akkoma/lib/mix/tasks/pleroma/search/meilisearch.ex
Norm d79c92f9c6
meilisearch: Move published date to lower priority
Currently, Akkoma sorts by published date first before everything else.
This however makes search results pretty bad since Meilisearch uses a
bucket sort algorithm in order of the ranking rules specified:
https://www.meilisearch.com/docs/learn/core_concepts/relevancy#behavior

Since the `published` attribute is a unix timestamp, the resulting
buckets are pretty small so the other rules essentially have little to
no effect on the rankings of search results.

This fixes that issue by moving the `published:desc` rule further down
so it still sorts by date, but only after considering everything else.

AFAIK attribute and sort doesn't really affect results for Akkoma since
the only attribute considered is the `content` attribute and the `sort`
parameter isn't used in Akkoma searches. Everything else is made to
match more closely to Meilisearch's defaults.
2023-08-11 11:07:14 -04:00

144 lines
3.8 KiB
Elixir

# Pleroma: A lightweight social networking server
# Copyright © 2017-2021 Pleroma Authors <https://pleroma.social/>
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Mix.Tasks.Pleroma.Search.Meilisearch do
require Pleroma.Constants
import Mix.Pleroma
import Ecto.Query
import Pleroma.Search.Meilisearch,
only: [meili_put: 2, meili_get: 1, meili_delete!: 1]
def run(["index"]) do
start_pleroma()
meili_version =
(
{:ok, result} = meili_get("/version")
result["pkgVersion"]
)
# The ranking rule syntax was changed but nothing about that is mentioned in the changelog
if not Version.match?(meili_version, ">= 0.25.0") do
raise "Meilisearch <0.24.0 not supported"
end
{:ok, _} =
meili_put(
"/indexes/objects/settings/ranking-rules",
[
"words",
"proximity",
"typo",
"exactness",
"attribute",
"published:desc",
"sort"
]
)
{:ok, _} =
meili_put(
"/indexes/objects/settings/searchable-attributes",
[
"content"
]
)
IO.puts("Created indices. Starting to insert posts.")
chunk_size = Pleroma.Config.get([Pleroma.Search.Meilisearch, :initial_indexing_chunk_size])
Pleroma.Repo.transaction(
fn ->
query =
from(Pleroma.Object,
# Only index public and unlisted posts which are notes and have some text
where:
fragment("data->>'type' = 'Note'") and
(fragment("data->'to' \\? ?", ^Pleroma.Constants.as_public()) or
fragment("data->'cc' \\? ?", ^Pleroma.Constants.as_public())),
order_by: [desc: fragment("data->'published'")]
)
count = query |> Pleroma.Repo.aggregate(:count, :data)
IO.puts("Entries to index: #{count}")
Pleroma.Repo.stream(
query,
timeout: :infinity
)
|> Stream.map(&Pleroma.Search.Meilisearch.object_to_search_data/1)
|> Stream.filter(fn o -> not is_nil(o) end)
|> Stream.chunk_every(chunk_size)
|> Stream.transform(0, fn objects, acc ->
new_acc = acc + Enum.count(objects)
# Reset to the beginning of the line and rewrite it
IO.write("\r")
IO.write("Indexed #{new_acc} entries")
{[objects], new_acc}
end)
|> Stream.each(fn objects ->
result =
meili_put(
"/indexes/objects/documents",
objects
)
with {:ok, res} <- result do
if not Map.has_key?(res, "indexUid") do
IO.puts("\nFailed to index: #{inspect(result)}")
end
else
e -> IO.puts("\nFailed to index due to network error: #{inspect(e)}")
end
end)
|> Stream.run()
end,
timeout: :infinity
)
IO.write("\n")
end
def run(["clear"]) do
start_pleroma()
meili_delete!("/indexes/objects/documents")
end
def run(["show-keys", master_key]) do
start_pleroma()
endpoint = Pleroma.Config.get([Pleroma.Search.Meilisearch, :url])
{:ok, result} =
Pleroma.HTTP.get(
Path.join(endpoint, "/keys"),
[{"Authorization", "Bearer #{master_key}"}]
)
decoded = Jason.decode!(result.body)
if decoded["results"] do
Enum.each(decoded["results"], fn %{"description" => desc, "key" => key} ->
IO.puts("#{desc}: #{key}")
end)
else
IO.puts("Error fetching the keys, check the master key is correct: #{inspect(decoded)}")
end
end
def run(["stats"]) do
start_pleroma()
{:ok, result} = meili_get("/indexes/objects/stats")
IO.puts("Number of entries: #{result["numberOfDocuments"]}")
IO.puts("Indexing? #{result["isIndexing"]}")
end
end