Question

我有一个postgres查询，当我将其作为查询运行时，它可以正常工作。但是，我想将其转换为pl/r，并能够动态输入开始日期和结束日期。

有效的SQL是：

with date as (
select d as first_day,
d + interval '1 month' - interval '1 day' as last_day
from generate_series('2010-01-01'::date,
                     '2018-12-01'::date,
                     '1 month') as d
) select last_day::date as snapshot_date from date;

想做一个pl / r：

DROP FUNCTION IF EXISTS standard.seq_monthly(min_date_str char, max_date_str char);
CREATE FUNCTION standard.seq_monthly(min_date_str char, max_date_str char)
RETURNS setof dates AS
$$ 
with date as (
select d as first_day,
d + interval '1 month' - interval '1 day' as last_day
from generate_series(min_date_str::date,
                     max_date_str::date,
                     '1 month') as d
) select last_day::date as snapshot_date from date;

$$
LANGUAGE 'plr';
select * from standard.seq_monthly('2010-01-01' , '2018-12-01')

但是，运行该函数时出现错误。错误是

在 R解析错误<- 函数（min_date_str，max_date_str）

尝试将最大最小日期也声明为日期。

我们非常感谢您的帮助。

Answer 1

PL / R是PostgreSQL中的procedural language扩展名（类似于plpython，plperl，plphp），可以在其中运行有效的兼容R语言代码。您正在尝试无法仅在R会话中运行的SQL，因此您的代码将在PG plr存储的函数中失败。

但是，不需要扩展，因为可以使用非常基本的SQL语言（通常效率更高）来处理您的需求，以按指定的输入范围返回所需的日期范围表：

CREATE OR REPLACE FUNCTION seq_monthly(min_date_str char, max_date_str char)
RETURNS TABLE(snapshot_date date) AS
$$ 
     with mydate as (
          select d as first_day,
                 DATE_TRUNC('month', d) 
                     + '1 MONTH'::INTERVAL 
                     - '1 DAY'::INTERVAL as last_day
          from generate_series(min_date_str::date,
                               max_date_str::date,
                               '1 month') as d
     ) 

     select last_day::date as snapshot_date from mydate;

$$
LANGUAGE SQL STABLE;

select * from seq_monthly('2010-01-31' , '2018-12-31');

Rextester demo

现在，如果您真的想要一个plr存储的函数，请在给定的日期范围内使用R的seq()：

CREATE FUNCTION standard.seq_monthly(min_date_str char, max_date_str char)
RETURNS setof dates AS
$$ 
    seq(as.Date(min_date_str), as.Date(max_date_str), by='month')
$$
LANGUAGE 'plr';

select * from standard.seq_monthly('2010-01-01' , '2018-12-01')

Answer 2

一种解决方案是不通过plr，而是编写一个sql查询：

with max_min_date as(
select max(snapshot_date) as max_date_str, min(snapshot_date) as min_date_str from data
) , 
ts as (
select d as first_day,
d + interval '1 month' - interval '1 day' as last_day
from generate_series((select min_date_str from max_min_date)::date,
                     (select max_date_str from max_min_date)::date,
                     '1 month') as d
) select last_day::date as snapshot_date from ts;

以防万一，您需要月末日期而不是月初：

with max_min_date as(
  with max_min_wrk as (
    select max(snapshot_date) as max_date_str, min(snapshot_date) as min_date_str from data
    ) select cast(date_trunc('month', max_date_str) as date) as max_date, cast(date_trunc('month', min_date_str) as date) as min_date from max_min_wrk
), 
ts as (
select d as first_day,
d + interval '1 month' - interval '1 day' as last_day
from generate_series((select min_date from max_min_date)::date,
                     (select max_date from max_min_date)::date,
                     '1 month') as d
) select last_day::date as snapshot_date from ts 
order by snapshot_date asc;

我正在尝试将Postgresql查询转换为plr函数

2 个答案: