diff --git a/migrations/000.396.sql b/migrations/000.396.sql new file mode 100644 index 0000000..7c29f96 --- /dev/null +++ b/migrations/000.396.sql @@ -0,0 +1,73 @@ +-- Add date column to v_analytics_product_categories + +-- We want to add a last modified date to v_analytics_product_categories, for Funnel. + +-- A row can be modified by any of the following reasons: +-- 1: when a new product_category mapping is added +-- -> solved by adding a product_category.inserted column +-- 2: when category display name is updated (en_GB only) +-- -> taken from category_texts.name. solved by adding a category_texts.updated column +-- 3: when category path is updated (en_GB only) +-- -> also from category_texts.name +-- 4: when any of the category's parent categories path is updated +-- -> we skip this for now. not worth the effort as it seems it is not even possible to edit category names in admin at the moment +-- -> I.e. for the full category path cities-places/geographical-locations/europe/scandinavia/finland/helsinki, +-- only changes to helsinki will updated the last_modified column. +-- -> solvable, but complicated, see end of this file for example +-- +-- The last modified field should be the latest of all date values. +-- +-- I don't include changes to categories table, I think only the category_texts table is relevant here. +-- +-- Finally I foresee a problem of how removing a product category mapping will reach Funnel. +-- If we delete a product_category row - how will they know? (since they use date column to partially read the data) + +-- Add category_texts.updated +ALTER TABLE category_texts ADD COLUMN updated timestamp with time zone NOT NULL DEFAULT now(); +CREATE TRIGGER updated_timestamp BEFORE UPDATE ON public.category_texts FOR EACH ROW EXECUTE PROCEDURE public.updated_timestamp(); + +-- Add product_category.inserted +ALTER TABLE product_category ADD COLUMN inserted timestamp with time zone NOT NULL DEFAULT now(); + +-- Add v_analytics_product_categories.last_updated +drop view if exists v_analytics_product_categories; +create view v_analytics_product_categories as +SELECT DISTINCT tree.id AS category_id, + getsafexmlpath(tree.path::character varying, 2) AS category_path, + category_texts.name AS category_name, + product_category.product_id, + greatest(product_category.inserted, category_texts.updated) last_updated +FROM product_category + JOIN categories ON product_category.category_id = categories.id + JOIN categories hidden_category ON hidden_category.name::text = 'hidden'::text + JOIN v_esales_categorytree_i18n tree ON (tree.lft < categories.lft AND tree.rgt > categories.rgt OR categories.id = tree.id) AND tree.id <> 1 AND (tree.lft < hidden_category.lft OR tree.lft > hidden_category.rgt) + JOIN category_texts ON tree.id = category_texts.category_id AND tree.locale_id = category_texts.locale_id +WHERE tree.locale_id = 3; + +grant select on v_analytics_product_categories to datastudio; +grant select on v_analytics_product_categories to funnel; + + +-- Finally, since we are adding this date field for Funnel's sake, it is of no help if all 320k rows have the exact same date. +-- Therefore we modify the dates in batches of 10K (Funnel's recommended limit). +-- We have 320503 rows in v_analytics_product_categories. The dates come from 137739 rows in product_category, and 1665 rows in category_texts +update product_category set inserted = inserted - (product_id::text || category_id::text)::int % 100 * interval '2 day'; +alter table category_texts disable trigger updated_timestamp; +update category_texts set updated = updated - (id % 10 + 1) * interval '1 month' where locale_id = 3; +alter table category_texts enable trigger updated_timestamp; +-- should create batches of roughly < 10k rows: select last_updated, count(*) from v_analytics_product_categories group by 1; + +-- For reference, if we need to include the modification dates for parent categories: +-- create view v_category_relatives as +-- select +-- categories.id category_id, +-- tree.id related_category_id, +-- category_texts.name category_name, +-- category_texts.updated name_update, +-- getsafexmlpath(tree.path, 2) category_path +-- from categories +-- join v_esales_categorytree_i18n tree on tree.locale_id = 3 and (tree.lft < categories.lft and tree.rgt > categories.rgt or categories.id = tree.id) and tree.id <> 1 +-- join category_texts on tree.id = category_texts.category_id +-- where category_texts.locale_id = 3; +-- this will get the latest modification date of all relatives of a category: +-- select max(name_update) from v_category_relatives where category_id = 1094;