对于不想保留的模式，可以将regexp_replace与捕获组一起使用，并在替换字符串中仅指定感兴趣的组。
请参阅Impala（impalad版本3.4.0）中的以下示例：

select
  addr_list,
  /*Concat is used just for visualization*/
  rtrim(ltrim(regexp_replace(addr_list,concat(
    /*Group of 169.254.*.* that should be excluded*/
    '(169\\.254\\.\\d{1,3}\\.\\d{1,3})', '|',
    /*Another group for 192.168.0.1*/
    '(192\.168\.0\.1)', '|',
    /*And the group that we need to keep*/
    '(\\d{1,3}\\.\\d{1,3}\\.\\d{1,3}\\.\\d{1,3})'

    /*So keep the third group in the output.
      Other groups will be replaced with empty string*/
  ), '\\3'), ','), ',') as ip_whitelist
from(values
 ('169.254.182.175,192.168.0.1,169.254.2.12,10.199.44.111,169.254.0.2' as addr_list),
 ('10.58.3.142,169.254.2.12'),
 ('192.168.0.1,192.100.0.2,154.16.171.3')
) as t

由于未知原因，e1d1e的工作方式不同，因为对于情况1和3，使用3作为返回组的同一regex根本不会返回任何内容。

select
  t.addr_list,
  rtrim(ltrim(regexp_replace(addr_list, r.regex, '\\3'), ','), ',') as ip_whitelist,
  regexp_extract(addr_list, r.regex, 3) as ip_wl_extract
from(values
  ('169.254.182.175,192.168.0.1,169.254.2.12,10.199.44.111,169.254.0.2' as addr_list),
  ('10.58.3.142,169.254.2.12'),
  ('192.168.0.1,192.100.0.2,154.16.171.3')
) as t
  cross join (
    select concat(
      '(169\\.254\\.\\d{1,3}\\.\\d{1,3})', '|',
      '(192\.168\.0\.1)', '|',
      '(\\d{1,3}\\.\\d{1,3}\\.\\d{1,3}\\.\\d{1,3})'
    ) as regex
  ) as r