mirror of
https://github.com/discourse/discourse.git
synced 2026-08-11 02:59:07 +08:00
4032dc2b05
The post_* linkage tables from #41177 become embed_* tables keyed by (owner_type, owner_id), because placeholders will also live in user bios and group/category/badge descriptions later. On top of the rename: - quotes carry source coordinates (topic id + post number) and links a typed target, so the importer can resolve both once the id maps exist - new hashtag and emoji linkage tables; mention_type, hashtag_type and link_target become integer enums and the string MentionType is gone - EmbedBuffer records for one owner kind, PlaceholderResolver resolves and renders per owner and reports unresolved embeds and orphaned placeholders - the posts and custom_emojis tables stay ignored until their conversion steps land in a follow-up PR - the lazy source-name lookups (users, groups, categories with their slug paths, synonym-folded tags) live in their own NameResolver, and the resolver spec is split by phase under placeholder_resolver/
20 lines
860 B
Ruby
Vendored
20 lines
860 B
Ruby
Vendored
# frozen_string_literal: true
|
|
|
|
# Custom emoji embeds (`:name:`). Only the source's custom emoji become embeds;
|
|
# standard `:smile:` shortcodes stay plain text. A custom emoji is keyed by name,
|
|
# and the name is exactly what a conflict renames at import, so there's nothing to
|
|
# resolve ahead of time — `name` is the source shortcode (without the surrounding
|
|
# colons) and the importer rewrites it through the emoji-name map.
|
|
#
|
|
# `placeholder` holds the token spliced into the owner's markdown; `owner_type`/
|
|
# `owner_id` name that owning record.
|
|
Migrations::Tooling::Schema.table :embed_emojis do
|
|
synthetic!
|
|
|
|
add_column :owner_type, :integer, required: true, enum: :embed_owner
|
|
add_column :owner_id, :numeric, required: true
|
|
add_column :placeholder, :text, required: true
|
|
add_column :name, :text, required: true
|
|
|
|
index :owner_type, :owner_id
|
|
end
|
