158 lines
5.6 KiB
Ruby
158 lines
5.6 KiB
Ruby
# frozen_string_literal: true
|
|
|
|
# This module adds PG full-text search capabilities to a model.
|
|
# A `search_data` association with a `search_vector` column is required.
|
|
#
|
|
# Declare the fields that will be part of the search vector with their
|
|
# corresponding weights. Possible values for weight are A, B, C, or D.
|
|
# For example:
|
|
#
|
|
# include PgFullTextSearchable
|
|
# pg_full_text_searchable columns: [{ name: 'title', weight: 'A' }, { name: 'description', weight: 'B' }]
|
|
#
|
|
# This module sets up an after_commit hook that updates the search data
|
|
# when the searchable columns are changed. You will need to implement the
|
|
# `#persist_pg_full_text_search_vector` method that does the actual insert or update.
|
|
#
|
|
# This also adds a `pg_full_text_search` scope so you can do:
|
|
#
|
|
# Model.pg_full_text_search("some search term")
|
|
|
|
module PgFullTextSearchable
|
|
extend ActiveSupport::Concern
|
|
|
|
LONG_WORDS_REGEX = %r([A-Za-z0-9+/@]{50,}).freeze
|
|
TSVECTOR_MAX_LENGTH = 1.megabyte.freeze
|
|
TEXT_SEARCH_DICTIONARY = 'english'
|
|
URL_SCHEME_REGEX = %r{(?<=\A|\W)\w+://(?=\w+)}.freeze
|
|
TSQUERY_DISALLOWED_CHARACTERS_REGEX = %r{[^a-zA-Z0-9 .@/\-_"]}.freeze
|
|
|
|
def update_search_data!
|
|
tsvector_sql_nodes = self.class.pg_full_text_searchable_columns.map do |column, weight|
|
|
tsvector_arel_node(column, weight)&.to_sql
|
|
end
|
|
|
|
persist_pg_full_text_search_vector(Arel.sql(tsvector_sql_nodes.compact.join(' || ')))
|
|
rescue ActiveRecord::StatementInvalid => e
|
|
raise unless e.cause.is_a?(PG::ProgramLimitExceeded) && e.message.include?('string is too long for tsvector')
|
|
|
|
Gitlab::AppJsonLogger.error(
|
|
message: 'Error updating search data: string is too long for tsvector',
|
|
class: self.class.name,
|
|
model_id: self.id
|
|
)
|
|
end
|
|
|
|
private
|
|
|
|
def persist_pg_full_text_search_vector(search_vector)
|
|
raise NotImplementedError
|
|
end
|
|
|
|
def tsvector_arel_node(column, weight)
|
|
return if self[column].blank?
|
|
|
|
column_text = self[column].gsub(LONG_WORDS_REGEX, ' ')
|
|
column_text = column_text[0..(TSVECTOR_MAX_LENGTH - 1)]
|
|
column_text = ActiveSupport::Inflector.transliterate(column_text)
|
|
|
|
Arel::Nodes::NamedFunction.new(
|
|
'setweight',
|
|
[
|
|
Arel::Nodes::NamedFunction.new(
|
|
'to_tsvector',
|
|
[Arel::Nodes.build_quoted(TEXT_SEARCH_DICTIONARY), Arel::Nodes.build_quoted(column_text)]
|
|
),
|
|
Arel::Nodes.build_quoted(weight)
|
|
]
|
|
)
|
|
end
|
|
|
|
included do
|
|
cattr_reader :pg_full_text_searchable_columns do
|
|
{}
|
|
end
|
|
end
|
|
|
|
class_methods do
|
|
def pg_full_text_searchable(columns:)
|
|
raise 'Full text search columns already defined!' if pg_full_text_searchable_columns.present?
|
|
|
|
columns.each do |column|
|
|
pg_full_text_searchable_columns[column[:name]] = column[:weight]
|
|
end
|
|
|
|
# When multiple updates are done in a transaction, `saved_changes` will only report the latest save
|
|
# and we may miss an update to the searchable columns.
|
|
# As a workaround, we set a dirty flag here and update the search data in `after_save_commit`.
|
|
after_save do
|
|
next unless pg_full_text_searchable_columns.keys.any? { |f| saved_changes.has_key?(f) }
|
|
|
|
@update_pg_full_text_search_data = true
|
|
end
|
|
|
|
# We update this outside the transaction because this could raise an error if the resulting tsvector
|
|
# is too long. When that happens, we still persist the create / update but the model will not have a
|
|
# search data record. This is fine in most cases because this is a very rare occurrence and only happens
|
|
# with strings that are most likely unsearchable anyway.
|
|
#
|
|
# We also do not want to use a subtransaction here due to: https://gitlab.com/groups/gitlab-org/-/epics/6540
|
|
after_save_commit do
|
|
update_search_data! if @update_pg_full_text_search_data
|
|
@update_pg_full_text_search_data = nil
|
|
end
|
|
end
|
|
|
|
def pg_full_text_search(query, matched_columns: [])
|
|
search_data_table = reflect_on_association(:search_data).klass.arel_table
|
|
|
|
joins(:search_data).where(
|
|
Arel::Nodes::InfixOperation.new(
|
|
'@@',
|
|
search_data_table[:search_vector],
|
|
Arel::Nodes::NamedFunction.new(
|
|
'to_tsquery',
|
|
[Arel::Nodes.build_quoted(TEXT_SEARCH_DICTIONARY), build_tsquery(query, matched_columns)]
|
|
)
|
|
)
|
|
)
|
|
end
|
|
|
|
private
|
|
|
|
def build_tsquery(query, matched_columns)
|
|
# URLs get broken up into separate words when : is removed below, so we just remove the whole scheme.
|
|
query = remove_url_scheme(query)
|
|
# Remove accents from search term to match indexed data
|
|
query = ActiveSupport::Inflector.transliterate(query)
|
|
# Prevent users from using tsquery operators that can cause syntax errors.
|
|
query = filter_allowed_characters(query)
|
|
|
|
weights = matched_columns.map do |column_name|
|
|
pg_full_text_searchable_columns[column_name]
|
|
end.compact.join
|
|
prefix_search_suffix = ":*#{weights}"
|
|
|
|
tsquery = Gitlab::SQL::Pattern.split_query_to_search_terms(query).map do |search_term|
|
|
case search_term
|
|
when /\A\d+\z/ # Handles https://gitlab.com/gitlab-org/gitlab/-/issues/375337
|
|
"(#{search_term + prefix_search_suffix} | -#{search_term + prefix_search_suffix})"
|
|
when /\s/
|
|
search_term.split.map { |t| "#{t}:#{weights}" }.join(' <-> ')
|
|
else
|
|
search_term + prefix_search_suffix
|
|
end
|
|
end.join(' & ')
|
|
|
|
Arel::Nodes.build_quoted(tsquery)
|
|
end
|
|
|
|
def remove_url_scheme(query)
|
|
query.gsub(URL_SCHEME_REGEX, '')
|
|
end
|
|
|
|
def filter_allowed_characters(query)
|
|
query.gsub(TSQUERY_DISALLOWED_CHARACTERS_REGEX, ' ')
|
|
end
|
|
end
|
|
end
|