Postgres 是否缓存了我们的查询，我们如何绕过它？

Question

我正在尝试运行以下 python3 代码：

import os
import psycopg2
import logging

# Set max attempts before giving up
MAX_ATTEMPTS = 5

# Set basic logging config to debug (i.e. log everything).
# By default, this will log te stdout (i.e. it will behave the same as print)
logging.basicConfig(level=logging.DEBUG)

# Grab DB url from env variable
database_url = os.environ.get('DATABASE_URL')

assert database_url is not None, 'DATABASE_URL env variable must be set to a postgres connection string.'

# Initiate psycopg2 and instantiate a cursor object
conn = psycopg2.connect(database_url)
cursor = conn.cursor()


# Define function to delete old records
def delete_old_records(cur):
    # execute a query to delete old records. We're going to refer to this as the "delete" command
    query = 'DELETE FROM my_table WHERE id NOT IN ( SELECT id FROM ( SELECT id FROM my_table ORDER BY id DESC LIMIT 1850 ) foo);'
    cur.execute(query)


# Set variables to keep track of loop
successful = False
attempts = 0

# While not successful and max attempts not reached
while not successful and attempts < MAX_ATTEMPTS:
    try:
        # Attempt to delete old records
        delete_old_records(cursor)
        # Set successful to True if no errors were encountered in the previous line
        successful = True
        # Log a message
        logging.info('Successfully truncated old records!')
    # If some psycopg2 error happens
    except psycopg2.Error as e:
        # Log the error
        logging.exception('Got exception when executing query')
        # Rollback the cursor and get ready to try again
        conn.rollback()
        # Increment attempts by 1
        attempts += 1

# If not able to perform operation after max attempts, log message to indicate failure
if not successful:
    logging.warning(f'Was not successfully able to truncate logs after {MAX_ATTEMPTS} retries. '
                    f'Check logs for traceback (console output by default).')

问题来了：

1. 代码执行成功且没有错误。但是，当我们在 postico（Mac 的 Postgres GUI）上运行以下命令（以下称为“count”命令）时：

   SELECT count(*) from my_table;
   

   我们得到 1860 而不是 1850（即行没有被删除）。

2. 在 psql 或 postico 中手动运行删除命令时，我们分别在 psql 或 postico 中运行 COUNT 命令时得到正确的结果。但是，在 ipython 中运行命令时，我们会得到不同的结果。

3. 当我在计算机 A 上的 ipython 上打开了与数据库的连接，我运行了删除命令，我在计算机 B 上打开另一个连接到 ipython 上的数据库并运行 count 命令，我看到了db 行数没有改变，即仍然是 1860，没有减少到 1850。

我怀疑缓存/记忆，但我不确定我的命令是否真的有效。 psycopg2、postico 或 postgres 本身是否有可能导致这种情况的东西？我们如何绕过它？我们在 postico 或 psycopg2/postgres 上看不到任何清晰的缓存。

Answer 1

不涉及缓存。 PostgreSQL 不缓存查询结果。

你只是忘记COMMIT删除事务，所以它的效果在任何并发事务中都不可见。