Planner(spec).plan() def snapshot(planner): """The planner's world as JSON (sets as sorted lists), stored with a run so later generator changes never re-read it.""" def plain(v): if isinstance(v, set): return sorted(v) if isinstance(v, dict): return {k: plain(x) for k, x in v.items()} if isinstance(v, list): return [plain(x) for x in v] return v return {"version": GENERATOR_VERSION, "steps": planner.steps, "accounts": planner.accounts, "objects": plain(planner.objects), "groups": plain(planner.groups), "follows": plain(planner.follows)} class Stored(Planner): """A Planner rebuilt from a run's plan.json instead of from the spec.""" def __init__(self, snap): self.spec = None self.steps = snap["steps"] self.accounts = snap["accounts"] self.objects = {r: {k: (set(v) if k in ("mentions", "followers") else v) for k, v in o.items()} for r, o in snap["objects"].items()} self.groups = {r: {k: (set(v) if k == "members" else v) for k, v in g.items()} for r, g in snap["groups"].items()} self.follows = {k: set(v) for k, v in snap["follows"].items()} def plan(self): return self.steps def digest(steps): return hashlib.sha256(json.dumps(steps, sort_keys=True, ensure_ascii=False).encode()).hexdigest() def main(args): spec = load(args[0] if args else "village") steps = plan(spec) if "--summary" in args: from collections import Counter print(json.dumps({"steps": len(steps), "sha": digest(steps), "by": dict(Counter(f"{s['phase']}:{s['verb']}" for s in steps))}, indent=1)) return 0 for s in steps: print(json.dumps(s, ensure_ascii=False)) return 0 if __name__ == "__main__": sys.exit(main(sys.argv[1:]))