-- sql/128-url-normalize-and-enrichment-worklist.sql -- Foundation for the datasheet-enrichment pipeline (Lever 3): -- (1) Fix malformed double-prefixed URLs (a scraper concatenated the base host in -- front of an already-absolute URL, e.g. "https://www.mouser.dehttps://www.mouser.de/…"). -- 503 rows affected (details_source_url 373, product_page_url 83, image_url 47). -- These cannot be fetched, so they block any enrichment run. -- (2) A guard so the ingest bug cannot re-introduce them. -- (3) v_datasheet_enrichment_worklist — the pipeline's input: every transceiver/switch -- that needs a field filled from source, with its best source URL + type. -- Idempotent; safe to re-run. The regex only matches a second scheme that appears -- BEFORE the first path slash, so legitimate URLs with an encoded URL in the query string -- (…?url=https://…) are never touched. create or replace function fn_strip_double_url_prefix(u text) returns text language sql immutable as $$ select case when u ~ '^https?://[^/]*https?://' then regexp_replace(u, '^https?://[^/]*(https?://)', '\1') else u end $$; comment on function fn_strip_double_url_prefix(text) is 'Removes a duplicated leading scheme+host from a URL, e.g. https://a.comhttps://a.com/x -> https://a.com/x. Only when the 2nd scheme precedes the first path slash (so ?url=https://… is safe).'; -- (1) one-time cleanup update transceivers set details_source_url = fn_strip_double_url_prefix(details_source_url), product_page_url = fn_strip_double_url_prefix(product_page_url), image_url = fn_strip_double_url_prefix(image_url) where details_source_url ~ '^https?://[^/]*https?://' or product_page_url ~ '^https?://[^/]*https?://' or image_url ~ '^https?://[^/]*https?://'; -- (2) ingest guard create or replace function transceiver_url_normalize() returns trigger language plpgsql as $$ begin NEW.details_source_url := fn_strip_double_url_prefix(NEW.details_source_url); NEW.product_page_url := fn_strip_double_url_prefix(NEW.product_page_url); NEW.image_url := fn_strip_double_url_prefix(NEW.image_url); NEW.datasheet_url := fn_strip_double_url_prefix(NEW.datasheet_url); return NEW; end $$; drop trigger if exists trg_transceiver_url_normalize on transceivers; create trigger trg_transceiver_url_normalize before insert or update on transceivers for each row execute function transceiver_url_normalize(); -- (3) enrichment worklist — pipeline input create or replace view v_datasheet_enrichment_worklist as select 'switch'::text as entity, s.id as entity_id, v.name as vendor, s.model as item, 'ports_config+max_speed'::text as need, coalesce(s.datasheet_url, s.product_page_url) as source_url, case when s.datasheet_url is not null then 'datasheet' when s.product_page_url is not null then 'product_page' else 'none' end as source_type from switches s join vendors v on v.id = s.vendor_id where (s.ports_config is null or s.ports_config = '{}'::jsonb) and s.max_speed_gbps is null union all select 'transceiver', t.id, v.name, t.part_number, 'form_factor+speed', coalesce(t.details_source_url, t.datasheet_url, t.product_page_url), case when t.datasheet_url is not null then 'datasheet' when t.details_source_url is not null then 'details_page' when t.product_page_url is not null then 'product_page' else 'none' end from transceivers t join vendors v on v.id = t.vendor_id where t.speed_gbps is null or t.speed_gbps <= 0; comment on view v_datasheet_enrichment_worklist is 'Input for the datasheet-enrichment pipeline: entities whose ports_config/max_speed (switch) or form_factor/speed (transceiver) must be filled from source, with the best available source URL and its type. source_type=none means URL discovery is needed first.'; insert into _migrations (filename, applied_at) select '128-url-normalize-and-enrichment-worklist.sql', now() where not exists (select 1 from _migrations where filename = '128-url-normalize-and-enrichment-worklist.sql');