Sitemap Generation

This document describes the automated XML sitemap generation system.

Overview

The system generates XML sitemaps for SEO, covering all public-facing entity types. Sitemaps follow the sitemaps.org protocol and are split into multiple files when exceeding the per-file item limit.

Console Command

File: console/controllers/SitemapController.php

php yii sitemap/index [--webStoreNavCode=WEB|WEB2]

• Uses AlreadyRunningFilter to prevent concurrent execution
• Supports multiple web stores via --webStoreNavCode option
• Items per file: 8,000 (ITEMS_COUNT_PER_FILE)

Entity Types

The sitemap covers these entity types, each with a dedicated data provider and last-modification extractor:

Entity	Data Provider	Last Modified Extractor
Products	SitemapProductDataProvider	ProductLastModificationExtractor
Categories	SitemapCategoryDataProvider	CategoryLastModificationExtractor
Content Pages	SitemapContentPageDataProvider	ContentPageLastModificationExtractor
Authors	SitemapAuthorDataProvider	ProductPersonLastModificationExtractor
Other person types (illustrators, translators, actors, compilers, performers, readers)	SitemapProductPersonDataProvider (one pass per type)	ProductPersonLastModificationExtractor
Publishers	SitemapPublisherDataProvider	ProductPublisherLastModificationExtractor
Series	SitemapSeriesDataProvider	ProductSeriesLastModificationExtractor
Product lists	SitemapProductListDataProvider (activeAndAvailable scope)	ProductList.updated
Campaign lists	SitemapCampaignListDataProvider (active + web-visible + public + url)	CampaignList.updated
Blog posts	SitemapBlogDataProvider (origin + active + store)	Blog.updated
Events	SitemapEventDataProvider (origin + active + visible)	Event.updated

Product lists use the model's own activeAndAvailable() visibility scope (is_active + within the visible_from/visible_to window) so the sitemap matches the storefront. They are emitted in the store's default language only — lists are language-specific (language_type_id) and resolve by their unique slug, so no hreflang alternates are fabricated.

Campaign lists mirror the storefront API's public lookup (CampaignListController::findModel): active() (within the visibility window — excludes expired campaigns) + availableForWeb() (do_not_show_in_web = 0) + the current store + forClientType(null) (no client-type restriction, i.e. what an anonymous crawler sees) + a non-empty url (campaigns are resolved by their stored slug, so urlless ones are unreachable). Same single-file, default-language, updated-based lastmod approach as product lists.

Blog posts are single-language (the language column); each published post (origin = not a draft, active) is emitted once, in its own language, at /{lang}/{blog}/{category}/{slug} (category = the post's mainTag). getBlogUrl() instantiates the v2 BlogUrlBuilder directly — blogs are not registered with the version-aware builder factory.

Events are global (no web_store_id), so they are emitted only into the main store's sitemap to avoid cross-store duplication, and only when currently visible() (within the visible_from/visible_to window — past events are excluded). Each event is multilingual via rr_event_translation: one URL per language that has a translation slug (the builder throws for languages without one, which the section skips). getEventUrl() goes through the wired forEvent() factory method.

Authors keep their own provider (author-specific bulk indexation approver + productPersonTops boost); the other PRODUCT_PERSON_TYPE pages go through the generic SitemapProductPersonDataProvider + getProductPersonUrl(), which routes the URL path by type (/illustrators/…, /translators/…, etc.). A person can appear under more than one type, yielding one URL per qualifying type.

Data providers directory: console/sitemap/

Last Modification Tracking

Each entity type has specialized repositories to determine the most recent modification datetime:

Directory: console/sitemap/fieldExtractor/repositories/

Category

• CategoryPageComponentUpdatedDateTimeRepository
• CategorySeoUpdatedDateTimeRepository
• CategoryStatisticUpdatedDateTimeRepository

Content Page

• ContentPageComponentUpdatedDateTimeRepository

Person (Author)

• ProductPersonMetaUpdatedDateTimeRepository

Publisher

• ProductPublisherMetaUpdatedDateTimeRepository

Series

• ProductSeriesDescriptionUpdatedDateTimeRepository

Output

Sitemaps are generated as XML files using PHP's XMLWriter. Each file contains up to 8,000 URLs with:

• <loc> — Full URL
• <lastmod> — Last modification datetime

Files are written to the web-accessible directory for search engine crawlers.

Monitoring

actionIndex records a sitemap/lastRun entry in rr_system_state on every fully-completed run, holding per-store/per-entity URL counts, the completion timestamp and the total runtime. The write is best-effort — a bookkeeping failure never breaks generation — and is only reached when the run finishes, so a stale finished_at is itself a signal that the run died partway.

The daily health report (common/reports/SystemHealthReport.php, section "Sitemap — generation freshness") surfaces this:

• Per-entity URL counts and last successful regen timestamp from the sitemap/lastRun state record.
• Partial-failure detection by comparing the oldest vs newest sitemap-*.xml mtimes on disk. A healthy run writes all sub-sitemaps within one multi-hour pass, so a spread wider than 24h means the cron died mid-run and the tail (content pages / authors / publishers / series) is frozen.
• Staleness detection when the newest file is older than 192h / 8 days (a weekly run was fully missed / the cron stopped firing; the bound sits above the normal ~168h weekly gap). Thresholds live on SystemHealthReport.

Resilience & failure handling

The product phase performs per-product/per-image S3 existence checks (fileExists → S3::doesObjectExist), so it is network-bound and by far the longest, most failure-prone step. To stop a failure there from freezing the cheap tail entities:

• Each entity section runs through runSection(), which catches any throwable, logs it, records it under errors[<section>] in the sitemap/lastRun state, and lets the run continue. A single failing section no longer aborts the rest.
• The master index (general()) is always rebuilt afterwards from whatever sub-sitemaps exist on disk (prepareDirectory never wipes).
• Recorded section errors are surfaced in the daily health report as both an action line and a per-store ⚠️ <section> failed note.

Note: the generation cron runs weekly (Sunday 07:11) and is scheduled via the Zone control panel (not in this repo). The cron generates on the build box and then scps the output to the FE web servers that serve it (rahvaraamat.ee and raoutlet.ee). The health report runs on the build box and checks the generated files, so it does not detect scp/delivery failures — a follow-up live HTTP freshness check against the served sitemap would close that gap.

Models Used

• Product — common/models/Product.php
• Category / CategoryPage — common/models/Category.php, common/models/CategoryPage.php
• ContentPage — common/models/ContentPage.php
• ProductPerson — common/models/ProductPerson.php
• ProductPublisher — common/models/ProductPublisher.php
• ProductSeries — common/models/ProductSeries.php
• WebStore — common/models/WebStore.php