SEO and marketplace feeds
Shop products and categories are not MODX resources, so nothing that walks the resource tree will find them. That is the problem all three sections below exist to solve.
Structured data
The product page prints JSON-LD: a Product with its offers, and a BreadcrumbList. Category pages print a listing. It is on by default and needs no configuration.
Product carries the name, the URL, the image, the description and one of two offer shapes:
- a plain
Offerwith a price, for a simple product; - an
AggregateOfferwithlowPrice,highPriceandofferCount, for a composite one.
Why JSON-LD rather than microdata in the markup
Decided for this storefront specifically, and the main reason is the composite product. Its price is not printed on the server: the price block is left empty and filled in by the shop's script once the buyer picks a size. There is nothing to hang an itemprop="price" on — the node does not exist in the server's response, so microdata would either lie with a zero or stay silent. The variants are known on the server, so JSON-LD prints an honest AggregateOffer where microdata has nothing to print.
Two smaller reasons in the same direction: markup spread across div attributes is lost by the owner at the first redesign, and one block has one escaping rule instead of one per attribute.
A product with no price in the current currency gets no offers block at all. That is deliberate: an Offer without a price is an error to Google and the product is dropped entirely, while price: 0 would promise it for free. The storefront does not sell that product either — its card shows no price.
This is not an override point. A site that wants its own markup prints it from its own product chunk rather than editing ours.
Sitemap
Off by default. Set pbshop_sitemap_route to the address you want — for example sitemap-shop.xml — and the route appears.
This is the fallback path, not the main one
The main path is to add the shop's URLs as a section of the sitemap your site already builds: Seo\CatalogUrls hands them over for exactly that.
The setting is empty by default — the opposite decision from the catalogue prefix, and for a reason. Almost every site already has somebody generating a sitemap, and /sitemap.xml is taken. A shop that grabbed that address on install would take the whole site's sitemap away and hand back a catalogue.
Writing sitemap.xml into the setting is allowed. It is just the site owner's conscious choice, not a package default.
One file while the catalogue fits in it; 5,000 URLs per file, not the 50,000 the standard permits, because a sitemap is read by a robot and nobody ever sees the response that failed. A bigger catalogue answers with a sitemap index rather than a first page: silently truncating would leave a 20,000-product shop with a sitemap that looks complete while the rest simply is not indexed — noticeable only by counting.
There is no changefreq or priority, and no robots.txt filtering. If part of your catalogue is closed to indexing, write your own controller over CatalogUrls and apply the filter there — the shop cannot reach the site-owned layer where that rule lives.
Open Graph is not implemented. Said plainly so it is not looked for: no OG tags are printed anywhere. That belongs to the site's page template.
Marketplace feeds
A YML feed for two Ukrainian marketplaces:
https://your-site/pbshop/feed/prom.xml
https://your-site/pbshop/feed/rozetka.xmlNot Yandex.Market
Both are YML descendants, which makes it easy to mistake them for one format with small differences. They were written from each marketplace's own documentation, not from Yandex's specification, and they are not the same format.
The address sits under pbshop/ rather than the catalogue prefix, on purpose: you paste it into a marketplace dashboard once, and it must not change because the owner renamed the storefront from /shop to /katalog. It also works when the storefront is switched off entirely — a shop may trade only on marketplaces.
Where the two differ
| Prom | Rozetka | |
|---|---|---|
| Variants | group_id glues offers into one product with options | no such field — each variant is a separate offer, glued by the marketplace at moderation |
| Portal category | accepted on the offer | only as an attribute on the category |
| Unknown tags | ignored | a stated cause of import failure |
| Stock | optional | required on every offer |
That last row is why the offer is assembled from an allow-list per marketplace rather than "like Prom, minus the extras". Prom documents that undescribed tags are ignored; Rozetka makes no such promise and names "extra or missing tags" as a reason an import fails.
Configuring it
| Setting | Default | What it does |
|---|---|---|
pbshop_feed_currency | empty | the feed's currency; empty means the shop's |
pbshop_feed_company | empty | the legal entity in the feed header |
pbshop_feed_vendor | empty | the manufacturer for products without one |
pbshop_feed_ttl | 60 | minutes the built file lives before rebuilding |
pbshop_feed_key | empty | non-empty requires ?key= in the address |
A feed has exactly one currency
Prices in pbShop are fixed per currency and never converted at a rate — see Currencies. Rozetka wants a rate for each currency, which is exactly the thing this shop does not have. Rozetka also accepts only UAH, USD and EUR.
feed_company empty means the tag is not printed: an empty <company> is read as an error by the marketplace rather than as "unspecified". feed_vendor exists because Rozetka requires <vendor> on every offer and there is no column for it — a product takes it from a vendor or brand constructor field.
The feed is cached on disk and rebuilt when stale, written atomically so a robot never collects a half-written file. An hour rather than "every time", because the robot comes on its own schedule and rebuilding per request means a full catalogue walk at someone else's command — which is to say, a way to take the shop down by requesting the feed ten times.
feed_key is empty because a feed is public by nature and a marketplace cannot log in. But a whole catalogue in one file is also what competitors collect, so there has to be a way to close the address.
What is not here
- Ecommerce events for GA4 or Metrica. Not printed.
shop.jsis where they would live. - 1C, МойСклад, CRM. Paid modules, and worth being blunt about: there is no groundwork for them in the core yet. No
external_idon a product, no price or stock ownership, noPRODUCT_*events, no exchange journal. If you are buying on the strength of a future sync, buy on what is on this site instead.
Canonical product URLs with a 301 from the wrong address, friendly URLs, and seo_title / seo_description on products and categories all work and are part of the catalogue — see Quick start. Filtered and sorted pages are closed with noindex,follow while pagination stays open.