Method addBuildTextIndexesStep

src/Storages/MergeTree/MergeTask.cpp:2802–2876

}

void MergeTask::addBuildTextIndexesStep(QueryPlan & plan, const IMergeTreeDataPart & data_part, const GlobalRuntimeContextPtr & global_ctx)
{
    if (global_ctx->text_indexes_to_merge.empty())
        return;

    auto original_header = plan.getCurrentHeader();
    auto read_column_names = original_header->getNameSet();

    IndicesDescription description_to_build;
    std::vector<MergeTreeIndexPtr> indexes_to_build;
    const auto storage_columns = global_ctx->storage_columns.getNameSet();
    const auto virtual_columns = global_ctx->virtual_columns.getNameSet();

    for (const auto & index : global_ctx->text_indexes_to_merge)
    {
        auto required_columns = index.expression->getRequiredColumns();

        bool read_any_required_column = std::ranges::any_of(required_columns, [&](const auto & column_name)
        {
            return read_column_names.contains(getColumnNameInStorage(column_name, storage_columns, virtual_columns));
        });

        if (!read_any_required_column)
            continue;

        auto index_ptr = MergeTreeIndexFactory::instance().get(global_ctx->metadata_snapshot, index, *global_ctx->data_settings);

        /// Rebuild index if merge may reduce rows because we cannot adjust parts offsets in that case.
        /// Build index if it is not materialized in the data part.
        if (global_ctx->merge_may_reduce_rows || !index_ptr->getDeserializedFormat(data_part.checksums, index_ptr->getFileName(), &data_part.getDataPartStorage()))
        {
            description_to_build.push_back(index);
            indexes_to_build.push_back(std::move(index_ptr));
        }
    }

    if (indexes_to_build.empty())
        return;

    if (!global_ctx->temporary_text_index_storage)
    {
        auto new_part_path = global_ctx->new_data_part->getDataPartStorage().getRelativePath();
        global_ctx->temporary_text_index_storage = createTemporaryTextIndexStorage(global_ctx->disk, new_part_path);
    }

    addSkipIndexesExpressionSteps(plan, description_to_build, global_ctx);

    MergeTreeWriterSettings writer_settings(
        global_ctx->data->getContext()->getSettingsRef(),
        global_ctx->context->getWriteSettings(),
        global_ctx->data->getSettings(),
        global_ctx->new_data_part,
        global_ctx->new_data_part->index_granularity_info.mark_type.adaptive,
        /*rewrite_primary_key=*/ false,
        /*save_marks_in_cache=*/ false,
        /*save_primary_index_in_memory=*/ false,
        /*blocks_are_granules_size=*/ false);