Merge branch 'feat/allow_alt_text_search_config' into 'develop'

allow user defined text search config in database

See merge request pleroma/pleroma!3275

docs/administration/CLI_tasks/database.md

@@ -141,3 +141,21 @@ but should only be run if necessary. **It is safe to cancel this.**
     ```sh
     mix pleroma.database ensure_expiration
     ```
+
+## Change Text Search Configuration
+
+Change `default_text_search_config` for database and (if necessary) text_search_config used in index, then rebuild index (it may take time). 
+
+=== "OTP"
+
+    ```sh
+    ./bin/pleroma_ctl database set_text_search_config english
+    ```
+
+=== "From Source"
+
+    ```sh
+    mix pleroma.database set_text_search_config english
+    ```
+
+See [PostgreSQL documentation](https://www.postgresql.org/docs/current/textsearch-configuration.html) and `docs/configuration/howto_search_cjk.md` for more detail.

docs/configuration/howto_search_cjk.md

@@ -0,0 +1,42 @@
+# How to enable text search for Chinese, Japanese and Korean
+
+Pleroma's full text search feature is powered by PostgreSQL's native [text search](https://www.postgresql.org/docs/current/textsearch.html), it works well out of box for most of languages, but needs extra configurations for some asian languages like Chinese, Japanese and Korean (CJK).
+
+
+## Setup and test the new search config
+
+In most cases, you would need an extension installed to support parsing CJK text. Here are a few extension you may choose from, or you are more than welcome to share additional ones you found working for you with the rest of Pleroma community.
+
+ * [a generic n-gram parser](https://github.com/huangjimmy/pg_cjk_parser) supports Simplifed/Traditional Chinese, Japanese, and Korean
+ * [a Korean parser](https://github.com/i0seph/textsearch_ko) based on mecab
+ * [a Japanese parser](https://www.amris.co.jp/tsja/index.html) based on mecab
+ * [zhparser](https://github.com/amutu/zhparser/) is a PostgreSQL extension base on the Simple Chinese Word Segmentation(SCWS)
+ * [another Chinese parser](https://github.com/jaiminpan/pg_jieba) based on Jieba Chinese Word Segmentation
+ 
+Once you have the new search config , make sure you test it with the `pleroma` user in PostgreSQL (change `YOUR.CONFIG` to your real configuration name)
+```
+SELECT ts_debug('YOUR.CONFIG', '安装和配置Nginx, ElixirとErlangをインストールします');
+```
+Check output of the query, and see if it matches your expectation.
+
+
+## Update text search config and index in database
+
+=== "OTP"
+
+    ```sh
+    ./bin/pleroma_ctl database set_text_search_config YOUR.CONFIG
+    ```
+
+=== "From Source"
+
+    ```sh
+    mix pleroma.database set_text_search_config YOUR.CONFIG
+    ```
+
+Note: index update may take a while.
+
+## Restart database connection
+Since some changes above will only apply with a new database connection, you will have to restart either Pleroma or PostgreSQL process, or use `pg_terminate_backend` SQL command without restarting either. 
+
+Now the search results of statuses should be much more friendly for your language of choice, the results for searching users and tags were not changed, as the default parsing/matching should work for most cases. 

lib/mix/tasks/pleroma/database.ex

@@ -167,4 +167,51 @@ defmodule Mix.Tasks.Pleroma.Database do
     end)
     |> Stream.run()
   end
+
+  def run(["set_text_search_config", tsconfig]) do
+    start_pleroma()
+    %{rows: [[tsc]]} = Ecto.Adapters.SQL.query!(Pleroma.Repo, "SHOW default_text_search_config;")
+    shell_info("Current default_text_search_config: #{tsc}")
+
+    %{rows: [[db]]} = Ecto.Adapters.SQL.query!(Pleroma.Repo, "SELECT current_database();")
+    shell_info("Update default_text_search_config: #{tsconfig}")
+
+    %{messages: msg} =
+      Ecto.Adapters.SQL.query!(
+        Pleroma.Repo,
+        "ALTER DATABASE #{db} SET default_text_search_config = '#{tsconfig}';"
+      )
+
+    # non-exist config will not raise excpetion but only give >0 messages
+    if length(msg) > 0 do
+      shell_info("Error: #{inspect(msg, pretty: true)}")
+    else
+      rum_enabled = Pleroma.Config.get([:database, :rum_enabled])
+      shell_info("Recreate index, RUM: #{rum_enabled}")
+
+      # Note SQL below needs to be kept up-to-date with latest GIN or RUM index definition in future
+      if rum_enabled do
+        Ecto.Adapters.SQL.query!(
+          Pleroma.Repo,
+          "CREATE OR REPLACE FUNCTION objects_fts_update() RETURNS trigger AS $$ BEGIN
+          new.fts_content := to_tsvector(new.data->>'content');
+          RETURN new;
+          END
+          $$ LANGUAGE plpgsql"
+        )
+
+        shell_info("Refresh RUM index")
+        Ecto.Adapters.SQL.query!(Pleroma.Repo, "UPDATE objects SET updated_at = NOW();")
+      else
+        Ecto.Adapters.SQL.query!(Pleroma.Repo, "DROP INDEX IF EXISTS objects_fts;")
+
+        Ecto.Adapters.SQL.query!(
+          Pleroma.Repo,
+          "CREATE INDEX objects_fts ON objects USING gin(to_tsvector('#{tsconfig}', data->>'content')); "
+        )
+      end
+
+      shell_info('Done.')
+    end
+  end
 end

lib/pleroma/activity/search.ex

@@ -64,7 +64,7 @@ defmodule Pleroma.Activity.Search do
     from([a, o] in q,
       where:
         fragment(
-          "to_tsvector('english', ?->>'content') @@ plainto_tsquery('english', ?)",
+          "to_tsvector(?->>'content') @@ plainto_tsquery(?)",
           o.data,
           ^search_query
         )
@@ -75,7 +75,7 @@ defmodule Pleroma.Activity.Search do
     from([a, o] in q,
       where:
         fragment(
-          "to_tsvector('english', ?->>'content') @@ websearch_to_tsquery('english', ?)",
+          "to_tsvector(?->>'content') @@ websearch_to_tsquery(?)",
           o.data,
           ^search_query
         )
@@ -86,7 +86,7 @@ defmodule Pleroma.Activity.Search do
     from([a, o] in q,
       where:
         fragment(
-          "? @@ plainto_tsquery('english', ?)",
+          "? @@ plainto_tsquery(?)",
           o.fts_content,
           ^search_query
         ),
@@ -98,7 +98,7 @@ defmodule Pleroma.Activity.Search do
     from([a, o] in q,
       where:
         fragment(
-          "? @@ websearch_to_tsquery('english', ?)",
+          "? @@ websearch_to_tsquery(?)",
           o.fts_content,
           ^search_query
         ),

priv/repo/migrations/20210121080964_add_default_text_search_config.exs

@@ -0,0 +1,11 @@
+defmodule Pleroma.Repo.Migrations.AddDefaultTextSearchConfig do
+  use Ecto.Migration
+
+  def change do
+    execute("DO $$
+    BEGIN
+    execute 'ALTER DATABASE '||current_database()||' SET default_text_search_config = ''english'' ';
+    END
+    $$;")
+  end
+end

priv/repo/optional_migrations/rum_indexing/20190510135645_add_fts_index_to_objects_two.exs

@@ -17,7 +17,7 @@ defmodule Pleroma.Repo.Migrations.AddFtsIndexToObjectsTwo do
 
     execute("CREATE FUNCTION objects_fts_update() RETURNS trigger AS $$
     begin
-    new.fts_content := to_tsvector('english', new.data->>'content');
+    new.fts_content := to_tsvector(new.data->>'content');
     return new;
     end
     $$ LANGUAGE plpgsql")