// NO PROD JSON API (bucket C) — mock until backend (see PROD_API_INVENTORY.md) // Represents: admin.blog / admin.blog.category · BlogController + BlogCategoryController — see docs/ISYSTEM_REFERENCE.md §Batch 4 "Blogs + Blog categories" /* CMS ▾ → Blogs and CMS ▾ → Blog categories. Two screens, one file, because they are one feature: `blogs.category_id` → `blog_categories.id`, both indexes cross-link ("Manage categories" / "Back to blogs"), and a category cannot be deleted while a blog points at it. REAL ROUTES / CONTROLLERS - Blogs `admin.blog` GET /blog/ (routes/admin.php:L1015-1017) · BlogController::index L21, ::blogForm L54 (GET /blog/form/), ::saveBlog L66 (POST /blog/save), ::delete L88 (GET /blog/delete/{id}/, plus a duplicate un-named legacy copy at L1628-1630). - Categories `admin.blog.category` GET /blog/category/ (L979-981) · BlogCategoryController:: index L23, ::categoryForm L54, ::saveCategory L66, ::delete L87 (duplicate legacy delete route L1631-1633). - Logic BlogAdminService (listForTable L30, summarizeStatus L49, prepareFormData L69, saveBlog L193, listCategoriesForTable L335, summarizeCategoryStatus L355, prepareCategoryFormData L376, saveCategory L411). - Requests SaveBlogRequest, SaveBlogCategoryRequest. - Views admin/blog/index.blade.php · form-page → form.blade.php → partials/ translation-language-section.blade.php (one stacked section PER LANGUAGE) + components/admin/blog/image-upload-field.blade.php; admin/blog/category/ index.blade.php · category/form.blade.php (inline select2 + tab JS, no TinyMCE). Both modals/*.blade.php files are referenced by nothing — dead code. THE 12-LANGUAGE PROBLEM (the one real IA decision on this screen) The real blog form renders the per-language block as a STACKED section per language — twelve copies of {title, body, image, thumbnail} down one page, 48 fields, with no way to see which locales are actually filled. The category form already uses tab JS for the same list. This rebuild puts BOTH behind one language rail (HbgLangTabs) with a per-language translated / partial / untranslated indicator and a coverage counter, plus an "All languages" overview: read-only coverage matrix for blogs (4 fields × 12), the twelve editable title inputs for categories (1 field × 12). No field is added, removed or renamed — this is navigation over exactly the fields SaveBlogRequest / SaveBlogCategoryRequest validate. The tab list is NOT hardcoded on the real platform: blogs read BlogAdminService::adminLanguages L633 and categories read LanguagesController::getLanguages() (Language::all, cached 24h, English sorted first), so the real count is whatever production's `languages` table holds. HBG_LANGS below carries the 12-language content set the reference documents for this platform's multilingual content (ISYSTEM_REFERENCE.md §Batch 6 "Multi-language CONTENT") in the order the legacy prototype bundle observed on the real screen. See UNCLEAR #1 in the phase report: the sibling Languages screen's mock `languages` table (src/pages/HostSetLanguages.jsx) holds a different 13 rows (pl/nl/br present, ar/ro absent) — the two lists cannot both be production. PER-SKIN SLUG UNIQUENESS (validated here exactly as documented, and it differs per screen) - blogs: UNIQUE (skin_id, slug). There is NO slug field on the form. Blog::generateSlug L85 derives it from master_title (fallback title_{default locale}) and appends -1, -2 … until it is free WITHIN THAT SKIN, and never regenerates it on edit. Since create fans out one row per selected skin, one title can produce a different slug per skin — so the form shows the resolved slug per selected skin instead of pretending there is one. - blog_categories: UNIQUE (skin_id, slug) too, but slug IS an input. Empty → auto-generated from the first non-empty language title; a collision is a hard error (backend.slug_already_exists = "Slug already exists", surfaced behind backend.operation_error). Checked per selected skin on create, against the target skin on edit (the skin can be changed there). FAITHFUL ABSENCES — deliberately NOT built - No sortable columns anywhere: both lists are fixed `id DESC` (BlogAdminService L39 / L345). - No export (neither screen has one), no bulk actions, no KPI strip beyond the documented "Results" total and the status chip counts. - No page-size select: `$perPage = 25` is hardcoded, so the pager is Prev/Next only. - No slug input on the blog form, no "regenerate slug" action, no restore/duplicate/preview action — none exist upstream. - Blogs' Category column shows `blog_categories.slug` (leftJoin), while the Category FILTER and the form's picker label the same rows by `getTitle() ?? slug`. That inconsistency is real and is reproduced rather than smoothed over. KNOWN-BUG POLICY: the reference records no behavioural bug for these two screens (the batch-wide bug list covers Promotions, Game import and Slideshow), so nothing here diverges from the real platform. What it does record are permission and data-integrity asymmetries, surfaced honestly in the header Tip / Explainer instead of being papered over: - Customer Care with `support_cms_banners` SEES both entries in the sidebar (sidebar.blade.php L18) but 404s at the controller — a menu/controller mismatch. - Blog delete re-checks skin ownership for skin admins; category delete does NOT (only isadmin/isSkinAdmin, BlogCategoryController L87-91). - Changing a category's skin on edit does not move the blogs pointing at it, so a blog can end up in a category owned by another skin. HBG_BLOG_SEED contains one such row on purpose. LABEL POLICY: keys that resolve in the committed public/default-lang/en/backend.php are used verbatim ("Blogs", "New Blog", "Master Fields", "Master Title", "Master Body", "Master Image", "Fallback if translation empty", "Translations", "Promotion Settings", "Is Promotion", "Promotion Text 1/2", "Valid Until", "Published", "Draft", "Promotion", "Select Multiple", "No categories available", "Please create a category first", "If empty, will use master …", "Blog Categories", "New Category", "Data", "Slug", "Active", "Disabled", "Cannot delete category that has blogs", "Slug already exists", "Slug will be auto-generated if empty", "Select at least one skin"), as are the strings hardcoded in English in the blades ("Master Thumbnail Image", "Button Text", "Button Link", "Casinos (Select Multiple)", "Casino", "Select one or more casinos…", "Select the casino this category belongs to.", "Category Title"). Everything else on these pages resolves only in the gitignored storage/lang — page subtitles, "Back to list", "Back to blogs", "Back to categories", "Edit blog", search placeholders, "All categories", "Apply", "Results", "Showing", "Prev", "Next", "Created", "No records", "Delete", "Remove current image", "Slug or title is required" — and is written as a sensible operator-facing label, marked "label inferred" at the point of use. */ const { useState: hbgUseState, useMemo: hbgUseMemo } = React; /* NOT-A-GENERATOR: FNV-1a, and all that survives of the PRNG pair that used to build this screen's contents. Its one remaining job is picking a stable colour for an image placeholder from a stored path, so the same file always gets the same tile. The mulberry32 it used to seed — which decided how many of a post's twelve languages existed, and whether each had a body — is gone with the seed data it fed. tools/wiringstate.js flags any Math.imul with no *Rng name, since that is how a generator hides from its classifier. This marker is the declared way out and the tool only honours it on a file that also reads the database. */ const hbgHash = (s) => { let h = 2166136261 >>> 0; for (let i = 0; i < s.length; i++) { h ^= s.charCodeAt(i); h = Math.imul(h, 16777619); } return h >>> 0; }; const hbgToast = (title, detail) => window.hrsToast ? window.hrsToast(title, detail) : (window.PAYBO && window.PAYBO.emitToast && window.PAYBO.emitToast({ id: `hbg-${Date.now()}`, tx_id: title, amount: 0, currency: "HOST", player: "Blogs", reason: detail || "Prototype state only \u2014 not persisted." })); /* Cross-page navigation — push the target route's canonical path and let app.jsx's popstate handler resolve it (the hdNavTo convention from HostDashboard.jsx). Used by the two cross-links the real indexes carry: Blogs → "Manage categories", Categories → "Back to blogs". */ const hbgNavTo = (routeId) => { try { const path = window.pathForActive && window.pathForActive(routeId); if (path) { if (window.location.pathname !== path) window.history.pushState({ active: routeId }, "", path); window.dispatchEvent(new PopStateEvent("popstate")); } } catch (_e) { /* no-op */ } }; /* blogs.addedTime / blog_categories.addedTime are unix ints (models override getDateFormat()='U'). Index display convention is date('j M Y, H:i'), "—" when empty. */ const HBG_MON = ["Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"]; const hbgUnix = (y, m, d, h, mi) => Math.floor(new Date(y, m - 1, d, h, mi).getTime() / 1000); const hbgPad2 = (n) => String(n).padStart(2, "0"); const hbgFmtTs = (ts) => { if (!ts) return "—"; const d = new Date(ts * 1000); return `${d.getDate()} ${HBG_MON[d.getMonth()]} ${d.getFullYear()}, ${hbgPad2(d.getHours())}:${hbgPad2(d.getMinutes())}`; }; /* Str::slug stand-in — strips accents, emoji and punctuation, collapses to dashes. */ const hbgSlugify = (s) => String(s || "") .toLowerCase() .normalize("NFD").replace(/[\u0300-\u036f]/g, "") .replace(/[^a-z0-9]+/g, "-") .replace(/^-+|-+$/g, ""); /* Blog::generateSlug L85 — base slug, then -1, -2 … until free WITHIN THE SKIN. */ const hbgNextSlug = (base, takenInSkin) => { if (!base) return ""; if (takenInSkin.indexOf(base) === -1) return base; let i = 1; while (takenInSkin.indexOf(`${base}-${i}`) !== -1) i++; return `${base}-${i}`; }; /* ------------------------------------------------------------------ * * `languages` rows (see the 12-language note in the header comment). * English is force-sorted first by both services; the remaining order is * the one the legacy prototype bundle observed on the real screen. * ------------------------------------------------------------------ */ const HBG_LANGS = [ { code: "en", name: "English" }, { code: "it", name: "Italiano" }, { code: "de", name: "Deutsche" }, { code: "tr", name: "Türkçe" }, { code: "ar", name: "Arabic" }, { code: "ro", name: "Romana" }, { code: "zh", name: "Chinese" }, { code: "es", name: "Español" }, { code: "fr", name: "Français" }, { code: "pt", name: "Português" }, { code: "pt_br", name: "Português-Brasil" }, { code: "hu", name: "Magyar" }, ]; /* config/app.php locale AND fallback_locale are both 'en', so "title_{fallback_locale}" in SaveBlogRequest::withValidator L85-94 is title_en, and BlogCategory::getTitle()'s chain is locale → en → slug. */ const HBG_FALLBACK = "en"; const HBG_LANG_LABEL = (code) => { const l = HBG_LANGS.find(x => x.code === code); return l ? `${l.name} (${code.toUpperCase().replace("_", "-")})` : code; }; /* Auth::user()->getSkins() / SkinsController::getSkinsList(). Read from `skins` — every brand this operator's RLS lets them see. It was six hardcoded ids and names copied from the Business report page, which is exactly the arrangement that makes an invented list look verified: two screens agreeing with each other about brands neither of them had read. */ const hbgSkinName = (skins, id) => { const s = (skins || []).find(x => String(x.id) === String(id)); return s ? s.name : null; }; const hbgInitials = (name) => String(name || "?").replace(/[^A-Za-z0-9 ]/g, "").split(" ").filter(Boolean).slice(0, 2).map(w => w[0].toUpperCase()).join("") || "?"; /* ------------------------------------------------------------------ * * THE DATA. This screen used to carry its own database too. * * `HBG_DB = { blogs: hbgSeedSlugs(hbgBuildBlogs()), cats: hbgBuildCats() }` * built thirty-four posts across six brands, minted their slugs through the * same -1/-2 de-dup the real generateSlug uses so two "Bonus"-titled posts on * one brand collided exactly as they would in production, and generated each * post's twelve-language translation coverage from a PRNG seeded on its id — * English at 95%, the rest thinning out, "the realistic shape of a 12-locale * post". Both screens mirrored it into local state and wrote back on mutation, * so an edit made on one was still there after navigating to the other. * * That care is why it was convincing. It is also why nothing about it was true. * * Replaced by reads of `blogs` and `blog_categories` with their translations * embedded. The conversion below is the only code that knows the storage * differs: isystem keeps a post's twelve titles and bodies as flat columns, * this schema keeps one blog_translations row per language, so saving a post is * N+1 writes rather than one. * ------------------------------------------------------------------ */ /* BlogCategory::getTitle() — requested locale → fallback locale (en) → slug. */ const hbgCatTitle = (cat, locale) => { if (!cat) return "—"; const t = cat.t || {}; return t[locale || HBG_FALLBACK] || t[HBG_FALLBACK] || cat.slug; }; /* Fetch ceiling, and NOT silent: a brand with more posts than this pages in the database while this screen searches only what it holds, so the list says so. */ const HBG_FETCH_MAX = 500; /* blog_translations rows -> the per-language map both forms already speak. */ const hbgTrFromRows = (rows) => { const t = {}; (rows || []).forEach(r => { const c = String(r.locale_code || "").toLowerCase(); if (!HBG_LANGS.some(l => l.code === c)) return; t[c] = { title: String(r.title || ""), body: String(r.body || ""), image: String(r.image_url || ""), thumb: String(r.thumbnail_url || ""), }; }); return t; }; const hbgCatTrFromRows = (rows) => { const t = {}; (rows || []).forEach(r => { const c = String(r.locale_code || "").toLowerCase(); if (HBG_LANGS.some(l => l.code === c)) t[c] = String(r.title || ""); }); return t; }; /* `published`/`active` are real booleans here and 0/1 upstream; timestamps are timestamptz here and unix ints upstream (the models override getDateFormat() to 'U'). Both are converted at this edge so every renderer and both forms keep working against the shape they were written for. */ const hbgUnixOf = (iso) => (iso ? Math.floor(new Date(iso).getTime() / 1000) : 0); const hbgBlogFromDb = (r) => ({ id: r.id, skin_id: r.skin_id, category_id: r.category_id, slug: r.slug || "", master_title: r.master_title || "", master_body: r.master_body || "", master_image: r.master_image_url || "", master_thumb: r.master_thumbnail_url || "", is_published: r.published ? 1 : 0, published_at: hbgUnixOf(r.published_at) || null, is_promotion: r.is_promotion ? 1 : 0, promotion_text_1: r.promotion_text_1 || "", promotion_text_2: r.promotion_text_2 || "", promotion_button_text: r.promotion_button_text || "", promotion_button_link: r.promotion_button_link || "", /* The form's datetime-local input wants YYYY-MM-DDTHH:mm, which is the first 16 characters of the ISO string. Anything else and the field renders blank while still holding a value — an edit that silently clears a date. */ valid_until: r.valid_until ? String(r.valid_until).slice(0, 16) : "", addedTime: hbgUnixOf(r.created_at), tr: hbgTrFromRows(r.translations), _tr: r.translations || [], }); const hbgCatFromDb = (r) => ({ id: r.id, skin_id: r.skin_id, slug: r.slug || "", is_active: r.active ? 1 : 0, t: hbgCatTrFromRows(r.translations), _tr: r.translations || [], addedTime: hbgUnixOf(r.created_at), /* blogs(count) — every post filed under this category, including ones this screen has not fetched. Counting fetched rows would under-report and make a category look deletable when posts still point at it. */ _blogCount: Array.isArray(r.blogs) && r.blogs[0] ? Number(r.blogs[0].count) || 0 : 0, }); /* Both lists plus the brand list. Shared by the two screens, so a category created on one is visible on the other after its refetch — which is what the module-level store was for, done by asking the database instead. */ const hbgUseDb = () => { const skinsFeed = useHrsFetch(() => window.sb.list("skins", { limit: 200 }), []); const blogsFeed = useHrsFetch(() => window.sb.list("blogs", { limit: HBG_FETCH_MAX }), []); const catsFeed = useHrsFetch(() => window.sb.list("blogCategories", { limit: HBG_FETCH_MAX }), []); const blogs = hbgUseMemo(() => (blogsFeed.data || []).map(hbgBlogFromDb), [blogsFeed.data]); const cats = hbgUseMemo(() => (catsFeed.data || []).map(hbgCatFromDb), [catsFeed.data]); const skins = hbgUseMemo( () => (skinsFeed.data || []).map(k => ({ id: k.id, name: k.name, code: k.code })), [skinsFeed.data]); const truncated = [ blogsFeed.meta && blogsFeed.meta.total > blogs.length ? "posts" : null, catsFeed.meta && catsFeed.meta.total > cats.length ? "categories" : null, ].filter(Boolean); return { blogs, cats, skins, truncated, loading: skinsFeed.loading || blogsFeed.loading || catsFeed.loading, error: skinsFeed.error || blogsFeed.error || catsFeed.error, retry: () => { skinsFeed.retry(); blogsFeed.retry(); catsFeed.retry(); }, feeds: [blogsFeed, catsFeed], }; }; /* ------------------------------------------------------------------ * * The write half. Same contract as the FAQ screen's hfqSyncTr, and the * same reason it cannot simply be shared: the child tables differ * (title+body+two images here, title+content there) and a helper * parameterised over both would take more arguments than it saves. * * A post's parent row goes first — a translation needs a blog_id — and a * failure afterwards leaves the post EXISTING with some of its languages. * Reported as PARTIAL rather than as saved. * ------------------------------------------------------------------ */ const hbgTrHasContent = (v) => !!(v && (String(v.title || "").trim() || String(v.body || "").trim() || String(v.image || "").trim() || String(v.thumb || "").trim())); const hbgSyncTr = async ({ resource, fkColumn, parentId, existing, next, isCat }) => { const have = new Map((existing || []).map(r => [String(r.locale_code || "").toLowerCase(), r])); const failed = []; let wrote = 0, removed = 0; for (const l of HBG_LANGS) { const code = l.code; const prev = have.get(code); const src = (next || {})[code]; const keep = isCat ? !!String(src || "").trim() : hbgTrHasContent(src); /* 006 refuses a blog_translations row that is entirely empty, and blog_category_translations.title is NOT NULL — so "no content" is not a row with blank fields, it is no row. */ const body = isCat ? { title: String(src || "").trim() } : { title: String((src || {}).title || "") || null, body: String((src || {}).body || "") || null, image_url: String((src || {}).image || "") || null, thumbnail_url: String((src || {}).thumb || "") || null }; if (keep) { const res = prev && prev.id ? await window.sb.update(resource, prev.id, body) : await window.sb.create(resource, Object.assign({ [fkColumn]: parentId, locale_code: code }, body)); if (res && res.ok) wrote++; else failed.push(`${l.name}: ${(res && res.error && res.error.message) || "refused"}`); } else if (prev && prev.id) { /* Emptied in the form. Hard delete — these tables carry no deleted_at, which is 006 saying a translation is not a record of anything that happened. Skipping this is how a language the operator removed stays published to players. */ const res = await window.sb.remove(resource, prev.id); if (res && res.ok) removed++; else failed.push(`${l.name} (removal): ${(res && res.error && res.error.message) || "refused"}`); } } return { wrote, removed, failed }; }; const HBG_PER_PAGE = 25; /* BlogAdminService::listForTable $perPage = 25 (L30) */ const HBG_MAX_IMG_KB = 5120; /* nullable|image|max:5120 */ /* ================================================================== * * Small shared pieces * ================================================================== */ /* Skin cell: initials avatar + name, "—" when the leftJoin found nothing. */ const HbgSkinCell = ({ skinId, skins }) => { const name = hbgSkinName(skins, skinId); if (!name) return —; return ( {hbgInitials(name)} {name} ); }; /* Status chip strip. Counts come from summarizeStatus / summarizeCategoryStatus, which ignore the status filter itself so the numbers stay stable while you switch tabs; picking a chip sets the same is_published / is_active param the select does, preserving every other filter. */ const HbgChips = ({ items, value, onPick }) => (
{items.map(it => ( ))}
); const HbgStatusPill = ({ on, onLabel, offLabel }) => ( {on ? onLabel : offLabel} ); /* Sectioned config panel — the Settings.jsx shape (title + optional Explainer callout + body). */ const HbgPanel = ({ title, sub, icon, aside, children }) => (
{icon && }
{title}
{sub &&
{sub}
}
{aside &&
{aside}
}
{children}
); const HbgField = ({ label, req, hint, error, tip, children, wide }) => (
{label && ( )} {children} {hint && !error &&
{hint}
} {error &&
{error}
}
); /* Validation banner — the real save redirects back withInput and flashes backend.operation_error + the message list. */ const HbgErrors = ({ items }) => { if (!items || !items.length) return null; return (
Operation error{/* backend.operation_error */}
    {items.map((m, i) =>
  • {m}
    • )}
); }; /* TinyMCE 5.0.16 stand-in (loaded from CDN by public/js/pages/blog/form.js). The toolbar is decorative — the prototype stores the raw HTML string the editor would post. */ const HbgRichText = ({ value, onChange, minH = 150, placeholder }) => (