django count(*) 慢查詢優化

　　分頁顯示是web開發常見需求，隨着表數據增加，200萬以上時，翻頁越到后面越慢，這個時候慢查詢成為一個痛點，關於count(*)慢的原因，簡單說會進行全表掃描，再排序，導致查詢變慢。這里介紹postgresql一種解決方案。對於大表，我們有時候並不需要返回精確的數值，可以采用模糊的總數代替。

　　原始語句

SELECT COUNT(*) AS "__count" FROM "my_table"

　　優化語句

SELECT reltuples::numeric FROM pg_class WHERE relname = table_name

 

介紹Django admin 分頁優化

　　

from django.contrib.auth.admin import UserAdmin
from django.core.paginator import Paginator, cached_property
from django.db import connections


class UserAdmin(UserAdmin):
    paginator = TablePaginator

class TablePaginator(Paginator):


    @cached_property
    def count(self):
        return (
            self._get_pgsql_estimate_count()
            if not self.object_list.query.where
            else super(LargeTablePaginator, self).count
        )

    def _get_pgsql_estimate_count(self):
        with connections["default"].cursor() as cursor:
            cursor.execute(
                "SELECT reltuples::numeric FROM pg_class WHERE relname = %s",
                [self.object_list.query.model._meta.db_table],
            )

            total_count = int(cursor.fetchone()[0])
        return total_count